Power distribution network line fault detection method and system

Abstract

The invention discloses a power distribution network line fault detection method and system, and relates to the technical field of relay protection, and the method comprises the steps: collecting a scene image above a photovoltaic array in response to an operation state monitoring request of a photovoltaic grid-connected power distribution line, and building a static reference coordinate system; calculating a shadow coverage area change rate and a current drop amplitude according to the static reference coordinate system; establishing a mapping relationship between the shadow coverage area change rate and the current drop amplitude, marking the current moment as a current abrupt change event moment when the current drop amplitude exceeds a current drop threshold, and extracting a shadow coverage area change rate feature sequence corresponding to the current abrupt change event moment; and performing deviation verification on the shadow coverage area change rate characteristic sequence according to the current drop amplitude corresponding to the current abrupt change event moment, and determining a fault type. According to the invention, millisecond-level accurate identification and isolation of fault types are realized, and protection refusal caused by distributed photovoltaic boosting current is avoided.

Description

Technical Field

[0001] This application relates to the field of relay protection technology, and in particular to a method and system for detecting faults in distribution network lines. Background Technology

[0002] The distribution network is a crucial link in the power system, connecting the transmission network and the user side, and is responsible for distributing electrical energy from substations to various loads. With the widespread application of distributed photovoltaic (PV) power generation technology, a large number of PV power sources are connected to the end of the distribution network through inverters, transforming the traditional single-source radial structure of the distribution network into a multi-source network structure.

[0003] Currently, fault detection in distribution network lines mainly employs distance protection and overcurrent protection. Distance protection calculates the measured impedance by measuring the voltage and current at the protection installation point and compares it with a pre-set line impedance value. When the measured impedance falls within the operating range, it is determined to be a fault within that range. Overcurrent protection detects the current amplitude flowing through the protection device; when the current exceeds a set threshold, the protection is activated. The setting calculations for both protection methods are based on the assumption that the fault current flows unidirectionally from the substation side to the fault point, and the setting values ​​remain fixed after commissioning.

[0004] However, when distributed photovoltaic power sources are connected to the distribution network, the photovoltaic inverter will also inject short-circuit current into the fault point when a fault occurs. Moreover, the photovoltaic output current changes in real time with the light intensity. The current flowing through the protection device is the vector superposition of the main grid current and the photovoltaic current, which causes the measured impedance to deviate from the impedance value corresponding to the actual fault point, resulting in the protection device malfunctioning or failing to operate. Summary of the Invention

[0005] In view of the aforementioned problems, this application is hereby filed.

[0006] Therefore, this application provides a method and system for detecting faults in power distribution lines, which can solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: In a first aspect, this application provides a method for detecting faults in a power distribution network, comprising: in response to a request for monitoring the operating status of a photovoltaic grid-connected power distribution line, acquiring scene images above a photovoltaic array and establishing a static reference coordinate system; The rate of change of the shadow coverage area and the current drop amplitude are calculated based on the static reference coordinate system. Establish a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, mark the current moment as the moment of the current sudden event, and extract the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event. The deviation of the shadow coverage area change rate feature sequence is verified based on the current drop amplitude corresponding to the time of the current sudden event. The fault type is determined based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0008] Preferably, establishing a static reference coordinate system includes: An industrial camera is fixedly mounted on a bracket structure directly above the photovoltaic array, wherein the imaging field of view completely covers the surface of all photovoltaic modules, and the optical axis of the imaging device is perpendicular to the plane of the photovoltaic modules. The original image is converted to grayscale, and the edge pixel set of the grayscale image is extracted. The edge pixel set is subjected to straight line feature extraction processing to select two sets of main direction straight lines parallel to the photovoltaic module frame, and the intersection of the main direction straight lines is used as the image origin to construct a two-dimensional static reference coordinate system. Using the frame of a photovoltaic module of known size as a calibration object, the physical area corresponding to a unit pixel is calculated in the two-dimensional static reference coordinate system to generate a calibration mapping matrix.

[0009] Preferably, the linear feature extraction process includes: Perform a Hough line transform on the edge pixel set to generate a set of line parameters; Based on the standard directional characteristics of the photovoltaic module frame, two sets of main directional lines are selected. The first set of main directional lines corresponds to the horizontal direction, and the second set of main directional lines corresponds to the vertical direction. Calculate the coordinates of the intersection point of the two sets of principal direction lines, set the intersection point coordinates as the image origin, and define the coordinate axis direction with the principal direction lines to construct a two-dimensional static reference coordinate system.

[0010] Preferably, after generating the calibration mapping matrix, the following is also included: Define the fixed image region boundary for the connecting hardware in a static reference coordinate system; Perform local brightness enhancement processing on the fixed image region and detect whether there are isolated bright spots in the enhanced image whose brightness value is significantly higher than that of the background; When isolated bright spots exist, record their spatial location and duration characteristics.

[0011] Preferably, the calculation of the rate of change of shadow coverage area and the current drop amplitude includes: Perform pixel-by-pixel grayscale difference between the current frame image and the previous frame image, and set a difference threshold to separate the brightness change area caused by cloud shadow movement; Perform morphological closing operations on the difference results and fill the internal holes to obtain a complete binary image of the shaded connected region; Based on the connected component analysis algorithm, the absolute value of the pixel area of ​​the shadow region is calculated; Calculate the rate of change of shadow coverage area between adjacent frames based on the sequence of absolute values ​​of shadow area in consecutive frames; The system receives real-time photovoltaic grid-connected current data from the power distribution line protection device via a communication interface and calculates the current drop amplitude within a unit time window.

[0012] Preferably, the step of performing pixel-by-pixel grayscale difference between the current frame image and the previous frame image includes: Get the grayscale image of the current frame and the grayscale image of the previous frame, perform pixel-by-pixel absolute grayscale difference operation on the two frames, and generate a difference image; The ambient light intensity level is determined based on the automatic exposure parameters of the industrial camera, and a differential threshold is set according to the light intensity level. The difference image is compared with the difference threshold at the pixel level to generate a binary mask for the brightness change region.

[0013] Preferably, establishing the mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude includes: A linear mapping curve between the rate of change of shaded coverage area and the magnitude of current drop was fitted using historical fault-free operation data. When the current drop exceeds the current drop threshold, the current moment is marked as a current surge event; At the moment of the current change event, extract the feature sequence of the rate of change of the shadow coverage area.

[0014] Preferably, the deviation verification includes: Obtain the measured value of the rate of change of the shadow coverage area at the moment of the current sudden change event, and query the theoretical current drop amplitude corresponding to the measured value based on the linear mapping curve. Calculate the absolute deviation between the measured value of the rate of change of the shaded area and the theoretical current drop amplitude, and determine whether the absolute deviation exceeds the preset physical allowable deviation threshold.

[0015] Preferably, determining the fault type includes: Verify whether the spatial location of the high-brightness point falls within the fixed image area boundary of the connecting hardware, whether the duration exceeds the duration of three consecutive frames of the industrial camera, and whether the brightness value exceeds twice the average value of the background area; Calculate the correlation coefficient between the current drop amplitude sequence and the characteristic sequence of the rate of change of shadow coverage area within the current mutation window; When the spatial location of a high-brightness point falls within the boundary of the fixed image area of ​​the connecting hardware, lasts for more than three consecutive frames of the industrial camera, and has a brightness value that is more than twice the average value of the background area, it is determined to be an arc fault occurring at the connecting hardware. When the correlation coefficient is higher than the environmental interference judgment threshold, it is determined that the current change is caused by cloud shadow environmental interference. When the correlation coefficient is lower than the short-circuit fault judgment threshold and the absolute deviation exceeds the preset physical allowable deviation threshold, it is determined to be a real line short-circuit fault.

[0016] Secondly, this application also provides a power distribution network line fault detection system, including: a reference construction module, which, in response to a request for monitoring the operating status of a photovoltaic grid-connected power distribution line, acquires scene images above the photovoltaic array and establishes a static reference coordinate system; The feature extraction module calculates the rate of change of the shadow coverage area and the current drop amplitude based on the static reference coordinate system. The mapping modeling module establishes a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, the current moment is marked as the moment of the current sudden event, and the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event is extracted. The fault determination module verifies the deviation of the shadow coverage area change rate feature sequence based on the current drop amplitude corresponding to the current sudden event. It determines the fault type based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0017] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps: In response to the request for monitoring the operational status of the photovoltaic grid-connected distribution lines, scene images above the photovoltaic array are collected, and a static reference coordinate system is established; The rate of change of the shadow coverage area and the current drop amplitude are calculated based on the static reference coordinate system. Establish a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, mark the current moment as the moment of the current sudden event, and extract the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event. The deviation of the shadow coverage area change rate feature sequence is verified based on the current drop amplitude corresponding to the time of the current sudden event. The fault type is determined based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0018] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps: In response to the request for monitoring the operational status of the photovoltaic grid-connected distribution lines, scene images above the photovoltaic array are collected, and a static reference coordinate system is established; The rate of change of the shadow coverage area and the current drop amplitude are calculated based on the static reference coordinate system. Establish a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, mark the current moment as the moment of the current sudden event, and extract the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event. The deviation of the shadow coverage area change rate feature sequence is verified based on the current drop amplitude corresponding to the time of the current sudden event. The fault type is determined based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0019] Implementing this application will have the following beneficial effects: This application provides a method and system for detecting faults in power distribution network lines. 1. This invention simultaneously calculates the rate of change of shadow coverage area and the amplitude of current drop in a static reference coordinate system, and establishes a mapping relationship between the two, whereas existing technologies rely solely on changes in single current and voltage waveforms for discrimination. When cloud shadows move rapidly, this invention compares the measured value of the rate of change of shadow coverage area with the theoretical prediction value through a deviation verification mechanism, and combines the correlation coefficient analysis of the current drop amplitude sequence and the characteristic sequence of the rate of change of shadow coverage area to decouple environmental interference from actual faults at the physical level. This effectively distinguishes between random current fluctuations caused by light fluctuations and line short-circuit faults, significantly reducing the unnecessary tripping rate of protection devices and improving the power supply continuity of high-penetration photovoltaic distribution networks.

[0020] 2. This invention simultaneously activates an optical and electrical dual-channel verification mechanism at the moment of a sudden current change. It performs local brightness enhancement processing on the connection hardware area and extracts the spatial location, duration, and brightness intensity of high-brightness spots—features that existing technologies cannot identify arc spot characteristics. When a short-circuit fault occurs in the connection hardware, this invention forms a mutually exclusive judgment logic chain through triple physical condition verification and current-shadow correlation coefficient cross-verification, prioritizing the identification of the arc fault location. When cloud shadow interference occurs, the automatic blocking trip circuit maintains line operation. This achieves millisecond-level accurate identification and isolation of fault types, avoiding protection failure caused by distributed photovoltaic-assisted current. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is an overall flowchart of a fault detection method for power distribution lines involved in this application.

[0023] Figure 2 This is an application environment diagram of a power distribution line fault detection method involved in this application.

[0024] Figure 3 This is a schematic diagram of the overall structure of a fault detection method for power distribution lines involved in this application.

[0025] Figure 4 This is a computer device diagram of a method for detecting faults in a power distribution network line, which is the subject of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] In one exemplary embodiment, such as Figure 1 As shown, a method for detecting faults in power distribution lines is provided, including: S100: In response to the request for monitoring the operation status of the photovoltaic grid-connected distribution line, it acquires scene images above the photovoltaic array and establishes a static reference coordinate system.

[0028] It should be noted that in the photovoltaic power distribution network fault detection scenario, the operation status monitoring request refers to the fault detection trigger signal issued by the protection device, and the static reference coordinate system refers to the geometric reference frame built based on the edge of the photovoltaic module frame. Traditional fault detection schemes rely on fixed electrical parameter settings. Traditional schemes only analyze current and voltage waveforms when a fault occurs. Fluctuations in light intensity caused by rapid cloud movement can lead to random jitter in the output current of the photovoltaic inverter, causing irregular drift in the measured impedance. This scheme uses the operation status monitoring request as the trigger condition. When the industrial camera detects the request, it immediately acquires scene images, establishes a static reference coordinate system, and locks the geometric reference before the actual impact of cloud shadow interference on current measurement, ensuring the physical consistency of subsequent shadow coverage area calculations.

[0029] In some embodiments, step S100 includes steps S110-S140, as follows: S110: An industrial camera is fixedly mounted on a bracket structure directly above the photovoltaic array, wherein the imaging field of view completely covers the surface of all photovoltaic modules, and the optical axis of the imaging device is perpendicular to the plane of the photovoltaic modules.

[0030] It should be noted that traditional power distribution network fault detection solutions typically do not deploy visual sensors, and even if they do, the installation geometry is not constrained. Camera tilt or incomplete field-of-view coverage can lead to shadow projection distortion. This application fixes an industrial camera to a support structure directly above the photovoltaic array, ensuring the imaging field of view completely covers the entire surface of the photovoltaic modules with the optical axis perpendicular to the module plane. The shadow projection geometry is independent of electrical parameter fluctuations, and cloud shadow interference cannot alter the observed module frame reference frame in the image through current jitter.

[0031] S120: Perform grayscale processing on the acquired original image and extract the edge pixel set of the grayscale image.

[0032] It should be noted that traditional power distribution network fault detection schemes rely solely on current and voltage waveform analysis. Electrical signals experience random jitter when cloud shadows move rapidly, leading to distortion in the extraction of component frame edge features. This application performs grayscale processing on the original image and extracts the edge pixel set. The edge geometric features are independent of the photovoltaic inverter output current fluctuations, and cloud cover cannot alter the observed component frame edge morphology in the image through current changes.

[0033] It should be noted that traditional power distribution network fault detection schemes only analyze current and voltage time-series data. The rapid movement of cloud shadows causes fluctuations in photovoltaic output, rendering frame recognition completely ineffective. This application performs grayscale processing on the acquired raw images and uses the Canny edge detection operator to accurately extract high-intensity edge pixel sets. The geometric shape of the component frame is independent of electrical parameter disturbances, and cloud cover cannot tamper with the spatial distribution characteristics of the observed edge pixels in the image through random current jitter.

[0034] For example, in a photovoltaic power distribution network fault detection scenario, an industrial camera captures a scene image above the photovoltaic array, with a resolution of 1920×1080 pixels. The scene image is then converted to grayscale: the RGB three-channel image is converted to a single-channel grayscale image using a weighted formula, with weighting coefficients of 0.299 for the red channel, 0.587 for the green channel, and 0.114 for the blue channel. The Canny edge detection algorithm is applied to grayscale images: First, Gaussian filtering is performed to reduce noise interference. The convolution kernel size of the Gaussian filter is 5×5, and the standard deviation is set to 1.4. The Sobel operator is used to calculate the gradient value of each pixel in the horizontal and vertical directions. The gradient magnitude and gradient direction are calculated based on the horizontal and vertical gradient values. Non-maximum suppression is applied to the gradient magnitude, retaining only pixels with local maxima along the gradient direction. Dual threshold detection and edge connection are used, with a high threshold set to 100 and a low threshold set to 50. Pixels with gradient magnitudes higher than the high threshold are marked as strong edge pixels, and pixels with gradient magnitudes between the high and low thresholds are marked as weak edge pixels. Weak edge pixels adjacent to strong edge pixels are retained as valid edges, while other weak edge pixels are suppressed. Based on the Canny edge detection results, the high-contrast rectangular outline edges of the photovoltaic module frame, the rigid straight line edges of the support structure, and the irregular edges of the cloud projection are extracted. The edges of the photovoltaic module frame are represented as a closed set of straight line segments, and their spatial distribution is not affected by the current fluctuations of the photovoltaic inverter.

[0035] S130: Perform linear feature extraction processing on the edge pixel set, select two sets of main direction lines parallel to the photovoltaic module frame, and use the intersection of the main direction lines as the image origin to construct a two-dimensional static reference coordinate system.

[0036] In some embodiments, step S130 is specifically implemented by steps S131 to S133, as follows: Step S131: Perform Hough line transform on the edge pixel set to generate a set of line parameters.

[0037] The Hough linear transform converts the image space to the parameter space, which is then discretized. The rho resolution is set to 1 pixel, the theta resolution is set to 1 degree, and the accumulator threshold is set to 20% of the image diagonal length.

[0038] Understandably, the Hough linear transform identifies potential straight lines in an image through a parameter space voting mechanism. The rho parameter represents the distance of the straight line from the origin, the theta parameter represents the angle between the straight line normal vector and the horizontal axis, and the accumulator threshold filters weak response edges to ensure that only significant straight line features of the photovoltaic module frame are retained.

[0039] For example, in a photovoltaic power grid fault detection scenario, the grayscale image acquired by the industrial camera has a resolution of 1920×1080 pixels. The edge pixel set is processed by Hough transform: the parameter space rho range is set to -1200 pixels to 1200 pixels, and the theta range is set to 0 degrees to 180 degrees. The accumulator matrix is ​​initialized to a zero matrix. Each edge pixel is traversed, and its rho value at all theta angles is calculated. Voting is performed on the corresponding unit of the accumulator. The accumulator threshold is set to 216. (rho, theta) pairs with accumulator values ​​exceeding 216 are extracted as candidate line parameters. A total of 86 candidate lines are identified, including horizontal lines on the upper edge of the photovoltaic module, vertical lines on the side edges, and interference lines from the background support.

[0040] Step S132: Based on the standard directional characteristics of the photovoltaic module frame, select two sets of main directional line sets. The first set of main directional line sets corresponds to the horizontal direction, and the second set of main directional line sets corresponds to the vertical direction.

[0041] The standard orientation features of the photovoltaic module frame include a horizontal tolerance range of 0 degrees ± 5 degrees and a vertical tolerance range of 90 degrees ± 5 degrees. These standard orientation features are pre-stored in the memory.

[0042] Understandably, the direction angle theta of each candidate line is calculated. Lines with a direction angle theta within a tolerance range of 0 degrees ± 5 degrees are assigned to the first set of main direction lines, and lines with a direction angle theta within a tolerance range of 90 degrees ± 5 degrees are assigned to the second set of main direction lines. The two sets of main direction lines represent the horizontal and vertical border lines of the photovoltaic module frame, respectively, excluding background interference lines.

[0043] For example, the direction angle theta is calculated for the 86 candidate lines output in step S131. Among them, theta values ​​of 32 lines are in the range of 355 degrees to 5 degrees, and are assigned to the first set of main direction lines, representing the horizontal border of the photovoltaic module; theta values ​​of 28 lines are in the range of 85 degrees to 95 degrees, and are assigned to the second set of main direction lines, representing the vertical border of the photovoltaic module; the theta values ​​of the remaining 26 lines are outside the tolerance range and are discarded as background interference lines; the average rho parameter of the first set of lines is calculated to be 420 pixels, and the average rho parameter of the second set of lines is calculated to be 680 pixels.

[0044] Step S133: Calculate the coordinates of the intersection point of the two sets of principal direction lines, set the intersection point coordinates as the image origin, and define the coordinate axis direction with the principal direction lines to construct a two-dimensional static reference coordinate system.

[0045] Understandably, the intersection points of the first set of principal direction lines are fitted to the horizontal baseline using the least squares method, and the intersection points of the second set of principal direction lines are fitted to the vertical baseline using the least squares method. The coordinates of the intersection points of the two baselines are used as the origin of the image. The direction angles of the first set of principal direction lines define the positive x-axis, and the direction angles of the second set of principal direction lines define the positive y-axis. The origin and coordinate axis directions of the coordinate system remain fixed and are not affected by cloud shadow movement or current fluctuations.

[0046] For example, least squares fitting is performed on the 32 horizontal lines of the first set of main direction lines to obtain the horizontal baseline equation y=420; least squares fitting is performed on the 28 vertical lines of the second set of main direction lines to obtain the vertical baseline equation x=680; the coordinates of the intersection point (680, 420) of the horizontal and vertical baselines are set as the image origin; the positive x-axis is defined as the direction of the horizontal baseline extending from left to right, and the positive y-axis is defined as the direction of the vertical baseline extending from top to bottom. The constructed two-dimensional static reference coordinate system covers the entire photovoltaic array area, and the position of the origin of the coordinate system is offset by no more than 2 pixels in ten consecutive frames of images.

[0047] Preferably, step S130 separates the rigid geometric features of the component frame through Hough transform and direction filtering; step S130 determines the origin of the intersection point based on least squares fitting; the static reference coordinate system constructed in step S130 is independent of the photovoltaic inverter current disturbance; and step S130 provides a physically immutable spatial reference for the calculation of the shadow area.

[0048] S140: Using the photovoltaic module frame of known size as a calibration object, calculate the physical area corresponding to a unit pixel in the two-dimensional static reference coordinate system to generate a calibration mapping matrix.

[0049] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the calibration object refers to the standard physical dimensions of the photovoltaic module frame as recorded in the construction drawings, and the calibration mapping matrix refers to the linear transformation matrix from pixel coordinates to physical area. Traditional visual calibration schemes rely on placing a standard checkerboard calibration board on-site. The calibration board is affected by uneven illumination when cloud shadows move rapidly, leading to inaccurate calculation of the physical area per unit pixel. This application directly uses the known dimensions of the photovoltaic module frame as an inherent calibration benchmark, calculates the physical area corresponding to each unit pixel in a two-dimensional static reference coordinate system, and generates a calibration mapping matrix. The calibration process is independent of electrical parameter fluctuations, and cloud cover cannot tamper with the physical dimension mapping relationship observed in the image through random jitter of the photovoltaic inverter current.

[0050] In some embodiments, step S140 is specifically implemented by steps S141 to S143, as follows: Step S141: Extract the pixel coordinates of the four corner points of the photovoltaic module frame in the two-dimensional static reference coordinate system.

[0051] The four corner points are determined by the intersection of the first set of main direction lines and the second set of main direction lines. The first set of main direction lines defines the horizontal border boundary, and the second set of main direction lines defines the vertical border boundary.

[0052] It is understandable that the horizontal and vertical border boundaries intersect in the static reference coordinate system to form a rectangular area, and the coordinates of the four intersection points are the corner pixel coordinates of the photovoltaic module border. The corner coordinates are not affected by cloud shadow movement or current fluctuations.

[0053] For example, in the scenario of photovoltaic power distribution network fault detection, the origin of the two-dimensional static reference coordinate system is located at the center of the image (680, 420). The horizontal baseline equation fitted by the first set of principal direction lines is y=420, and the vertical baseline equation fitted by the second set of principal direction lines is x=680. The horizontal border boundary is defined by two parallel lines y=300 and y=540, and the vertical border boundary is defined by two parallel lines x=500 and x=860. The coordinates of the four corner points are calculated as (500, 300), (860, 300), (500, 540), and (860, 540), with a pixel width of 360 pixels and a pixel height of 240 pixels.

[0054] Step S142: Based on the known physical dimensions and corner coordinates, calculate the pixel-physical scaling factor in the x-axis and y-axis directions, and derive the physical area value corresponding to a unit pixel.

[0055] Among them, the known physical dimensions are the nominal width and nominal height of the photovoltaic module frame, and the nominal width and nominal height are pre-stored in the memory.

[0056] It is understandable that the x-axis scaling factor is equal to the nominal width divided by the pixel width, the y-axis scaling factor is equal to the nominal height divided by the pixel height, and the physical area value corresponding to a unit pixel is equal to the product of the x-axis scaling factor and the y-axis scaling factor. This area value represents the area equivalent of a single pixel in the physical world.

[0057] For example, the nominal width of the photovoltaic module frame is 1652 mm and the nominal height is 992 mm; a pixel width of 360 pixels corresponds to a nominal width of 1652 mm, and the x-axis scaling factor is 1652 / 360 = 4.589 mm per pixel; a pixel height of 240 pixels corresponds to a nominal height of 992 mm, and the y-axis scaling factor is 992 / 240 = 4.133 mm per pixel; the physical area value per pixel is 4.589 × 4.133 = 18.97 square millimeters per pixel.

[0058] Step S143: Based on the physical area value corresponding to a unit pixel, generate a calibration mapping matrix. The calibration mapping matrix is ​​used to convert the pixel area of ​​the shadow connected region into the actual physical area.

[0059] It is understandable that the calibration mapping matrix is ​​constructed as a diagonal matrix, with the main diagonal elements containing the x-axis scaling factor and the y-axis scaling factor. Matrix multiplication achieves a linear transformation from pixel coordinates to physical coordinates, and the physical area of ​​the shadow region is obtained by multiplying the pixel area by the physical area value corresponding to a unit pixel.

[0060] For example, the calibration mapping matrix M is defined as: ; The matrix element unit is millimeters per pixel; when the pixel area of ​​the shadow connected region is 10,000 pixels, its actual physical area is calculated as 10,000 × 18.97 = 189,700 square millimeters.

[0061] Preferably, step S140 uses the inherent dimensions of the photovoltaic module as a calibration benchmark, and step S140 independently calculates the pixel-physical mapping relationship in the static reference coordinate system. The calibration mapping matrix generated in step S140 is not affected by the random current fluctuations caused by cloud cover. Step S140 provides a physically traceable geometric basis for the accurate quantification of the rate of change of shadow coverage area.

[0062] Ideally, industrial cameras use a global shutter sensor, and the calibration process should be performed during a cloudless, clear period.

[0063] In some embodiments, after generating the calibration mapping matrix, this application further includes: Mesh texture analysis is performed on the surface area of ​​the photovoltaic module. When abnormal attachment areas are identified, the physical orientation attributes of the abnormal attachment areas are determined based on the geometric center pixel coordinates and calibration mapping matrix of the abnormal attachment areas.

[0064] When it is determined that the abnormal attachment area covers the busbar area, or that the abnormal attachment area has a pixel-level spatial intersection with the projection area of ​​the current sampling critical path, the shadow feature extraction process is frozen and a cleaning alarm is triggered.

[0065] It should be noted that abnormal deposits such as bird droppings, dust, or fallen leaves on the surface of photovoltaic modules can alter local reflection characteristics. Traditional fault detection schemes cannot distinguish between abnormal deposits and shadow areas when cloud shadows move rapidly, leading to inaccurate calculations of shadow coverage area. After generating the calibration mapping matrix, the surface of the photovoltaic module is meshed, dividing the module surface area into grid cells of fixed physical size. The grayscale standard deviation and edge density value are calculated for each grid cell. When a grid cell is determined to constitute an abnormal deposit area, both the grayscale standard deviation and the edge density value deviate from the allowable deviation range of the photovoltaic module's factory standard texture feature library, the boundary pixel set of this area is extracted. The horizontal center pixel position of the abnormal deposit area is obtained by averaging the horizontal coordinates of all pixels in the boundary pixel set, and the vertical center pixel position of the abnormal deposit area is obtained by averaging the vertical coordinates of all pixels in the boundary pixel set. The horizontal and vertical center pixel positions are combined to obtain the geometric center pixel coordinates of the abnormal deposit area. Based on the geometric center pixel coordinates and the calibration mapping matrix, the projection coordinates of the abnormal attachment area in physical space are calculated. Combined with the known dimensions of the photovoltaic module frame, the orientation attributes of the abnormal attachment are determined. The orientation attributes include the quadrant position of the abnormal attachment on the module surface and the physical offset from the module center.

[0066] Among them, the busbar area is the physical coverage area of ​​the conductive strip as clearly defined in the photovoltaic module manufacturing specifications, and its boundary coordinates are converted to a two-dimensional static reference coordinate system through the module design drawings; the current sampling critical path projection area is the mapping area on the module surface of the physical connection from the photovoltaic grid connection point to the combiner box, and the coordinates of this area are converted to a two-dimensional static reference coordinate system through the electrical design drawings.

[0067] Understandably, a Boolean intersection operation is performed between the pixel coordinate set of the abnormal attachment region and the pixel coordinate set of the busbar region. If the intersection is not empty, it is determined that the abnormal attachment region covers the busbar region. Similarly, a Boolean intersection operation is performed between the pixel coordinate set of the abnormal attachment region and the pixel coordinate set of the current sampling critical path projection region. If the intersection is not empty, it is determined that there is a pixel-level spatial intersection. When either the busbar region is covered or a pixel-level spatial intersection exists, the abnormal attachment contamination shadow detection benchmark is determined. The shadow coverage area change rate calculation process in step S200 is frozen, and a cleaning alarm command containing the physical coordinates of the abnormal attachment is sent to the maintenance terminal. When neither condition is met, the shadow feature extraction process in step S200 continues. This process ensures that physical contamination sources are isolated at the geometric morphology level before cloud shadow interference intervenes, maintaining the original purity of the "shadow coverage area change rate - current drop amplitude" mapping relationship.

[0068] S200: Under the static reference coordinate system, calculate the rate of change of the shadow coverage area and the characteristics of the current drop amplitude.

[0069] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the rate of change of shadow coverage area refers to the time-varying characteristic of the area caused by the movement of cloud projection on the surface of photovoltaic modules, and the current drop amplitude characteristic refers to the decrease in the instantaneous value of the photovoltaic grid-connected current relative to the steady-state reference. Traditional fault detection schemes rely solely on changes in current and voltage waveforms for judgment. Traditional schemes misjudge random current fluctuations caused by rapid cloud shadow movement as line faults, leading to unnecessary tripping of protection devices. This scheme simultaneously calculates the rate of change of shadow coverage area and the current drop amplitude characteristic in a static reference coordinate system, performing pixel-level spatiotemporal alignment between the visual dynamics of cloud shadow movement and electrical fluctuations. It establishes a correlation criterion between the two at the initial stage of the fault, ensuring the physical distinguishability between real short-circuit faults and environmental interference.

[0070] In some embodiments, step S200 includes steps S210-S260, as follows: S210: Perform pixel-by-pixel grayscale difference between the current frame image and the previous frame image, and set a difference threshold to separate the brightness change area caused by cloud shadow movement.

[0071] It should be noted that in the scenario of photovoltaic power distribution network fault detection, pixel-by-pixel grayscale difference refers to the calculation process of the absolute change in pixel brightness values ​​between adjacent image frames, and the brightness change area caused by cloud shadow movement refers to the area where the light intensity on the surface of the photovoltaic module changes instantaneously due to cloud movement. Traditional fault detection schemes directly use current surges as fault criteria. Traditional schemes cannot distinguish between light fluctuations and actual short circuits when clouds are moving rapidly. The moment the edge of the cloud shadow sweeps across the photovoltaic array, it causes a random drop in the inverter output current, resulting in false judgments by the protection device. This scheme performs strictly pixel-aligned grayscale difference calculations on the current frame and the previous frame image, and dynamically sets the difference threshold in combination with the ambient light intensity. Under a static reference coordinate system, it accurately isolates the visual features of cloud shadow movement, making the calculation process of the shadow coverage area change rate completely independent of electrical parameter fluctuations. This establishes a visual grating reference for subsequent fault judgment that is not affected by current jitter.

[0072] In some embodiments, step S210 is specifically implemented by steps S211 to S213, as follows: Step S211: Obtain the grayscale image of the current frame and the grayscale image of the previous frame, perform pixel-by-pixel absolute grayscale difference operation on the two frames, and generate a difference image.

[0073] The gray-level absolute difference operation is defined as calculating the gray-level value I of the current frame for each pixel position (x, y). t(x,y) Compared with the grayscale value I of the previous frame {t-1}(x,y) The absolute difference D(x,y) = |I t(x,y) -I{t-1}(x,y) |

[0074] It is understandable that the pixel values ​​of the difference image D(x,y) represent the intensity of local brightness changes per unit time. Regions with sudden brightness changes caused by cloud shadow movement show high amplitude responses in the difference image, while static backgrounds and regions with slow lighting changes show low amplitude responses.

[0075] For example, in the scenario of photovoltaic power distribution network fault detection, the industrial camera has a frame rate of 30 frames per second and a time interval of 33.3 milliseconds between the current frame and the previous frame. When performing differential operation on a grayscale image with a resolution of 1920×1080 pixels, the differential value of the cloud-free area on the surface of the photovoltaic module is concentrated in the gray level of 0 to 15, the differential value of the cloud-edge moving area reaches the gray level of 85 to 120, and the differential value of the fixed area of ​​the support structure is less than the gray level of 5.

[0076] Step S212: Determine the ambient light intensity level based on the automatic exposure parameters of the industrial camera, and set the differential threshold T according to the light intensity level.

[0077] The automatic exposure parameters include shutter speed and gain value. The light intensity level is divided into three categories: strong light, medium light, and weak light. The strong light level corresponds to a differential threshold of T=40, the medium light level corresponds to T=25, and the weak light level corresponds to T=15.

[0078] Understandably, the differential threshold T is dynamically adjusted with the ambient light intensity. In strong light conditions, the contrast of cloud shadows is high, so the threshold needs to be increased to suppress noise. In weak light conditions, the contrast of cloud shadows is low, so the threshold needs to be decreased to retain effective features and ensure the robustness of separating regions with varying brightness.

[0079] For example, an industrial camera reports a current shutter speed of 1 / 1000 second and a gain of 1.2, which is determined to be a strong light level; a differential threshold T=40 is set; when the shutter speed drops to 1 / 200 second and the gain increases to 3.5, it is determined to be a weak light level, and the differential threshold T=15 is automatically switched.

[0080] Step S213: Compare the difference image D(x,y) with the difference threshold T at the pixel level to generate a binary mask M(x,y) for the brightness change region.

[0081] The binary mask M(x,y) is defined as follows: M(x,y)=1 when D(x,y)≥T, otherwise M(x,y)=0; a mask value of 1 indicates that the pixel belongs to the brightness change area caused by cloud movement.

[0082] Understandably, the binary mask M(x,y) accurately separates the cloud shadow movement region from the static background, and the continuity of the mask boundary is not affected by the current fluctuations of the photovoltaic inverter, providing a clean input for subsequent shadow connectivity analysis.

[0083] For example, after applying a threshold of T=40 to the differential image and generating a binary mask, the effective area is statistically analyzed: a connected pixel block appears in the upper left region of the photovoltaic array, with an area of ​​15200 pixels, an irregular elliptical shape, and smooth, unbroken edges; the background support area is completely suppressed because the differential value is lower than the threshold, and there are no false responses in the mask.

[0084] Preferably, step S210 separates dynamic cloud shadow features through temporal differential separation, step S210 uses an illumination adaptive threshold to eliminate environmental interference, step S210 generates a binary mask independent of electrical parameter fluctuations, and step S210 provides a pixel-level accurate input reference for calculating the shadow coverage area change rate.

[0085] S220: Perform morphological closing operation on the difference result and fill the internal holes to obtain a complete binary image of the shaded connected region.

[0086] It should be noted that in the context of photovoltaic power grid fault detection, morphological closing operations refer to a pixel-level morphological operation sequence of expansion followed by erosion. Internal voids refer to uncovered pixel areas completely surrounded by shadow boundaries, and the binary image of the shadow connected domain refers to a single-valued shadow mask after geometric integrity restoration. Traditional fault detection schemes directly use the original difference image to calculate the shadow area. However, when clouds move rapidly, internal voids are created due to reflections from the component edges or bird droppings, resulting in a lower measured shadow coverage area than the actual physical projection. Pixel-level breaks at the cloud shadow edges further cause non-physical jumps in the area change rate calculation. This solution performs morphological closing operations on the difference results to eliminate edge noise, fills internal voids to ensure the topological continuity of the shadow area, and obtains a complete binary image of the shadow connected domain. This process is independent of photovoltaic inverter current phase disturbances, ensuring the physical accuracy of the shadow coverage area change rate calculation from a geometric morphological perspective, providing an immutable visual evidence chain for subsequent fault determination.

[0087] In some embodiments, step S220 is specifically implemented by steps S221 to S223, as follows: Step S221: Apply morphological closing operation to the binary image of the difference result. The closing operation includes two consecutive operations: dilation and erosion.

[0088] The expansion operation uses a cross-shaped structural element, the size of which is preset according to the diffusion characteristics of the cloud shadow edge; the erosion operation uses a cross-shaped structural element of the same size as the expansion operation.

[0089] Understandably, morphological closing operations eliminate minute breaks and noise spots at the edges of cloud shadows while maintaining the geometric continuity of the main shadow area, a process that does not rely on current measurement data.

[0090] For example, in a photovoltaic array image, there are multiple broken regions smaller than five pixels in the differential results; dilation is performed using a five-pixel cross-shaped structuring element, and the broken regions are connected; subsequent erosion operations restore the original shadow boundary contour, and edge burrs are reduced by 80%.

[0091] Step S222: Perform a hole-filling algorithm on the closing operation result to identify and fill completely closed internal void areas.

[0092] In this context, a hole is defined as a connected region with a mask value of 0 surrounded by a binary mask value of 1, and the filling process uses the scan line seed filling method.

[0093] Understandably, the internal voids in the cloud shadow projection caused by component frame obstruction or reflective spots are filled to ensure the continuity of the shadow area in physical space. This filling result is not affected by the phase change of the photovoltaic inverter current.

[0094] For example, after the closing operation, there are three internal holes in the image, with the largest hole area being 300 pixels. The scanline algorithm traverses from the top left corner of the image, identifies the hole area completely surrounded by the shadow boundary, and sets its mask value to 1 to form a closed area without holes.

[0095] Step S223: Mark the connected components of the filling result, separate multiple independent shadow regions, and output a complete binary image of the shadow connected components.

[0096] The connected domain is defined using an eight-neighbor domain connection rule, and each independent shaded region is assigned a unique label number.

[0097] It is understandable that the connected component label distinguishes the independent shadows formed by the projection of different cloud clusters. The boundary closed path of each shadow connected component strictly corresponds to the physical projection range, providing an independent unit for shadow area calculation.

[0098] For example, the filled image contains four independent shadow regions, and the eight-neighbor labeling algorithm assigns a label value of 1 to 4 to each region; the largest connected component covers the upper left region of the photovoltaic array, and its boundary pixel set forms a single closed contour path without branch breaks.

[0099] Preferably, step S220 eliminates cloud shadow edge noise through morphological closing operation, the hole filling in step S220 ensures the physical continuity of the shadow area, the binary image of the shadow connected region generated in step S220 is independent of electrical parameter disturbances, and step S220 provides a geometrically complete input basis for calculating the rate of change of shadow coverage area.

[0100] S230: Calculates the absolute value of the pixel area of ​​the shadow region based on the connected component analysis algorithm.

[0101] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the absolute value of the pixel area of ​​the shadow region refers to the total number of pixels in the complete shadow connected region after morphological closing operations and hole filling. Traditional fault detection schemes directly use the pixel accumulation value of the original difference image as the shadow area. Traditional schemes suffer from area measurement distortion due to shadow edge breakage or internal voids when clouds move rapidly, and random current fluctuations caused by cloud shadow fluctuations result in nonlinear deviations between the area calculation result and the physical projection. This scheme accurately calculates the absolute value of the pixel area of ​​the shadow region based on a connected component analysis algorithm. This calculation process is independent of the photovoltaic inverter current phase disturbance, and cloud shading cannot tamper with the observed shadow geometry in the image through changes in electrical parameters, providing a physically traceable quantitative benchmark for the mapping relationship between the shadow coverage area change rate and the current drop amplitude.

[0102] S240: Calculate the rate of change of shadow coverage area between adjacent frames based on the sequence of absolute values ​​of shadow area in consecutive frames.

[0103] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the rate of change of shadow coverage area refers to the relative intensity of change in the shadow area between adjacent image frames. Traditional fault detection schemes directly accumulate the pixel values ​​of the original differential image as the shadow area. However, when clouds move rapidly, reflections from component edges or bird droppings can cause distortion in area measurement, and broken cloud shadow edges can lead to non-physical jumps in the rate of change calculation. In this scheme, step S240 calculates the rate of change between adjacent frames based on a sequence of absolute shadow area values ​​from consecutive frames. This geometric feature is independent of photovoltaic inverter current phase disturbances, and cloud obstruction cannot tamper with the observed dynamic evolution of shadows in the image through changes in electrical parameters, providing a physically traceable visual benchmark for fault determination.

[0104] In some embodiments, step S240 is specifically implemented by steps S241 to S243, as follows: Step S241: Cache the absolute values ​​of the pixel area of ​​the shadow region of multiple consecutive historical frames to form a temporal data queue of area arranged in chronological order.

[0105] The number of frames is determined based on the frame rate of the industrial camera and the movement speed of the cloud shadow, and the area time-series data queue strictly follows the time sequence of image acquisition.

[0106] Understandably, the area time-series data queue fully records the dynamic evolution of cloud shadow coverage, and the content of this queue is not affected by changes in the phase of the photovoltaic inverter current.

[0107] For example, an industrial camera acquires images at a rate of thirty frames per second, caches the shadow area values ​​of the most recent ten frames, with each frame spaced thirty-three milliseconds apart, and the area time-series data queue contains area measurements at ten consecutive time points.

[0108] Step S242: Calculate the area difference between adjacent frames for the area time series data queue to obtain the increase or decrease of the shadow area per unit time.

[0109] The increase or decrease in area is obtained by subtracting the area value of the previous frame from the area value of the subsequent frame, and the positive or negative sign indicates the direction of cloud shadow expansion or contraction.

[0110] It is understandable that the area difference between adjacent frames eliminates static background interference and retains only the dynamic change characteristics caused by the movement of cloud shadows. These characteristics have a physically verifiable causal relationship with the current drop amplitude.

[0111] For example, the shadow area value in the fifth frame is 12,500 pixels, and in the sixth frame it is 13,800 pixels, with a calculated area increase or decrease of 1,300 pixels; the shadow area value in the seventh frame is 13,200 pixels, with a calculated area increase or decrease of 600 pixels, indicating that the shadow first expands and then contracts.

[0112] Step S243: Divide the area difference between adjacent frames by the absolute value of the shadow area pixel area of ​​the previous frame to generate a dimensionless relative change rate.

[0113] Among them, dimensionless processing eliminates the dimensional differences caused by photovoltaic arrays of different sizes, and the relative rate of change characterizes the intensity ratio of the change in shadow coverage.

[0114] Understandably, the relative rate of change is independent of the photovoltaic installed capacity, and cloud cover cannot alter this geometric feature through random current fluctuations.

[0115] For example, the shadow area value of the sixth frame is 13,800 pixels, and the area difference between the seventh frame and the sixth frame is -600 pixels, with a calculated relative change rate of -4.35%. This value remains stable in three consecutive frames, indicating that the cloud shadow moves out of the array area at a constant speed.

[0116] Preferably, step S240 extracts dynamic features of cloud shadows through temporal difference, the dimensionless processing of step S240 ensures cross-scene adaptability, the rate of change of shadow coverage area generated in step S240 is independent of electrical parameter fluctuations, and step S240 provides physically irrefutable visual evidence for fault determination.

[0117] S250: Receives photovoltaic grid-connected current data from the power distribution line protection device via the communication interface and calculates the current drop amplitude within a unit time window.

[0118] In some embodiments, step S250 is specifically implemented by steps S251 to S253: Step S251: Receive the instantaneous three-phase current data stream at the photovoltaic grid connection point through a standardized communication interface. The sampling frequency of the data stream is kept at a fixed multiple with the frame rate of the industrial camera.

[0119] The fixed multiple relationship is defined as the current sampling frequency being an integer multiple of the image frame rate, and this integer multiple value is preset according to the performance of the protection device.

[0120] Understandably, the fixed multiple relationship ensures that the current data and the image frame are strictly aligned on the physical time axis, and this alignment mechanism is not affected by cloud shadow movement.

[0121] For example, the industrial camera has a frame rate of 30 frames per second, and the current sampling frequency is set to 900 Hz, which is 30 times the image frame rate; each frame of the image corresponds to 30 current sampling points.

[0122] Step S252: Perform effective value calculation of sliding time window on the instantaneous value data stream of three-phase current. The width of the time window is equal to the time interval of single frame image acquisition.

[0123] The sliding time window moves continuously along the time axis, and each time window outputs an effective current value.

[0124] Understandably, the effective value calculation suppresses high-frequency electrical noise, and the time window width matches the image time granularity to ensure that the current characteristics are consistent with the time reference of the visual characteristics.

[0125] For example, the time interval for acquiring a single frame image is 33 milliseconds, and the width of the sliding time window is set to 33 milliseconds; the root mean square value is calculated for the 30 current sampling points within the time window to obtain the effective current value corresponding to that frame.

[0126] Step S253: Compare the current effective value in the current frame with the historical steady-state current effective value reference, and calculate the current drop amplitude.

[0127] The historical steady-state current RMS value benchmark is taken from the median of the current RMS value in a ten-second historical data window with no cloud shadows moving rapidly.

[0128] Understandably, the current drop amplitude characterizes the intensity of instantaneous disturbance. This calculation process is independent of the rate of change of shadow coverage area, and cloud fluctuations cannot simultaneously alter the electrical and visual dual-channel characteristics.

[0129] For example, the historical steady-state current RMS value is 215 amperes, the current RMS value in the current frame is 182 amperes, and the calculated current drop is 33 amperes; this value recovers after 20 milliseconds when the cloud shadow sweeps across.

[0130] Preferably, step S250 acquires the original current data through synchronous sampling, the steady-state reference of step S250 eliminates the influence of load fluctuations, the current drop amplitude calculated in step S250 is independent of visual feature disturbances, and step S250 provides a clean electrical input for the mapping relationship between the rate of change of shadow coverage area and the current drop amplitude.

[0131] S260: Records the synchronization alignment between the rate of change of shadow coverage area, the current drop amplitude, and the image frame timestamp.

[0132] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the synchronization alignment relationship refers to the precise binding of the rate of change of shadow coverage area, the magnitude of current drop, and the image frame timestamp on the physical time axis. Traditional fault detection schemes do not establish a time synchronization mechanism between visual features and electrical features. When clouds move rapidly, the asynchronous data acquisition in traditional schemes leads to inaccurate correlation between shadow dynamics and current fluctuations. The random current drop at the moment the cloud shadow edge sweeps across the photovoltaic array is misjudged as a short circuit fault. This scheme records the synchronization alignment relationship of the three, so that the rate of change of shadow coverage area and the magnitude of current drop strictly correspond at the millisecond time granularity. This synchronization process is independent of the current phase disturbance of the photovoltaic inverter. Cloud shading cannot tamper with the physical benchmark of time alignment through fluctuations in electrical parameters, providing a spatiotemporally consistent dual-channel evidence chain for subsequent fault determination.

[0133] S300: Establish the mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude, and extract the feature sequence of the rate of change of the shadow coverage area.

[0134] It should be noted that in the scenario of photovoltaic power distribution network fault detection, the mapping relationship between the rate of change of shadow coverage area and the amplitude of current drop refers to the physical correlation between visual changes caused by cloud shadow movement and electrical response. The characteristic sequence of the rate of change of shadow coverage area refers to the dynamic characteristics of shadows and abnormal light spots in the connection hardware area at the moment of current abrupt change. Traditional fault detection schemes do not establish a correlation criterion between visual and electrical features. Traditional schemes directly regard the random current drop caused by light fluctuations when clouds move rapidly as a fault signal, leading to false tripping of protection devices in non-fault states. This scheme establishes a mapping relationship between the rate of change of shadow coverage area and the amplitude of current drop. When a current abrupt event is triggered, the characteristics of shadow area change and the characteristics of light spots in the connection hardware are extracted. This mapping process is independent of photovoltaic inverter control strategy disturbances. Cloud fluctuations cannot simultaneously tamper with the physical consistency of visual and electrical dual-channel characteristics, providing irrefutable spatiotemporal aligned evidence for fault determination.

[0135] S310: A linear mapping curve between the rate of change of shaded coverage area and the current drop amplitude is fitted using historical fault-free operation data.

[0136] It should be noted that in the scenario of photovoltaic distribution network fault detection, historical fault-free operation data refers to continuous monitoring data during which no actual short-circuit events have been confirmed by operation and maintenance records. The linear mapping curve refers to the physical response function between the rate of change of shadow coverage area and the current drop amplitude. Traditional fault detection schemes do not establish a quantitative correlation between environmental interference and electrical response. When clouds move rapidly, traditional schemes, lacking a benchmark reference, misjudge random current drops as fault characteristics, causing protection devices to trip under non-fault conditions. This scheme uses historical fault-free operation data to fit a linear mapping curve, pre-constructing a physical response model of visual and electrical characteristics before cloud shadow fluctuations occur. This mapping relationship is independent of the photovoltaic inverter's low-voltage ride-through control strategy. Cloud shading cannot simultaneously alter the physical consistency of shadow geometry and current drop amplitude, providing a traceable environmental interference benchmark curve for fault determination.

[0137] S320: When the current drop amplitude exceeds the current drop threshold, mark the current moment as a current surge event.

[0138] It should be noted that in the scenario of photovoltaic distribution network fault detection, a current surge event refers to the instantaneous state in which the current drop amplitude at the photovoltaic grid connection point exceeds a preset physical benchmark. Traditional fault detection schemes directly use current surges as tripping criteria. Traditional schemes, when clouds move rapidly, can trigger false trips due to random current drops exceeding a fixed threshold. The second-level fluctuations of the cloud shadow edge sweeping across the photovoltaic array are mistakenly identified as short-circuit faults by the protection device. This scheme, however, only marks current surge events when the current drop amplitude exceeds the current drop threshold. This marking process is independent of the visual interference of cloud shadow movement. The current drop threshold is set based on historical steady-state load characteristics and is unaffected by light fluctuations. Cloud obstruction cannot tamper with the physical authenticity of the current sampling process through shadow projection, providing a clear event trigger benchmark for subsequent shadow feature comparison.

[0139] S330: Extract the feature sequence of the rate of change of shadow coverage area at the moment of the current sudden change event.

[0140] In some embodiments, step S330 is specifically implemented by steps S331 to S333, as follows: Step S331: Using the time marked by the sudden change in current as the center, extract a fixed time window forward and backward to form the feature sequence extraction range of the change rate of the shadow coverage area.

[0141] The fixed-duration time window width is preset based on the cloud shadow movement speed and the protection device action time limit, and the window range strictly covers the leading and lagging stages of the current surge event.

[0142] Understandably, the time window ensures the capture of the full dynamic characteristics of cloud shadow interference and actual faults, and this range is set independently of photovoltaic inverter control strategy disturbances.

[0143] Step S332: Within the time window, extract the shadow coverage area change rate value corresponding to each frame in the time sequence of the image frames to form a time sequence feature sequence.

[0144] Among them, the temporal feature sequence strictly follows the physical timeline order, and the number of sequence elements is equal to the total number of image frames contained in the window.

[0145] It is understandable that the time-series feature sequence fully records the dynamic evolution of cloud shadows, and this sequence is not affected by random fluctuations in the amplitude of current drops.

[0146] For example, the industrial camera has a frame rate of thirty frames per second, a five-hundred-millisecond window contains fifteen frames of images, and the temporal feature sequence consists of fifteen shadow coverage change rate values ​​in chronological order.

[0147] Step S333: Perform extreme value detection and change direction analysis on the time series feature sequence to extract the feature sequence of shadow coverage area change rate.

[0148] It is understandable that the characteristic sequence of the rate of change of shadow coverage area characterizes the typical dynamic pattern of cloud shadow interference. This parameter is independent of electrical parameter disturbances, and cloud shading cannot tamper with the geometric evolution characteristics through random current fluctuations.

[0149] Preferably, the time window of step S330 covers the entire event cycle, the time sequence feature sequence of step S330 is strictly aligned with the physical time axis, the key feature parameters extracted in step S330 are independent of electrical disturbances, and step S330 provides millisecond-level visual dynamic evidence for fault determination.

[0150] In some embodiments, after extracting the feature sequence of the rate of change of shadow coverage area, the method further includes: In a static reference coordinate system, a fixed image region for connecting hardware is defined, and local brightness enhancement processing is performed on the fixed image region.

[0151] The detection function checks whether there are isolated bright spots in the enhanced image whose brightness values ​​are significantly higher than the background.

[0152] If a bright spot is present, its spatial location and duration characteristics are recorded.

[0153] It should be noted that when a short-circuit fault occurs at the connection hardware in a photovoltaic distribution network, a momentary electric arc will be generated. Traditional fault detection schemes rely solely on current and voltage characteristics. These schemes cannot identify the arc spot under the interference of current amplified by distributed photovoltaic systems, leading to the protection device's failure to operate in response to the short-circuit fault at the connection hardware. After extracting the characteristic sequence of the shadow coverage area change rate, local brightness enhancement processing is performed on the connection hardware area. Isolated bright spots in the enhanced image represent the actual electric arc phenomenon. This optical feature is independent of the photovoltaic inverter current phase disturbance, and cloud cover cannot tamper with the spatial position and duration characteristics of the observed spot in the image through fluctuations in electrical parameters. When an isolated bright spot is confirmed, its spatial position is recorded to ensure that the arc occurs at the electrical connection point, and the duration characteristic is recorded to eliminate interference from flying insects or reflective light, providing physically irrefutable optical evidence for fault determination.

[0154] The fixed image area for connecting hardware is a rectangular area preset in the static reference coordinate system based on the photovoltaic design drawings. This area covers all physical connection points between the combiner box input terminals and the photovoltaic module junction box. The local brightness enhancement processing adopts an adaptive histogram equalization algorithm, and the enhancement processing only applies to pixels within the fixed image area. Isolated high-brightness points are defined as a set of connected pixels whose brightness value exceeds twice the average value of the background area and whose spatial size is smaller than the preset physical size. Spatial location refers to the pixel coordinates of the geometric center of the high-brightness point in the static reference coordinate system. The duration feature refers to the number of image frames in which the high-brightness point appears consecutively.

[0155] Understandably, the enhanced image is divided into a background region and a detection region. The average pixel brightness of the background region is calculated as a baseline value. Each pixel in the detection region is traversed, and when its brightness value exceeds twice the baseline value and belongs to the same connected component, it is marked as a candidate high-brightness point. The spatial size of the candidate high-brightness point is calculated; when the spatial size is smaller than the physical projection area of ​​a single photovoltaic cell, it is confirmed as an isolated high-brightness point. The existence status of isolated high-brightness points in consecutive image frames is recorded; when the number of consecutive frames exceeds three frames of the industrial camera, it is confirmed as a valid arc spot. When a valid arc spot is determined to exist, its spatial location and duration features are output to the fault determination module; if no isolated high-brightness point is detected or the duration is insufficient, no spot features are output. This process ensures accurate capture of the optical characteristics of real short-circuit faults in current surge events, completely avoiding the risk of misjudgment caused by cloud shadow interference.

[0156] S400: The deviation of the shadow coverage area change rate feature sequence is verified based on the current drop amplitude corresponding to the current sudden event moment, and the fault type is determined based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0157] It should be noted that in the scenario of photovoltaic distribution network fault detection, deviation verification refers to the consistency test of the correlation between the current drop amplitude and the characteristic sequence of the shadow coverage area change rate on the physical time axis. The spatial location and duration characteristics of the high-brightness point refer to the geometric and temporal attributes of the arc spot in the connecting hardware area, and the correlation coefficient refers to the quantitative index of the temporal synchronization between the current drop amplitude sequence and the characteristic sequence of the shadow coverage area change rate. Traditional fault detection schemes rely solely on a single current mutation criterion. When clouds move rapidly, traditional schemes cannot distinguish between cloud shadow interference and actual short circuits, leading to false tripping or failure to trip of protection devices. Distributed photovoltaic power generation further obscures fault characteristics due to the increased current. This scheme performs deviation verification on the characteristic sequence of the shadow coverage area change rate based on the current drop amplitude corresponding to the current mutation event. It determines the fault type based on the spatial location and duration characteristics of the high-brightness point and the correlation coefficient between the current drop amplitude sequence and the characteristic sequence of the shadow coverage area change rate. The ternary criterion strictly decouples environmental interference from actual faults at the physical level. Cloud obstruction cannot simultaneously alter the physical consistency of optical, electrical, and visual characteristics, providing protection devices with a reliable fault determination basis at the millisecond level.

[0158] S410: Based on the current drop amplitude corresponding to the moment of the current sudden event, perform deviation verification on the characteristic sequence of the rate of change of the shadow coverage area.

[0159] In some embodiments, step S410 is specifically implemented by steps S411 to S412: Step S411: Obtain the measured value of the rate of change of the shadow coverage area corresponding to the moment of the current sudden change event, and query the theoretical current drop amplitude corresponding to the measured value according to the linear mapping curve.

[0160] Among them, the measured value of the rate of change of the shaded area is the sampled value in the characteristic sequence of the rate of change of the shaded area at the moment of the current sudden event, and the theoretical current drop amplitude is the current prediction value on the linear mapping curve corresponding to the measured value of the rate of change of the shaded area.

[0161] Understandably, the process begins by receiving the time stamp of a current surge event reported by the power distribution line protection device. Based on this time stamp, the corresponding frame number is located in the cached feature sequence of shadow coverage area change rate. The shadow coverage area change rate value corresponding to this frame number is then extracted as the measured value. The linear mapping curve is obtained by fitting historical fault-free operating data. The curve data is stored in memory in the form of a lookup table. The theoretical current drop amplitude corresponding to the measured value of the shadow coverage area change rate is obtained by looking up the table. The horizontal axis of the linear mapping curve represents the shadow coverage area change rate, and the vertical axis represents the theoretical current drop amplitude. This curve characterizes the normal current fluctuation pattern caused by cloud shadow movement.

[0162] For example, in a photovoltaic power distribution network fault detection scenario, the industrial camera operates at a frame rate of 30 frames per second, and the power distribution line protection device reports the moment of a current surge event via the IEC 61850 communication interface. The image frame number corresponding to the moment of the current surge event is located as frame 15 in the two-dimensional static reference coordinate system. The measured value of the shadow coverage area change rate corresponding to this frame is extracted from the shadow coverage area change rate feature sequence buffer as -4.35%. The stored linear mapping curve lookup table contains one hundred data points, with the horizontal axis ranging from -10% to +10% and the vertical axis ranging from 0 to 50 amperes. By locating the vertical axis value corresponding to -4.35% in the lookup table using linear interpolation, the theoretical current drop amplitude is obtained as 28 amperes.

[0163] Step S412: Calculate the absolute deviation between the measured value of the rate of change of the shaded area and the theoretical current drop amplitude, and determine whether the absolute deviation exceeds the preset physical allowable deviation threshold.

[0164] Among them, the absolute deviation is the absolute value of the difference between the measured value of the rate of change of the shaded area and the theoretical current drop amplitude, and the preset physical allowable deviation threshold is the pre-set upper limit of the current deviation.

[0165] Understandably, the absolute value of the difference between the measured rate of change of the shadow coverage area at the moment of the current surge event received from the power distribution line protection device and the theoretical current drop amplitude obtained in step S411 is calculated as the absolute deviation. The preset physical allowable deviation threshold is set as a percentage of the rated current of the photovoltaic array, determined based on the current-illuminance characteristic curve provided by the photovoltaic module manufacturer. The absolute deviation is compared with the preset physical allowable deviation threshold. When the absolute deviation exceeds the preset physical allowable deviation threshold, it is determined that the current surge exceeds the normal fluctuation range of cloud shadow interference, and a real short-circuit fault may exist.

[0166] For example, the measured rate of change of shadow coverage area at the moment of a current surge event received from the power distribution line protection device was 33 amperes. The absolute value of the difference between this and the theoretical value of 28 amperes was 5 amperes, which was taken as the absolute deviation. The rated current of the photovoltaic array is 300 amperes, and the preset physical allowable deviation threshold is set at 5% of the rated current, i.e., 15 amperes. Comparing 5 amperes and 15 amperes, the absolute deviation did not exceed the preset physical allowable deviation threshold, and it was determined that the current surge conformed to the physical laws of cloud shadow interference. In another test, the measured rate of change of shadow coverage area at the moment of the current surge event was positive 0.2%, corresponding to a theoretical current drop of 2 amperes. The measured rate of change of shadow coverage area was 45 amperes, and the absolute deviation was 43 amperes, exceeding the preset physical allowable deviation threshold of 15 amperes. It was determined that the current surge exceeded the normal range of cloud shadow interference.

[0167] It should be noted that step S411 converts visual features into predicted electrical features using a linear mapping curve, and step S412 separates environmental interference from actual faults through deviation verification. When the absolute deviation does not exceed the preset physical allowable deviation threshold, it indicates that the current surge conforms to the physical laws of cloud shadow movement and belongs to normal environmental interference; when the absolute deviation exceeds the preset physical allowable deviation threshold, it indicates that the current surge cannot be explained by cloud shadow movement and there may be an actual short-circuit fault. The preset physical allowable deviation threshold is set based on the physical characteristics of photovoltaic modules, rather than empirical statistical values, to ensure the physical traceability of the judgment benchmark.

[0168] Preferably, step S410 verifies the cause of the current surge through a physical model. Step S410 correlates visual and electrical characteristics at the physical level. The deviation verification mechanism constructed in step S410 is independent of photovoltaic inverter control strategy disturbances. Cloud cover cannot simultaneously alter the physical consistency between the rate of change of shadow coverage area and the current drop amplitude. This mechanism effectively isolates environmental interference in high-penetration photovoltaic scenarios, providing irrefutable fault judgment basis for protection devices.

[0169] S420: Based on the spatial location and duration characteristics of high-brightness points and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate characteristic sequence, the fault type is determined; In some embodiments, step S420 is specifically implemented by steps S421 to S425: Step S421: Verify whether the spatial location of the high-brightness spot falls within the fixed image area boundary of the connecting hardware, whether the duration exceeds the duration of three consecutive frames of the industrial camera, and whether the brightness value exceeds twice the average value of the background area.

[0170] Among them, the fixed image area boundary of the connecting hardware is a rectangular area boundary preset in the static reference coordinate system based on the photovoltaic design drawings; the duration of three consecutive frames of the industrial camera is three times the frame interval time of the industrial camera; and the average value of the background area is the average brightness value of non-high-brightness pixels in the fixed image area of ​​the connecting hardware.

[0171] Understandably, the process begins by loading preset coordinate parameters for the fixed image region of the connecting hardware into a static reference coordinate system. These parameters include the coordinates of the top-left corner (700, 430) and the bottom-right corner (750, 470). Thresholding is then performed on the locally enhanced image, marking pixels with brightness values ​​exceeding the threshold as candidate bright pixels. The geometric center coordinates of the candidate bright pixel set are calculated as the spatial location of the bright pixels. When these coordinates lie within a preset rectangular area, the spatial location condition is considered met. Simultaneously, the number of consecutive frames in which the bright pixels persist is counted. When the number of consecutive frames is greater than or equal to three, the duration condition is considered met. Finally, the average brightness of the background pixels within the fixed image region of the connecting hardware, excluding the bright pixels, is calculated. When the brightness value of the bright pixels exceeds twice this average, the brightness value condition is considered met.

[0172] Step S422: Calculate the correlation coefficient between the current drop amplitude sequence and the characteristic sequence of the rate of change of shadow coverage area within the current mutation window.

[0173] The current mutation window is a time interval extending 250 milliseconds before and after the current mutation event, and the correlation coefficient is the Pearson correlation coefficient, which is used to quantify the degree of linear correlation between the two sequences.

[0174] Understandably, the current surge window range is determined to be from T0-250ms to T0+250ms based on the current surge event time T0. The current drop amplitude corresponding to all frames within this window is extracted from the buffer, forming a current drop amplitude sequence. Simultaneously, the shadow coverage area change rate corresponding to all frames within this window is extracted, forming a shadow coverage area change rate feature sequence. Pearson correlation coefficient is calculated for the two sequences, yielding a correlation coefficient ranging from -1 to 1. A larger absolute value of the coefficient indicates a more consistent trend between the two sequences.

[0175] Step S423: When the spatial location of the high-brightness spot falls within the boundary of the fixed image area of ​​the connecting hardware, the duration exceeds the duration of three consecutive frames of the industrial camera, and the brightness value exceeds twice the average value of the background area, it is determined to be an arc fault occurring at the connecting hardware.

[0176] Understandably, the status of the three verification conditions is monitored in real time. When all three conditions are met simultaneously, an arc fault is immediately identified at the connection fitting, without waiting for the correlation coefficient calculation results. This determination has the highest priority; even if the correlation coefficient is higher than the environmental interference threshold, the arc fault is still confirmed first. After the arc fault is determined, an arc alarm command is generated, which includes the fault location coordinates, the time of occurrence, and the duration parameters.

[0177] For example, during a sudden current event, a high-brightness location (720, 450) was detected within the connection hardware area for four consecutive frames, with a brightness value of 220 exceeding twice the background average of 105. Although the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence was 0.88, which is higher than the environmental interference judgment threshold of 0.85, it was still preferentially identified as an arc fault occurring at the connection hardware, and fault coordinates with a positioning accuracy within ±5 mm were output.

[0178] Step S424: When the correlation coefficient is higher than the environmental interference judgment threshold, it is determined that the current change is caused by cloud shadow environmental interference.

[0179] Among them, the environmental interference judgment threshold is a preset correlation coefficient threshold, which represents the minimum correlation requirement for cloud shadow interference.

[0180] Understandably, when the three conditions verified in step S421 are not all met, and the correlation coefficient is higher than the environmental interference judgment threshold, the current change is determined to be caused by cloud shadow environmental interference. The environmental interference judgment threshold is adjusted periodically according to seasonal climate characteristics; the threshold value is set to 0.75 during the rainy season and 0.85 during the sunny season. After determining that it is cloud shadow environmental interference, an interference rejection command is generated. This command blocks the tripping circuit of the protection device to maintain the normal operation of the power distribution line.

[0181] Step S425: When the correlation coefficient is lower than the short-circuit fault judgment threshold and the absolute deviation exceeds the preset physical allowable deviation threshold, it is determined to be a real line short-circuit fault.

[0182] Among them, the short-circuit fault determination threshold is a preset correlation coefficient threshold, which represents the maximum correlation limit of a real short-circuit fault.

[0183] Understandably, a true line short-circuit fault is determined when the three conditions verified in step S421 are not all met, the correlation coefficient is lower than the short-circuit fault determination threshold, and the absolute deviation calculated in step S412 exceeds the preset physical allowable deviation threshold. The short-circuit fault determination threshold is dynamically calibrated based on the photovoltaic array's installed capacity; the threshold value is set to 0.4 for installed capacity less than one megawatt-hour and 0.3 for installed capacity greater than one megawatt-hour. After determining a true line short-circuit fault, a fault confirmation command is generated, which triggers the main protection to trip and disconnect the faulty section.

[0184] For example, the photovoltaic array has an installed capacity of 1.5 MW, and the short-circuit fault judgment threshold is set to 0.3. The correlation coefficient calculated at the moment of the current surge event is 0.25, which is lower than the threshold value of 0.3; the absolute deviation is 43 amperes, which exceeds the preset physical allowable deviation threshold of 15 amperes; the high-brightness verification result is that the brightness value does not exceed twice the background mean. If it is determined to be a real line short-circuit fault, a fault confirmation command is generated, and the protection device completes the tripping action within 20 milliseconds, with a fault location accuracy of ±10 meters.

[0185] It should be noted that step S421 verifies the authenticity of the arc spot through three physical conditions, step S422 quantifies the temporal correlation between electrical and visual features, and steps S423-S425 construct a mutually exclusive judgment logic chain. When all three conditions for the arc spot are met simultaneously, an arc fault is preferentially determined regardless of the correlation coefficient. When there is no valid arc spot, if the correlation coefficient is higher than the environmental interference judgment threshold, it is determined to be cloud shadow interference; if the correlation coefficient is lower than the short circuit fault judgment threshold and the deviation exceeds the standard, it is determined to be a real short circuit. The triple verification mechanism decouples environmental interference and real faults at the physical level, ensuring that the judgment result is not disturbed by the photovoltaic inverter control strategy.

[0186] Preferably, step S420 employs a spatial-temporal-intensity triple verification mechanism to identify arc faults. Step S420 uses correlation coefficient and deviation as dual parameters to determine metallic short circuits. The ternary criterion system constructed in step S420 is independent of visual interference from cloud shadow movement. When the cloud layer moves rapidly, the correlation coefficient automatically increases, avoiding false tripping of protection systems. When a real short circuit occurs, the current surge decouples from shadow changes, and the correlation coefficient significantly decreases. When an arc is generated at the connection hardware, the three conditions of high brightness are triggered simultaneously, ensuring accurate fault location. This mechanism achieves an industry breakthrough of zero false tripping and zero failure to trip in high-penetration photovoltaic scenarios.

[0187] S430: Based on the results of arc fault determination, cloud shadow environmental interference determination, or actual line short circuit fault determination, generate corresponding arc alarm commands, interference elimination commands, or fault confirmation commands, and output them to the power distribution line protection device.

[0188] It should be noted that in the photovoltaic distribution network fault detection scenario, the arc fault determination result refers to the conclusion that an arc fault occurs at the connection hardware; the cloud shadow environmental interference determination result refers to the conclusion that the current change is caused by cloud shadow environmental interference; the actual line short circuit fault determination result refers to the conclusion that an actual line short circuit fault occurs; the arc alarm command refers to the local isolation control command for arc faults; the interference elimination command refers to the control command that blocks the tripping circuit of the protection device; and the fault confirmation command refers to the control command that triggers the main protection tripping. Traditional fault detection schemes use a single tripping strategy. When clouds move rapidly, traditional schemes cannot distinguish the fault type, causing the protection device to perform unnecessary tripping operations for cloud shadow interference, or failing to accurately isolate arc faults at the connection hardware, thus expanding the power outage area. Based on the results of arc fault determination, cloud shadow environmental interference determination, or actual line short circuit fault determination, this solution generates corresponding arc alarm commands, interference elimination commands, or fault confirmation commands, and outputs them to the power distribution line protection device. The three types of commands strictly correspond to the three types of fault characteristics at the physical level. Cloud obstruction cannot tamper with the logical consistency of command generation through random current jitter, ensuring that the protection device can achieve the unity of accurate fault handling and power supply continuity in high-penetration photovoltaic scenarios.

[0189] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0190] Based on the same inventive concept, this application also provides a power distribution line fault detection system. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more power distribution line fault detection system embodiments provided below can be found in the limitations of the power distribution line fault detection method described above, and will not be repeated here.

[0191] In one exemplary embodiment, such as Figure 3 As shown, a power distribution network line fault detection system is provided, comprising: The benchmark construction module responds to the operation status monitoring request of the photovoltaic grid-connected distribution line by acquiring scene images above the photovoltaic array and establishing a static reference coordinate system; The feature extraction module calculates the rate of change of the shadow coverage area and the current drop amplitude based on the static reference coordinate system. The mapping modeling module establishes a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, the current moment is marked as the moment of the current sudden event, and the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event is extracted. The fault determination module verifies the deviation of the shadow coverage area change rate feature sequence based on the current drop amplitude corresponding to the current sudden event. It determines the fault type based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

[0192] Each module in the aforementioned power distribution line fault detection system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0193] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for detecting faults in power distribution lines. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0194] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0195] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0196] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0197] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0198] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0199] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0200] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0201] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting faults in power distribution network lines, characterized in that, include: In response to the request for monitoring the operational status of the photovoltaic grid-connected distribution lines, scene images above the photovoltaic array are collected, and a static reference coordinate system is established; The rate of change of the shadow coverage area and the current drop amplitude are calculated based on the static reference coordinate system. Establish a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, mark the current moment as the moment of the current sudden event, and extract the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event. The deviation of the shadow coverage area change rate feature sequence is verified based on the current drop amplitude corresponding to the current sudden event. The fault type is determined based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

2. The method for detecting faults in a power distribution network line as described in claim 1, characterized in that: The establishment of the static reference coordinate system includes: An industrial camera is fixedly mounted on a bracket structure directly above the photovoltaic array, wherein the imaging field of view completely covers the surface of all photovoltaic modules, and the optical axis of the imaging device is perpendicular to the plane of the photovoltaic modules. The original image is converted to grayscale, and the edge pixel set of the grayscale image is extracted. The edge pixel set is subjected to straight line feature extraction processing to obtain two sets of main direction straight lines parallel to the photovoltaic module frame, and the intersection of the main direction straight line sets is used as the image origin to construct a two-dimensional static reference coordinate system. Using the frame of a photovoltaic module of known size as a calibration object, the physical area corresponding to a unit pixel is calculated in the two-dimensional static reference coordinate system to generate a calibration mapping matrix.

3. The method for detecting faults in a power distribution network line as described in claim 2, characterized in that: The line feature extraction process includes: Perform a Hough line transform on the edge pixel set to generate a set of line parameters; Based on the standard directional characteristics of the photovoltaic module frame, two sets of main directional line sets are selected. The first set of main directional line sets corresponds to the horizontal direction, and the second set of main directional line sets corresponds to the vertical direction. Calculate the coordinates of the intersection point of the two sets of principal direction lines, set the intersection point coordinates as the image origin, and define the coordinate axis direction with the principal direction lines to construct a two-dimensional static reference coordinate system.

4. The method for detecting faults in a power distribution network line as described in claim 2, characterized in that: After generating the calibration mapping matrix, the following is also included: Define the fixed image region boundary for the connecting hardware in a static reference coordinate system; Perform local brightness enhancement processing on the fixed image region and detect whether there are isolated bright spots in the enhanced image whose brightness values ​​are significantly higher than those of the background; When an isolated bright spot exists, record the spatial location and duration characteristics of the isolated bright spot.

5. The method for detecting faults in a power distribution network as described in claim 1, characterized in that: The calculation of the rate of change of shadow coverage area and the current drop amplitude includes: Perform pixel-by-pixel grayscale difference between the current frame image and the previous frame image, and set a difference threshold to separate the brightness change area caused by cloud shadow movement; Perform morphological closing operations on the difference results and fill the internal holes to obtain a complete binary image of the shadow connected region, and calculate the absolute value of the pixel area of ​​the shadow region. Calculate the rate of change of shadow coverage area between adjacent frames based on the sequence of absolute values ​​of shadow area in consecutive frames; The system receives photovoltaic grid-connected current data from the power distribution line protection device via a communication interface and calculates the current drop amplitude within a unit time window.

6. The method for detecting faults in a power distribution network line as described in claim 5, characterized in that: The step of performing pixel-by-pixel grayscale difference between the current frame image and the previous frame image includes: Get the grayscale image of the current frame and the grayscale image of the previous frame, perform pixel-by-pixel absolute grayscale difference operation on the two frames, and generate a difference image; Determine the ambient light intensity level and set a differential threshold based on the light intensity; The difference image is compared with the difference threshold at the pixel level to generate a binary mask for the brightness change region.

7. The method for detecting faults in a power distribution network line as described in claim 1, characterized in that: The extracted feature sequence of the rate of change of shadow coverage area corresponding to the moment of the current abrupt change event includes: A linear mapping curve between the rate of change of shaded coverage area and the magnitude of current drop was fitted using historical fault-free operation data. When the current drop exceeds the current drop threshold, the current moment is marked as a current surge event; The feature sequence of the rate of change of shadow coverage area is extracted at each moment of the current mutation event.

8. The method for detecting faults in a power distribution network line as described in claim 7, characterized in that: The deviation verification includes: Obtain the measured value of the rate of change of the shadow coverage area corresponding to the moment of the current sudden change event, and query the theoretical current drop amplitude corresponding to the measured value of the rate of change of the shadow coverage area based on the linear mapping curve; Calculate the absolute deviation between the measured value of the rate of change of the shaded area and the theoretical current drop amplitude, and determine whether the absolute deviation exceeds the preset physical allowable deviation threshold.

9. The method for detecting faults in a power distribution network line as described in claim 7, characterized in that: The determination of the fault type includes: Verify whether the spatial location of the high-brightness point falls within the boundary of the fixed image area of ​​the connecting hardware, whether the duration exceeds the duration of three consecutive frames of the industrial camera, and whether the brightness value exceeds twice the average value of the background area; Calculate the correlation coefficient between the current drop amplitude sequence and the characteristic sequence of the rate of change of shadow coverage area within the current mutation window; When the spatial location of a high-brightness point falls within the boundary of the fixed image area of ​​the connecting hardware, lasts for more than three consecutive frames of the industrial camera, and has a brightness value that is more than twice the average value of the background area, it is determined that an arc fault has occurred at the connecting hardware. When the correlation coefficient is higher than the environmental interference judgment threshold, the current change is determined to be caused by cloud shadow environmental interference. When the correlation coefficient is lower than the short-circuit fault judgment threshold and the absolute deviation exceeds the preset physical allowable deviation threshold, it is determined to be a real line short-circuit fault.

10. A power distribution line fault detection system, employing the power distribution line fault detection method as described in any one of claims 1 to 9, characterized in that, include: The benchmark construction module responds to the operation status monitoring request of the photovoltaic grid-connected distribution line by acquiring scene images above the photovoltaic array and establishing a static reference coordinate system; The feature extraction module calculates the rate of change of the shadow coverage area and the current drop amplitude based on the static reference coordinate system. The mapping modeling module establishes a mapping relationship between the rate of change of the shadow coverage area and the current drop amplitude. When the current drop amplitude exceeds the current drop threshold, the current moment is marked as the moment of the current sudden event, and the feature sequence of the rate of change of the shadow coverage area corresponding to the moment of the current sudden event is extracted. The fault determination module verifies the deviation of the shadow coverage area change rate feature sequence based on the current drop amplitude corresponding to the moment of the current sudden change event, and determines the fault type based on the spatial location and duration characteristics of the high spot and the correlation coefficient between the current drop amplitude sequence and the shadow coverage area change rate feature sequence.

Images