CN114782442B - Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence - Google Patents
Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN114782442B CN114782442B CN202210707802.0A CN202210707802A CN114782442B CN 114782442 B CN114782442 B CN 114782442B CN 202210707802 A CN202210707802 A CN 202210707802A CN 114782442 B CN114782442 B CN 114782442B
- Authority
- CN
- China
- Prior art keywords
- dust
- image
- photovoltaic cell
- severity
- cluster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 24
- 239000000428 dust Substances 0.000 claims abstract description 275
- 238000004140 cleaning Methods 0.000 claims abstract description 41
- 238000001931 thermography Methods 0.000 claims abstract description 36
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 238000010586 diagram Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 150000001875 compounds Chemical class 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 241001270131 Agaricus moelleri Species 0.000 claims description 2
- 230000002776 aggregation Effects 0.000 claims description 2
- 238000004220 aggregation Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000009826 distribution Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 238000005286 illumination Methods 0.000 description 3
- 238000010191 image analysis Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000013210 evaluation model Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000017525 heat dissipation Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/92—Dynamic range modification of images or parts thereof based on global image properties
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention relates to a photovoltaic cell panel intelligent inspection method and a system based on artificial intelligence, which comprises the steps of obtaining a photovoltaic cell panel image and a corresponding thermal imaging image, carrying out image processing to obtain an orthographic image and an orthographic image of the photovoltaic cell panel, converting the orthographic image into a binary image, analyzing dust pixels in the image to obtain a convex hull area of each dust cluster, obtaining the dust severity according to the sum of the highest temperature of the thermal imaging image area corresponding to the convex hull area and the dust gray level of the dust cluster, obtaining the aggregate dust severity of each dust cluster and the global dust severity of the photovoltaic cell panel according to the minimum distance between the dust cluster and the dust cluster in the dust cluster range and the dust severity of the convex hull area of the dust cluster, comparing threshold values to judge whether cleaning is needed or not, obtaining the next inspection time through calculation, and based on artificial intelligence and image processing, compared with the traditional mode, the system is more accurate and intelligent, saves cost and reasonably applies resources.
Description
Technical Field
The application relates to the technical field of artificial intelligence and photovoltaic cleaning, in particular to a photovoltaic cell panel intelligent inspection method and system based on artificial intelligence.
Background
The output performance of photovoltaic cell panel receives the influence of deposit at surface dust, make photovoltaic cell's efficiency reduce, the dust has direct influence to the panel and receives solar irradiance and heat dissipation, and enable the panel surface to receive the corruption, there is great difference in the laying dust influence in different areas, photovoltaic power plant is being looked into by the dust, the influence is not only the generated energy, still influence photovoltaic power plant's life-span, current scheme adopts regular manual cleaning or robot clearance usually, the shortcoming is, the dust situation of every panel is usually not considered, direct according to presetting time or artificial detection, clear up one by one the panel, make the cleaning cost increase.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a photovoltaic cell panel intelligent inspection method and a photovoltaic cell panel intelligent inspection system based on artificial intelligence.
In order to achieve the purpose, the invention adopts the following technical scheme: a photovoltaic cell panel intelligent inspection method based on artificial intelligence comprises the following steps:
The above-mentioned
The middle inspection method comprises the following steps:
if the severity of the global dust of the photovoltaic panels in all the photovoltaic cell panels is larger than the threshold value, the photovoltaic cell panels are judged to need cleaning, a cleaning signal is sent to the robot, the robot cleans the photovoltaic cell panels, next inspection time is set after cleaning is finished, and if the severity of the global dust of the photovoltaic panels in all the photovoltaic cell panels is larger than the threshold value, the next inspection time is directly set.
The above-mentioned
The method of obtaining an orthoimage is as follows:
normalizing the collected image to obtain key points of the image, and labeling the key points to obtain label data;
inputting the acquired image and label data into a key point extraction network for training, and outputting a key point thermodynamic diagram;
and transforming the key points extracted from the key point thermodynamic diagram by utilizing perspective transformation to obtain an orthoimage.
The key point extraction network training comprises: thermodynamic diagram loss function
The expression is:
wherein (i, j) is the coordinate of the key point of the photovoltaic cell panel based on the image coordinate system,
the score of a photovoltaic panel keypoint representing category C at position (i, j),
heatmap representing a group route, N represents the number of key points in the group route,
、
is a hyper-parameter.
The above-mentioned
The middle dust cluster acquisition method comprises the following steps:
performing gray level conversion on the orthographic image of the photovoltaic cell panel to obtain a gray level image, and then performing threshold segmentation on the gray level image to obtain a binary image;
counting the coordinates of the dust pixels in the binary image, and carrying out the operation on the coordinates of the dust pixels
And clustering to obtain a plurality of dust clusters.
The above-mentioned
The method for obtaining the medium dust severity comprises the following steps:
acquiring a convex hull area of each dust cluster by using a convex hull algorithm;
obtaining the area of each convex hull area in the corresponding thermal imaging image, and carrying out temperature analysis on the area to obtain the highest temperature T of the area;
using the sum of the dust levels of each dust cluster
And obtaining the dust severity degree P of each dust cluster from the highest temperature T of the area corresponding to the convex hull area of each dust cluster, wherein the expression is as follows:
in the formula:
as the sum of the dust grayscales of each dust cluster,
is an indication of the temperature sensor of the current environment,
、
to adjust the factor.
The described
The medium aggregation dust severity is obtained as follows:
calculating the minimum distance between the dust pixel in each dust cluster and the dust pixel in other dust clusters
：
In the formula
Respectively representing sets of coordinates of dust pixels in two dust clusters,
representing the Euclidean distance between the two sets, and min representing the Euclidean distance between the minimum dust pixels in the sets;
the aggregate dust severity of the convex hull region of each dust cluster was calculated according to the following expression
：
Wherein P is the dust severity of the current dust cluster,
in order to set the minimum distance threshold value,
indicates a minimum distance from the current dust cluster
The number of dust clusters in the range is,
is shown as
Is arranged at
The dust severity of the dust clusters in the range,
is shown as
Is arranged at
The minimum distance of a dust cluster within the range from the current dust cluster.
The next inspection time obtaining method comprises the following steps:
after interval time inspection, a dust severity sequence can be obtained, and the next inspection time can be calculated by fitting a polynomial to the sequence, wherein the method comprises the following steps:
carrying out three-term polynomial fitting on the global dust severity sequence of the photovoltaic cell panel to obtain a fitting equation:
in the formula
Is the global dust severity of the photovoltaic panel,
to detect the time， 
B, c, d are coefficients;
will be provided with
=
Substituting to obtain day data
Then the inspection control is carried out by adopting a half-folding mode to obtain the next inspection time
Comprises the following steps:
in the formula (I), the compound is shown in the specification,
the time of the last round-trip is the time of the last round-trip,
representing an upward rounding function, while setting a cleaning threshold to
When it comes to
When the solar panel is cleaned, the photovoltaic cleaning robot is directly called at the moment T to clean the solar panel.
This technical scheme still provides a photovoltaic cell board wisdom system of patrolling and examining based on artificial intelligence, including image collector, image processor, data processor, control judgement ware, robot:
the image collector is used for shooting and collecting the images through a thermal imaging double-spectrum camera, and the collected images are photovoltaic cell panel images and thermal imaging images corresponding to the photovoltaic cell panel images;
the image processor is used for processing the photovoltaic cell panel image collected by the image collector to obtain an orthoimage of the photovoltaic cell panel, performing image enhancement and performing image conversion;
the data processor: performing data label training, pixel coordinate statistics, clustering operation, convex hull region calculation, convex hull region dust severity, aggregate dust severity, global dust severity calculation, and polynomial fitting calculation on next inspection time;
the control judger comprises a control module and a judgment module;
the robot is used for cleaning the corresponding photovoltaic cell panel after receiving the cleaning signal.
The control module and the judgment module are used for sending a signal to the image collector by the control module to collect a photovoltaic cell panel image, obtaining the global dust severity of the photovoltaic cell panel through the image processor and the data processor, comparing and judging the global dust severity and a threshold value by the judgment module, if the global dust severity is judged to be required to be cleaned, sending a signal to the robot by the controller for cleaning, if the global dust severity is judged not to be required to be cleaned, setting next inspection time, and when the next inspection time is reached, continuing sending the signal to the image collector by the control module for collection.
The invention has the beneficial effects that:
a severity evaluation model of dust coverage of the battery panel is established, and the dust distribution characteristics of the battery panel, the temperature characteristics of a dust coverage area and the distance between dust clusters can be effectively considered, so that the inspection equipment can be controlled more reasonably.
Drawings
FIG. 1 is a flow chart of a photovoltaic cell panel intelligent inspection method based on artificial intelligence;
FIG. 2 is a schematic diagram of a global dust severity sequence of a photovoltaic cell panel 1 in the intelligent routing inspection method for a photovoltaic cell panel based on artificial intelligence;
FIG. 3 is a schematic diagram of a global dust severity sequence of a photovoltaic cell panel 2 in the intelligent inspection method for photovoltaic cell panels based on artificial intelligence according to the present invention;
fig. 4 is a block diagram of the photovoltaic cell panel intelligent inspection system based on artificial intelligence.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application, as detailed in the appended claims.
Example 1:
the embodiment provides a photovoltaic cell panel intelligent inspection method based on artificial intelligence as shown in fig. 1, which includes:
the method comprises the steps of collecting images of the photovoltaic cell panels, obtaining an orthoimage and an orthothermal imaging image of each photovoltaic cell panel without a background through key point detection and perspective transformation, and has the advantages of unifying the standard of image analysis and enabling the detection accuracy to be higher.
Wherein, the step of obtaining the orthoimage is as follows:
firstly, a thermal imaging dual-spectrum camera collects a photovoltaic cell panel image and a corresponding thermal imaging image, the photovoltaic cell panel image is used as sample data of a training network, and the thermal imaging dual-spectrum camera is deployed on an unmanned aerial vehicle;
then, labeling image data of the photovoltaic cell panel, labeling x and y coordinates of key points of the photovoltaic cell panel based on an image coordinate system, wherein the x and y coordinates comprise four corner points of the photovoltaic cell panel, the category of the total key points is 4, and then convolving a labeled photovoltaic cell panel key point scatter diagram with a Gaussian kernel to obtain a key point thermodynamic diagram, namely, the number of label data thermodynamic diagrams generated by each image is 4, and finally connecting the label data thermodynamic diagrams together to form a 4-channel thermodynamic diagram, wherein the thermodynamic diagram has the characteristic that the pixel value of the thermodynamic diagram output by a network conforms to Gaussian distribution and the value range of the thermodynamic diagram is between [0 and 1 ];
further, the image data and the label data are sent to a key point extraction network for training, and the method comprises the following steps:
normalizing the image, namely changing the value range of the image matrix into a floating point number between [0 and 1] so as to facilitate the model to be better converged, and also performing normalization on the label;
the method comprises the steps that a photovoltaic cell panel key point Encoder and a photovoltaic cell panel key point Decoder are trained end to end through acquired images and labeled label data, the Encoder performs Feature extraction on the images, the input of the Encoder is normalized image data, the output of the Encoder is a Feature map, the Decoder performs upsampling and Feature extraction on the Feature map and finally generates a photovoltaic cell panel key point thermodynamic diagram, the input of the Decoder is a Feature map generated by an Encoder, the output of the Decoder is a photovoltaic cell panel key point thermodynamic diagram, the Encoder-Decoder is designed in many ways, the embodiment proposes that the Encoder and the Decoder are used for extracting by using a common pre-training backbone network, for example, the Encoder and the Decoder are used for extracting by using an Encoder and a Decoder in a matching manner, and the like
The method has the advantages that convergence of the network is facilitated, and after the final training is finished, the model compression and optimization acceleration technology is adopted to reduce the redundancy of network parameters and improve the calculation efficiency of the network.
When the training is carried out, the training device can be used,
in the formula (I), the compound is shown in the specification,
the photovoltaic panel keypoints representing category C score at location (i, j), the higher the score the more likely the keypoints of the photovoltaic panel.
Denotes Heatmap of ground truth. N represents the number of key points in the ground route.
、
The hyper-parameters need to be set manually.
Further, perspective transformation is performed by using the obtained key points of the photovoltaic cell panel, a homography matrix H needs to be introduced, specific operations can be estimated by a four-point method based on the key points of the corner points of the photovoltaic cell panel and four points of an orthographic image, it needs to be stated that four points of the orthographic image need to be determined according to the resolution of a camera, the four points determine the size of the image of the photovoltaic cell panel after perspective transformation, for example, when the original unmanned aerial vehicle is used for orthoscopic shooting, the size of the image is occupied by the photovoltaic cell panel in the obtained image (300 ), and then four points of the orthographic image need to select 4 corner points based on the size to estimate the homography matrix. Therefore, an orthographic image of each photovoltaic cell panel can be obtained through perspective transformation.
And (4) performing perspective transformation correction on the thermal imaging image of the photovoltaic cell panel to obtain an orthometric thermal imaging image of the photovoltaic cell panel.
Further, single-scale SSR image enhancement is carried out on the normal image to eliminate the problem of uneven illumination, and the single-scale SSR is single-scale
The algorithm, implementer can also adopt other white balance algorithms to eliminate the illumination influence, mainly reducing the influence caused by illumination to the follow-up twin network reasoning.
the purpose of the step is to
And converting the obtained orthographic image into dust pixels in a binary image analysis image, analyzing and processing the dust pixels to obtain dust clusters, and extracting the dust gray sum in a gray image corresponding to the dust clusters. It should be noted that when there is a dense dust cover on the part of the photovoltaic cell panel, it will cause the hot spot effect of the panel, so to avoid this, it is adopted
Intermediate ortho-thermal imaging image and
the gray scale map processing of (1).
The dust cluster acquisition method comprises the following steps:
the method comprises the steps of carrying out gray level conversion on an orthographic image of a photovoltaic cell panel to obtain a gray level image, then carrying out threshold segmentation on the gray level image to obtain a binary image a, setting a threshold value which needs to be manually debugged, and setting an empirical value of 100. In the binary image, 255 pixels are classified into one type, 0 pixels are classified into 1 type, and 255 pixels are classified into grid lines and dust;
and performing opening operation on the binary image a, eliminating the influence of the grid line to obtain a binary image b, and then counting the coordinates of dust pixels in the processed binary image b of the track camera, wherein the pixels with the pixel value of 255 are dust, namely counting the pixel coordinates with the pixel value of 255.
And performing DBSCAN clustering on the coordinates of the dust pixels, wherein in the embodiment, the parameter empirical value of the clustering is that the radius is 3, the threshold value of the core point is 5, and finally a plurality of dust clusters are obtained.
Wherein, the dust severity calculation steps are as follows:
each dust cluster represents a local dust set, and a convex hull region of each dust cluster is obtained by using a convex hull algorithm;
and extracting a dust gray sum G (each dust gray needs to be normalized and then the gray is summed, and the value of a visible dust pixel in a gray map is very close to white) of the original image for each dust cluster as a characteristic value for evaluating the dust depth, wherein the dust density of the convex hull area can be effectively reflected through the dust gray sum.
the purpose of this step is to analyze the convex hull region in the dust cluster to obtain the dust severity of the corresponding convex hull region of the dust cluster.
The method for acquiring the dust severity of the convex hull region comprises the following steps:
acquiring a thermal imaging image area corresponding to each convex hull area, and analyzing the temperature of the dust distribution area, wherein the highest temperature T of the thermal imaging image dust distribution area is directly obtained at the position because the thermal imaging can directly represent the temperature of the battery plate;
obtaining the dust severity P of each convex hull area of the photovoltaic cell panel:
in the formula (I), the compound is shown in the specification,
the reading of the temperature sensor in the current environment is avoided, the large formula error caused by weather and environment temperature is avoided,
、
to adjust the factors, so that the two terms of the evaluation model have different weights,
the depth characteristic value of the image dust distribution area (dust of the convex hull area),
、
the values in this embodiment are 0.1 and 0.005, respectively.
the purpose of this step is to carry out the analysis to the dust cluster condition that has the photovoltaic cell board in a plurality of convex closure regions, obtains corresponding aggregate dust severity and global dust severity, because when there are a plurality of dust clusters in photovoltaic cell board, causes great influence to photovoltaic cell board electricity generation.
The method for obtaining the severity of the aggregated dust comprises the following steps:
analyzing the closest distance between the dust clusters, namely acquiring the minimum distance between the dust pixel in each dust cluster and the dust pixel in another dust cluster
：
In the formula (I), the compound is shown in the specification,
and d represents the distance between the two sets, and in Euclidean distance representation, min represents the Euclidean distance between the smallest dust pixels in the sets, and the smaller the distance, the more possible two dust clusters are and the same.
Further, the severity of aggregated dust for each dust cluster was obtained as follows:
for each dust cluster, the minimum distance D between the dust cluster and other dust clusters is obtained, and then a minimum distance threshold G1 is set, where an empirical value is 20, that is, the aggregate dust severity O of each dust cluster can be obtained as the aggregate dust severity of the convex hull region corresponding to the dust cluster:
p is the dust severity of the current dust clusterDegree, n represents the number of dust clusters having a minimum distance from the current dust cluster within the range of G1,
indicating the dust severity of the ith dust cluster in the range of G1,
indicating the minimum distance of the ith dust cluster in the range of G1 from the current dust cluster.
So that
The larger the value of the term. The larger the final O value, the more likely the dust region is to merge with other regions and the dust coverage is more severe.
The O evaluation value considers the dust coverage of the convex hull area of the cell panel, the temperature information and the distance between dust clusters, has obvious characteristics, and can effectively reflect the influence of the dust coverage of the cell panel on the cell panel. Through calculating can be better wash photovoltaic cell board, avoid causing the panel to damage. The largest O in a photovoltaic panel is usually taken as the global dust severity M of the photovoltaic panel.
The special case is that when the photovoltaic cell panel only has 1 convex hull area, the global dust severity of the photovoltaic cell panel is represented by P, or when no convex hull area exists in the range of G1, the maximum value of the dust severity in all the convex hull areas is used as the global dust severity of the photovoltaic cell panel.
The method comprises the steps of judging whether cleaning is carried out or not according to the severity of global dust, and calculating next inspection time.
The inspection method comprises the following steps:
first, a cleaning threshold is set
，
The setting of (2) also needs manual debugging, the value of the embodiment is 10, and for the photovoltaic cell panel, when the global dust severity is greater than the threshold value, namely M>
During the process, a cleaning signal is sent to the photovoltaic cleaning robot to clean the photovoltaic panel, the signal contains information such as the position and the number of the photovoltaic cell panel and the position of the convex hull area of the photovoltaic cell panel, wherein the position of the convex hull area of the photovoltaic cell panel is the mass center of the convex hull area, and the signal can be directly obtained in an OpenCV (open cell vehicle vision system) library.
The method comprises the following specific steps:
firstly, uniformly cleaning all photovoltaic cell panels at one time, and taking the time as initial time;
setting intermittent inspection time, setting in the embodiment, performing inspection on the 5 th day, calling a robot to clean a photovoltaic cell panel if the severity of the global dust of a certain photovoltaic cell panel is larger than a threshold value, and waiting for 5 days to perform next inspection after the cleaning is finished; if the global dust severity of the photovoltaic cell panel is not found to be larger than the threshold value, waiting for 2 days, namely 7 days, and carrying out next inspection;
if the inspection on the 7 th day finds that the severity of the global dust of a certain photovoltaic cell panel is larger than the threshold value, a robot is called to clean the photovoltaic cell panel, and the next inspection is carried out after 5 days of cleaning; if the global dust severity of the photovoltaic cell panel is not found to be greater than the threshold, then a global dust severity sequence for time-series inspection of each photovoltaic cell panel can be obtained, as shown in fig. 2 and fig. 3, where the abscissa of the sequence is time and the ordinate of the sequence is dust severity, for example, in fig. 2, the sequences obtained after inspection of the photovoltaic cell panel 1 and the photovoltaic cell panel 2 in fig. 3 on the 5 th day and the 7 th day are (5, a), (7, b) and (5, a '), (7, b') are respectively fitted with a three-term polynomial to obtain a fitting equation:
then Y =
Substituting to obtain rough days data T1, and performing patrol control by using a halving method, wherein if the obtained rough days data is 18, the 18 th day can reach
Therefore, the next inspection time T is obtained as follows:
wherein T2 is the last time of polling
Day 7 after the initial clean-up. round represents an rounding-up function.
Setting a cleaning threshold value of
When it comes to
And (3) directly calling the photovoltaic cleaning robot to clean the battery panel at the time T, wherein the experience value of G2 is 2.
After the next inspection is carried out at the time T, if the severity of the global dust of a certain photovoltaic cell panel is found to be larger than a threshold value, a robot is called to clean the photovoltaic cell panel, and the next inspection is carried out after 5 days of cleaning; and if the severity of the global dust of the photovoltaic cell panel is not found to be larger than the threshold value, continuing to fit the polynomial fitting sequence to obtain next inspection time.
So circulate, until satisfying clean threshold value, begin to clean photovoltaic cell board, realize patrolling and examining photovoltaic cell board's intelligence.
Control module
Direction image collector
Transmit signal acquisitionImage processor for photovoltaic cell panel image
And a data processor
Obtaining the global dust severity of the photovoltaic cell panel, and judging the module
Comparing the global dust severity with a threshold value, judging that the robot needs cleaning if the robot needs cleaning, and sending the cleaning request to the controller
Sending a signal for cleaning, if the signal is judged not to need cleaning, setting next inspection time, and when the next inspection time is reached, continuing to send a signal to the image collector by the control module
Signal collection;
robot
The corresponding photovoltaic cell panel is cleaned after the cleaning signal is received.
The method has practical value in the application of routing inspection, operation and maintenance of the photovoltaic power station and the roof photovoltaic power station, compared with the traditional method without judgment, the high-cost cleaning mode of cleaning all the photovoltaic cell panels one by one is directly carried out on the cell panels according to the preset time or manual detection, and the method is based on artificial intelligence and image processing, accurately positions the photovoltaic cell panels needing to be cleaned, saves resources, and enables routing inspection equipment and cleaning resources to be utilized more reasonably.
The above embodiments are merely illustrative and should not be construed as limiting the scope of the invention, which is intended to be covered by the claims.
Claims (8)
1. A photovoltaic cell panel intelligent inspection method based on artificial intelligence is characterized by comprising the following steps:
the described
The method for obtaining the medium dust severity comprises the following steps:
acquiring a convex hull area of each dust cluster by using a convex hull algorithm;
obtaining the area of each convex hull area in the corresponding thermal imaging image, and carrying out temperature analysis on the area to obtain the highest temperature T of the area;
using the sum of the dust gray levels of each dust cluster
The highest temperature T of the region corresponding to the convex hull region of each dust cluster is obtainedDust severity P of the dust cluster, expressed as:
in the formula:
as the sum of the dust grayscales of each dust cluster,
is an indication of the temperature sensor of the current environment,
、
is an adjustment factor;
2. The intelligent inspection method for photovoltaic cell panels based on artificial intelligence as claimed in claim 1, wherein the method is characterized in that
The middle inspection method comprises the following steps:
if the severity of the global dust of the photovoltaic panels in all the photovoltaic cell panels is larger than the threshold value, the photovoltaic panels are judged to need cleaning, a cleaning signal is sent to the robot, the robot cleans the photovoltaic cell panels, next inspection time is set after cleaning is finished, and if the severity of the global dust of no photovoltaic panels in all the photovoltaic cell panels is larger than the threshold value, the next inspection time is directly set.
3. The intelligent inspection method for photovoltaic cell panels based on artificial intelligence as claimed in claim 1, wherein the method is characterized in that
The method for obtaining the orthoimage comprises the following steps:
normalizing the collected image to obtain key points of the image, and labeling the key points to obtain label data;
inputting the collected images and label data into a key point extraction network for training, and outputting a key point thermodynamic diagram;
and transforming the key points extracted from the key point thermodynamic diagram by utilizing perspective transformation to obtain an orthoimage.
4. The artificial intelligence based photovoltaic cell panel intelligent inspection method according to claim 3, wherein the key point extraction network training comprises: thermodynamic diagram loss function
The expression is:
wherein (i, j) is the coordinate of the key point of the photovoltaic cell panel based on the image coordinate system,
the score of a photovoltaic panel keypoint representing category C at position (i, j),
heatmap representing a group route, N represents the number of key points in the group route,
、
is a hyper-parameter.
5. The intelligent inspection method for photovoltaic cell panels based on artificial intelligence as claimed in claim 1, wherein the method is characterized in that
The middle dust cluster acquisition method comprises the following steps:
performing gray level conversion on the orthographic image of the photovoltaic cell panel to obtain a gray level image, and then performing threshold segmentation on the gray level image to obtain a binary image;
counting the coordinates of the dust pixels in the binary image, and performing the coordinate calculation on the dust pixels
And clustering to obtain a plurality of dust clusters.
6. The artificial intelligence based photovoltaic cell panel intelligent inspection method according to claim 1, wherein the method includes
The medium aggregation dust severity is obtained as follows:
calculating the minimum of the dust pixel in each dust cluster and the dust pixels in other dust clustersDistance between two adjacent plates
：
In the formula
Respectively representing sets of coordinates of dust pixels in two dust clusters,
representing the Euclidean distance between the two sets, and min representing the Euclidean distance between the minimum dust pixels in the sets;
the aggregate dust severity of the convex hull region of each dust cluster was calculated according to the following expression
：
Wherein P is the dust severity of the current dust cluster,
in order to set the minimum distance threshold value,
indicates a minimum distance from the current dust cluster
The number of dust clusters in the range is,
is shown as
Is arranged at
The dust severity of the dust clusters in the range,
is shown as
Is arranged at
The minimum distance of a dust cluster within the range from the current dust cluster.
7. The intelligent photovoltaic cell panel inspection method based on artificial intelligence according to claim 2, wherein the next inspection time obtaining method is as follows:
after interval time inspection, a dust severity sequence can be obtained, and the next inspection time can be calculated by fitting a polynomial to the sequence, wherein the method comprises the following steps:
carrying out three-term polynomial fitting on the global dust severity sequence of the photovoltaic cell panel to obtain a fitting equation:
where Y is the global dust severity of the photovoltaic panel,
in order to detect the time of day,
b, c, d are coefficients;
will be provided with
=
Substituting to obtain day data
Then the inspection control is carried out by adopting a half-folding mode to obtain the next inspection time
Comprises the following steps:
in the formula (I), the compound is shown in the specification,
the time of the last round-robin is the time of the round-robin,
represents an upward rounding function while setting a cleaning threshold to
When is coming into contact with
And when the solar panel is cleaned, the photovoltaic cleaning robot is directly called at the T moment.
8. The utility model provides a photovoltaic cell board wisdom system of patrolling and examining based on artificial intelligence which characterized in that, includes image collector, image processor, data processor, control judgement ware, robot:
the image collector is used for shooting and collecting the images through a thermal imaging double-spectrum camera, the collected images are used for obtaining photovoltaic cell panel images and corresponding thermal imaging images thereof, and the photovoltaic cell panel images and the corresponding thermal imaging images thereof are sent to the image processor;
the image processor receives and processes the photovoltaic cell panel image and the thermal imaging image sent by the image collector to obtain an ortho image of the photovoltaic cell panel and an ortho image of the thermal imaging image; converting an orthographic image of the photovoltaic cell panel into a gray-scale image, performing threshold segmentation on the gray-scale image to obtain a binary image, and sending the binary image to a data processor;
the data processor:
receiving the binary image sent by the image processor, clustering the coordinates of the dust pixels in the binary image to obtain a plurality of dust clusters, and extracting the dust gray sum in the gray image corresponding to the dust clusters;
acquiring a convex hull area of each dust cluster in a gray scale image, and obtaining the dust severity of the convex hull area according to the temperatures of the convex hull area in the dust cluster and the corresponding area of the convex hull area in a thermal imaging image;
obtaining the aggregated dust severity degree of each dust cluster according to the minimum distance between each dust cluster and the dust clusters in the range of the dust cluster and the dust severity degree of the convex hull area of the dust cluster, taking the aggregated dust severity degree of each dust cluster as the aggregated dust severity degree of the convex hull area corresponding to the dust cluster, selecting the maximum value of the aggregated dust severity degree of the convex hull area as the global dust severity degree of the photovoltaic cell panel, and sending the global dust severity degree of the photovoltaic cell panel to a control judger;
the control judger comprises a control module and a judgment module:
a judging module: receiving the global dust severity sent by the data processor, comparing and judging the global dust severity with a threshold, if the global dust severity is judged to need cleaning, sending a cleaning signal to the control module, if the global dust severity is judged not to need cleaning, setting next inspection time, and sending a collecting signal to the control module when the next inspection time is reached;
a control module:
when a cleaning signal sent by the judging module is received, sending a signal to the robot to control the robot to clean the corresponding photovoltaic cell panel;
when the acquisition signal sent by the judgment module is received, the signal acquisition photovoltaic cell panel image and the corresponding thermal imaging image are continuously sent to the image acquisition device;
the robot is used for cleaning the corresponding photovoltaic cell panel after receiving a signal sent by a control module of the control judger.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210707802.0A CN114782442B (en) | 2022-06-22 | 2022-06-22 | Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210707802.0A CN114782442B (en) | 2022-06-22 | 2022-06-22 | Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114782442A CN114782442A (en) | 2022-07-22 |
| CN114782442B true CN114782442B (en) | 2022-09-02 |
Family
ID=82420457
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210707802.0A Active CN114782442B (en) | 2022-06-22 | 2022-06-22 | Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114782442B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115035282B (en) * | 2022-08-12 | 2022-11-22 | 南通华豪巨电子科技有限公司 | High-quality photovoltaic cell panel orthographic image acquisition method and system based on track camera |
| CN115717890A (en) * | 2022-10-13 | 2023-02-28 | 上海中电电子系统科技股份有限公司 | Clean room cleanliness detection method based on robot inspection |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112184650A (en) * | 2020-09-21 | 2021-01-05 | 河南颂达信息技术有限公司 | Photovoltaic cell panel batch grid breakage detection method and system based on artificial intelligence |
| CN112329618A (en) * | 2020-11-04 | 2021-02-05 | 夏永霞 | Photovoltaic cleaning robot control method and system based on artificial intelligence |
-
2022
- 2022-06-22 CN CN202210707802.0A patent/CN114782442B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114782442A (en) | 2022-07-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114782442B (en) | Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence | |
| CN108961235B (en) | Defective insulator identification method based on YOLOv3 network and particle filter algorithm | |
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| Di Tommaso et al. | A multi-stage model based on YOLOv3 for defect detection in PV panels based on IR and visible imaging by unmanned aerial vehicle | |
| CN111080691A (en) | Infrared hot spot detection method and device for photovoltaic module | |
| CN110210060B (en) | Prediction method for surface area ash degree of solar photovoltaic panel | |
| CN112529875A (en) | Photovoltaic module glass burst early warning method and system based on artificial intelligence | |
| CN112871938B (en) | Photovoltaic power plant cleans control integration management system | |
| CN114677362B (en) | Surface defect detection method based on improved YOLOv5 | |
| CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
| CN115797813B (en) | Water environment pollution detection method based on aerial image | |
| CN111291814A (en) | Crack identification algorithm based on convolution neural network and information entropy data fusion strategy | |
| CN115861190A (en) | Comparison learning-based unsupervised defect detection method for photovoltaic module | |
| Prabhakaran et al. | Deep Learning-Based Model for Defect Detection and Localization on Photovoltaic Panels. | |
| CN114612406A (en) | Photovoltaic panel defect detection method based on visible light and infrared vision | |
| CN115115634A (en) | Photovoltaic array hot spot detection method based on infrared image | |
| CN112183403A (en) | Photovoltaic cleaning robot cleaning speed adjusting method based on computer vision | |
| CN116703932A (en) | CBAM-HRNet model wheat spike grain segmentation and counting method based on convolution attention mechanism | |
| CN113962973A (en) | Power transmission line unmanned aerial vehicle intelligent inspection system and method based on satellite technology | |
| CN116797944A (en) | Detection method and system for identifying cleanliness of photovoltaic panel based on unmanned aerial vehicle image | |
| CN117218457B (en) | Self-supervision industrial anomaly detection method based on double-layer two-dimensional normalized flow | |
| CN117269168B (en) | New energy automobile precision part surface defect detection device and detection method | |
| CN117593499A (en) | Hydropower station electromechanical equipment fault identification method based on distributed routing inspection strategy | |
| CN115953678A (en) | Pavement damage detection method based on local gray extreme point and feature fusion | |
| CN113284103B (en) | Substation equipment defect online detection method based on space transformation fast R-CNN model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230419 Address after: No. 1 Kaixuan Road, Qidong Economic Development Zone, Nantong City, Jiangsu Province, 226000 Patentee after: QIDONG CHUANGLYU GREENING ENGINEERING Co.,Ltd. Address before: 226000 No. 1000, Xinhong Road, Huilong Town, Qidong City, Nantong City, Jiangsu Province Patentee before: TUOLUNSI SEMICONDUCTOR EQUIPMENT QIDONG Co.,Ltd. |