CN114937001B - Dust early warning method based on infrared detection - Google Patents
Dust early warning method based on infrared detection Download PDFInfo
- Publication number
- CN114937001B CN114937001B CN202210356868.XA CN202210356868A CN114937001B CN 114937001 B CN114937001 B CN 114937001B CN 202210356868 A CN202210356868 A CN 202210356868A CN 114937001 B CN114937001 B CN 114937001B
- Authority
- CN
- China
- Prior art keywords
- early warning
- photovoltaic
- entering
- image
- segmentation
- 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
- 239000000428 dust Substances 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 230000011218 segmentation Effects 0.000 claims abstract description 34
- 238000003708 edge detection Methods 0.000 claims abstract description 27
- 238000004140 cleaning Methods 0.000 claims abstract description 15
- 230000002159 abnormal effect Effects 0.000 claims abstract description 11
- 238000007689 inspection Methods 0.000 claims abstract description 5
- 238000012706 support-vector machine Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 238000010248 power generation Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 238000001931 thermography Methods 0.000 description 2
- 230000005457 Black-body radiation Effects 0.000 description 1
- OYPRJOBELJOOCE-UHFFFAOYSA-N Calcium Chemical compound [Ca] OYPRJOBELJOOCE-UHFFFAOYSA-N 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000001429 visible spectrum Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- 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
- G06T2207/10048—Infrared image
-
- 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
Abstract
The invention discloses a dust early warning method based on infrared detection, which comprises the following steps: s1, obtaining an optimal cleaning day according to dust early warning; s2, judging whether the current cleaning day is the optimal cleaning day, if so, entering S8, and if not, entering S3; s3, scanning the discrete rate of the photovoltaic string, judging whether the photovoltaic string is abnormal, if so, entering S7 after waiting for the preset time normally, and if so, entering S4; s4, starting the unmanned aerial vehicle to carry out inspection on the group strings with abnormal discrete rate and obtaining infrared photos; s5, performing edge detection on the infrared photo to obtain a final segmentation image; s6, performing hot spot detection on the segmented image, if the segmented image does not enter S7, entering S8 after performing hot spot early warning; s7, judging whether the current-day discrete rate scanning is finished, if yes, entering S8, and if not, returning to S3; s8, cleaning the photovoltaic string; s9, ending; the invention effectively carries out early warning on the photovoltaic hot spots in advance, and solves the problems of potential safety hazard and economic loss caused by the fact that the existing dust early warning system cannot find the hot spots in time.
Description
Technical Field
The invention relates to the field of photovoltaic power generation, in particular to a dust early warning method based on infrared detection.
Background
Because solar energy has the advantages of no pollution, wide distribution, reproducibility and the like, the solar energy is widely regarded as a clean energy source with the most development prospect. However, the influence of environmental dust cannot be avoided when the photovoltaic module works outdoors, and a great amount of dust accumulated on the surface can reduce the power generation efficiency and the income of the module, so that a dust early warning system is necessary to be established, the coverage condition of the dust is monitored in time, and the photovoltaic module is cleaned when a certain threshold value is reached.
The existing dust early warning system is mainly used for evaluating whether an early warning threshold value is reached or not and cleaning based on the change condition of dust coverage proportion and photovoltaic power generation capacity. However, in actual operation, we find that the existing early warning system has certain hysteresis, that is, although the dust amount does not reach the early warning threshold, the existing dust may already cause the occurrence of photovoltaic hot spots, further cause damage to the photovoltaic module, and even fire disaster, so that improvement of the existing dust early warning system is needed.
Disclosure of Invention
In order to solve the existing problems, the invention provides a dust early warning method based on infrared detection, which comprises the following specific scheme:
a dust early warning method based on infrared detection comprises the following steps:
s1, obtaining an optimal cleaning day according to dust early warning of a photovoltaic string;
s2, judging whether the current date is the optimal cleaning date, if so, entering a step S8, and if not, entering a step S3;
s3, carrying out discrete rate scanning on the photovoltaic string, judging whether the discrete rate is abnormal, if so, waiting for a preset time, and then entering a step S7, and if so, entering a step S4;
s4, starting the unmanned aerial vehicle to carry out inspection on the string with abnormal discrete rate in the step S3 and obtaining an infrared photo;
s5, performing edge detection on the infrared photo obtained in the step S4 to obtain a final segmentation image of the photovoltaic module;
s6, performing hot spot detection on the photovoltaic module after the split image, if no hot spot is detected, entering a step S7, and if the hot spot is detected, performing hot spot early warning and entering a step S8;
s7, judging whether the solar photovoltaic string discrete rate scanning is finished or not, if so, entering a step S8, and if not, returning to the step S3;
s8, cleaning the photovoltaic string;
s9, ending.
Preferably, the discrete rate in step S3 is used to evaluate the overall operation condition of the photovoltaic power station combiner box group string current, where the discrete rate=standard deviation of the group string current/average value of the group string current is 100%, and the step of discrete rate scanning includes:
s3.1, scanning the photovoltaic group strings under the same inverter, taking out the current of each photovoltaic group string, and calculating the discrete rate of the current;
s3.2, judging whether the discrete rate in the group string is abnormal according to the calculated result of the discrete rate.
Preferably, the edge detection algorithm in step S5 is an edge detection segmentation algorithm based on a parallel Canny operator, and the specific steps include:
s5.1, filtering and denoising the infrared photo;
s5.2, carrying out image enhancement on the infrared photo taking the photovoltaic string as a foreground background; the foreground and background contrast is increased by a contrast enhancement method so as to make foreground and background segmentation in HSV space by utilizing an H value;
s5.3, performing foreground segmentation on the infrared photo processed in the step S5.2;
s5.4, converting the image obtained after the foreground segmentation in the step S5.3 into a YCrCb color space by an RGB color space, separating a Cb channel, and performing Canny operator edge detection under the Cb channel, so that edges between the photovoltaic modules can be clearly seen, and a connected domain of the independent photovoltaic modules is obtained;
and S5.5, performing contour analysis and further filtering on the connected domain obtained in the step S5.4 by utilizing a contour tracking algorithm to obtain a final segmentation image of the photovoltaic module.
Preferably, the filtering denoising in step S5.1 uses a guided filter to preserve edge information.
Preferably, in the step S5.2, gamma transformation is adopted to enhance contrast, and when the gamma transformation factor is greater than 1, the details of the highlight region of the image can be increased, so as to meet the expected requirement.
Preferably, in the step S5.3, the foreground segmentation is performed in HSV space by using an H value by using a color image segmentation method, the segmentation result is converted into an R channel corresponding to the RGB image, and then a single-threshold segmentation is performed again to obtain a foreground binary image.
Preferably, the step of performing Canny operator edge detection in step S5.4 includes:
s5.4.1, the Canny operator comprises a high Canny operator with high double threshold and a low Canny operator with low double threshold which are connected in parallel, and the edge detection graph obtained by the high Canny operator is subjected to expansion operation;
s5.4.2, performing exclusive-or operation on the result obtained by the expansion operation and the foreground binary image to obtain an exclusive-or operation intermediate result image;
s5.4.3, detecting more edges as much as possible by the low Canny operator on the premise of sacrificing a certain accuracy, performing AND operation on the edges obtained by the low Canny operator and the edge detection diagram obtained by the high Canny operator after the edges are taken out, and adding morphological operation to obtain a final result image.
Preferably, the specific steps of the hot spot detection algorithm in step S6 include:
s6.1, extracting the characteristics of hot spots in the B channel, the Cr channel and the V channel;
s6.2, identifying the extracted features;
and S6.3, identifying the hot spots by using a hot spot identification algorithm based on the identified characteristics.
Preferably, the step S6.1 uses a feature extraction method of an adaptive sliding window to extract features, and the specific steps include:
s6.1.1 dividing the long side H of a solar photovoltaic panel in the photovoltaic module into 3 areas, wherein the long side H with the cross is equal to H/2, and the width is equal to the wide self-adaptive sliding window of the solar photovoltaic panel;
s6.1.2, calculating an artificial feature in each of said sliding windows, wherein said artificial feature refers to a pixel ratio alpha below the B-channel exceeding 10,
the calculation formula is that
Wherein bin i The cumulative frequency of pixel values i in the statistical histogram is represented, and H and W represent the height and width of the solar photovoltaic panel;
s6.1.3, extracting other features in the sliding window with the largest artificial feature, wherein the other features are features of extracting hot spots in a B channel, a Cr channel and a V channel, and forming a final input feature vector with the artificial feature of the sliding window.
Preferably, the hot spot recognition algorithm in the step S6.3 adopts a hot spot recognition algorithm based on a support vector machine, and adopts zero-mean normalization to make the data set according to the training set: test set = 4:1 ratio random partitioning.
The invention has the beneficial effects that:
the invention can effectively early warn the photovoltaic hot spots caused by dust in advance, thereby solving the problems of potential safety hazard and economic loss caused by the fact that the existing dust early warning system cannot find the hot spots in time.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a graph showing the effect obtained after edge detection according to the present invention;
FIG. 3 is a foreground binary image obtained after foreground segmentation according to the present invention;
FIG. 4 is a high Canny operator edge detection graph;
FIG. 5 is a high Canny operator result expansion graph;
FIG. 6 is an intermediate result diagram of an exclusive OR operation;
FIG. 7 is a graph of the results of the operations;
FIG. 8 is a split binary diagram;
fig. 9 is a schematic diagram of an adaptive sliding window.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a dust early warning method based on infrared detection includes the following steps:
s1, obtaining an optimal cleaning day according to dust early warning of a photovoltaic string; the dust early warning has more mature models, so that the optimal cleaning day D can be obtained from a general dust early warning system.
S2, judging whether the current date is the optimal cleaning date, if so, proceeding to step S8, and if not, proceeding to step S3.
S3, carrying out discrete rate scanning on the photovoltaic string, judging whether the discrete rate is abnormal, if so, waiting for a preset time, and then entering a step S7, and if so, entering a step S4.
The discrete rate in the step S3 is used for evaluating the overall running condition of the string current of the combiner box group of the photovoltaic power station, and the smaller the discrete rate value is, the more concentrated the branch current curves of the combiner boxes are, and the more stable the power generation condition is. Discrete rate = standard deviation of group string current/average value of group string current 100%.
Specific examples are as follows: the range of string current discrete values can be divided into four classes according to production operation and maintenance experience:
(1) The discrete rate is < =5%, and the operation is stable;
(2) 5% < discrete rate < = 10%, working well;
(3) 10% < discrete rate < = 20%, operation to be improved;
(4) The discrete rate is >20% and the string fails.
(5) Discrete rate = 100%, group string does not generate/drop string.
For example, the following junction boxes are connected into 8 strings in total, and the current conditions at a certain moment are as follows:
| string set | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| Electric current | 3.15 | 3.16 | 1.1 | 3.18 | 3.2 | 3.15 | 3.17 | 3.13 |
Calculated, the dispersion ratio= 25.12% >20%, and a certain component is judged to be faulty. It is checked that group string 3 is faulty.
The step of discrete rate scanning comprises the following steps:
s3.1, scanning the photovoltaic group strings under the same inverter, taking out the current of each photovoltaic group string, and calculating the discrete rate of the current;
s3.2, judging whether the discrete rate in the group string is abnormal or not according to the calculated result of the discrete rate.
S4, starting the unmanned aerial vehicle to carry out inspection on the group string with abnormal discrete rate in the step S3 and obtaining an infrared photo. The infrared photo is obtained by infrared heat map of the photovoltaic panel. Infrared thermography techniques use mid-wave (MWIR, about 3-5 μm) or long-wave (LWIR, about 7-14 μm) infrared sensors to acquire thermal images or thermographic images of the object under inspection. According to the law of black body radiation by planck, all objects emit infrared radiation in proportion to their temperature. For example, the turned-off bulb may emit very low power radiation, which is at wavelengths well above 1 μm, typically outside the range visible to the unaided human eye. If the light switch is turned on, the temperature will rise and the bulb will emit red light into the visible spectrum due to the higher radiant energy emitted by the bulb. If the temperature continues to rise, the radiation may change from red to purple, and the lamp light then appears white. Therefore, the infrared thermal imaging technique is a possible technique for determining the surface temperature of an object and checking whether an abnormality occurs in the temperature distribution of a human body or a solar photovoltaic panel.
And S5, performing edge detection on the infrared photo obtained in the step S4 to obtain a final segmentation image of the photovoltaic module, as shown in FIG. 2.
Edges in a physical sense generally correspond to discontinuities in the physical, brightness and geometric properties of a scene object that provide important visual information. The edges in a common physical sense correspond to significant changes in reflectivity, illumination, direction and depth of the scene surface. Considering that the solar photovoltaic panel approximates a rectangle, the segmentation algorithm fully utilizes the priori knowledge to convert the segmentation problem of the solar photovoltaic panel into the edge detection problem, obtains the outline points of the solar photovoltaic panel according to the edge detection algorithm, constructs the minimum circumscribed rectangle to approximate the solar photovoltaic panel, and realizes the segmentation of the solar photovoltaic panel.
The edge detection algorithm in the step S5 is an edge detection segmentation algorithm based on a parallel Canny operator, and the specific steps include:
s5.1, filtering and denoising the infrared photo; the filtering denoising adopts a guide filter to filter so as to preserve edge information.
S5.2, carrying out image enhancement on the infrared photo taking the photovoltaic string as a foreground background; the foreground and background contrast is increased by a contrast enhancement method so as to utilize an H value for foreground and background segmentation in HSV space. The contrast is enhanced by adopting a gamma conversion method, and when the gamma conversion factor is larger than 1, the details of the highlight area of the image can be increased, so that the method meets the expected requirement.
S5.3, performing foreground segmentation on the infrared photo processed in the step S5.2; specifically, the foreground segmentation is performed in the HSV space by using the H value by using a color image segmentation method, the segmentation result is converted into an R channel corresponding to the RGB image, and then single-threshold segmentation is performed once again, so that a foreground binary image is obtained, as shown in fig. 3.
S5.4, converting the image obtained after the foreground segmentation in the step S5.3 into a YCrCb color space by an RGB color space, separating a Cb channel, and performing Canny operator edge detection under the Cb channel, so that edges between the photovoltaic modules can be clearly seen, and a connected domain of the independent photovoltaic modules is obtained;
the step of performing Canny operator edge detection comprises the following steps:
s5.4.1, the Canny operator comprises a high Canny operator with high double threshold and a low Canny operator with low double threshold which are connected in parallel, and the edge detection graph obtained by the high Canny operator is subjected to expansion operation, wherein the edge detection graph of the high Canny operator is shown in fig. 4; the high Canny operator result expansion map is shown in fig. 5.
S5.4.2, performing exclusive-or operation on the result obtained by the expansion operation and the foreground binary image to obtain an exclusive-or operation intermediate result image, as shown in fig. 6.
S5.4.3, the low Canny operator detects more edges as much as possible under the premise of sacrificing a certain accuracy, and performs AND operation with an edge detection diagram obtained by the high Canny operator after the opposite, wherein the AND operation result diagram is shown in fig. 7, and morphological operation is added to obtain a final result image, namely a segmentation binary diagram, as shown in fig. 8.
And S5.5, carrying out contour analysis and further filtering on the connected domain obtained in the step S5.4 by utilizing a contour tracking algorithm to obtain a final segmentation image of the photovoltaic module. Specific examples are as follows: firstly, calculating the area S of the connected domain according to the contour points contour In order to improve the fault tolerance, the upper limit of the area of the connected domain is set to 10000, the lower limit is set to 1000, and the solar photovoltaic panel with larger and smaller area caused by the change of the flight height of the unmanned aerial vehicle is reserved as far as possible. Considering that the solar photovoltaic panel is close to a rectangle, even if the segmentation result is deviated, the whole is close to the rectangle, an index is designed, and the index is defined as the convex hull area of the connected domain (S convex_hull ) The area difference ratio with the connected domain is calculated as follows:
if eta is less than or equal to 0.28, the connected domain is reserved, otherwise, the connected domain is abandoned. η is mainly used to filter the background of the misclassification.
S6, performing hot spot detection on the photovoltaic module after the split image, if no hot spot is detected, entering a step S7, and if the hot spot is detected, performing hot spot early warning and entering a step S8;
among them, hot spots are a common fault occurring on solar photovoltaic panels, negatively affecting the normal operation of the solar photovoltaic panels. One of the main causes of hot spots is dust shielding. When a certain cell of the solar photovoltaic panel is shaded, it reduces the current through that cell. Other normal cells that produce higher voltages typically reverse bias the shaded cells, which can be significant in power consumption. High power losses in the small range can lead to local overheating of the solar panel and thus to hot spots, which can lead to serious power losses and, in the worst case, even to fires. By researching the principle of hot spot generation, it is not difficult to find that the solar photovoltaic panel with the hot spot can generate local area overheat during normal operation, the phenomenon is not obvious under visible light, but the hot spot can generate local bright spots in an infrared heat map due to higher temperature. The hot spots under the infrared heat map have obvious difference from the normal area, so that the machine vision method is favorable for hot spot detection under the infrared heat map.
The hot spot detection algorithm in step S6 specifically includes:
s6.1, extracting the characteristics of hot spots in the B channel, the Cr channel and the V channel;
experimental observation can find that the hot spots are closer to white than normal solar light Fu Banliang in RGB color channel pictures, and the color of a normal solar photovoltaic panel is closer to yellow in the hot spot area pixel point RGB color space. Ideally the yellow is [255, 0] in the RGB color space, and the B channel is a very small integer, even if it is not pure yellow, most of which is actually below 10. The white color is [255, 255, 255] in the RGB color space, and the B-channel value is an integer close to 255 even if the hot spot is not pure white, so that the difference between the pixel value of the hot spot region and the pixel value of the normal solar photovoltaic panel region is about 200 in the B-channel. Therefore, the B channel is a color channel which can obviously distinguish hot spots from normal solar photovoltaic panels, and the Cr channel and the V channel are similar characteristic extraction channels.
As shown in fig. 9, the feature extraction method using the adaptive sliding window extracts features, which specifically includes the steps of:
s6.1.1 dividing the long side H of a solar photovoltaic panel in the photovoltaic module into 3 areas, wherein the long side H with the cross is equal to H/2, and the width is equal to the wide self-adaptive sliding window of the solar photovoltaic panel;
s6.1.2, an artificial feature is calculated in each sliding window, where artificial feature refers to a pixel ratio alpha of more than 10 under the B channel,
the calculation formula is that
Wherein bin i The cumulative frequency of pixel values i in the statistical histogram is represented, and H and W represent the height and width of the solar photovoltaic panel;
s6.1.3, extracting other features in the sliding window with the largest artificial features, wherein the other features are features of hot spots extracted in the B channel, the Cr channel and the V channel, and forming final input feature vectors with the artificial features of the sliding window.
S6.2, identifying the extracted features;
and S6.3, identifying the hot spots by using a hot spot identification algorithm based on the identified features.
The hot spot recognition algorithm in the step S6.3 adopts a hot spot recognition algorithm based on a Support Vector Machine (SVM), and adopts zero-mean normalization to normalize a data set according to a training set: test set = 4:1 ratio random partitioning.
The related concept of support vector machine (support vector machines, SVM) derives from VC theory proposed by Vapnik and cores in 1963, an algorithm for assigning labels to new objects by example learning, mainly to solve the binary linear classification problem.
The central idea of the SVM is to find a separation hyperplane which can correctly distinguish the training set data and the geometric distance from the positive and negative sample points closest to the hyperplane in the training set data to the hyperplane is the largest, and the hyperplane can ensure that the training set data is correctly classified with the largest confidence. The core keywords of the SVM are separation hyperplane, interval maximization, soft interval and core skill.
Since solar photovoltaic panels with hot spots account for a few of all solar photovoltaic panels, specific examples are as follows: 250 positive samples are extracted from a solar photovoltaic panel of 300 infrared heat maps of a certain photovoltaic power station, and 400 negative samples are extracted in order to balance the positive and negative samples as much as possible in consideration of the fact that the number of actual negative samples is far greater than that of positive samples. If the artificial feature input feature vector in step S6.1.3 is 16-dimensional in total, i.e. 12-dimensional features are extracted in the B channel, 2-dimensional features are extracted in the Cr channel, 2-dimensional features are extracted in the V channel, and 16-dimensional features are extracted in total. Since the support vector machine is "distance" sensitive, the input feature vector requires a normalization operation, here Zero-mean normalization (Zero-Score Normalization), to normalize the data set to a training set: test set = 4:1 ratio randomly divides training and test sets. The photovoltaic panels with hot spots are correctly identified, the average identification time is 820ms, and the actual operation and maintenance requirements can be met.
S7, judging whether the solar photovoltaic string discrete rate scanning is finished or not, if so, entering a step S8, and if not, returning to the step S3;
s8, cleaning the photovoltaic string; the operation and maintenance personnel go to the appointed problem string, and the string dust is cleaned, so that the hidden hot spot trouble or the influence of dust on power generation is eliminated.
S9, ending.
Although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. The dust early warning method based on infrared detection is characterized by comprising the following steps of:
s1, obtaining an optimal cleaning day according to dust early warning of a photovoltaic string;
s2, judging whether the current date is the optimal cleaning date, if so, entering a step S8, and if not, entering a step S3;
s3, carrying out discrete rate scanning on the photovoltaic string, judging whether the discrete rate is abnormal, if so, waiting for a preset time, and then entering a step S7, and if so, entering a step S4;
s4, starting the unmanned aerial vehicle to carry out inspection on the string with abnormal discrete rate in the step S3 and obtaining an infrared photo;
s5, performing edge detection on the infrared photo obtained in the step S4 to obtain a final segmentation image of the photovoltaic module;
s6, performing hot spot detection on the photovoltaic module after the split image, if no hot spot is detected, entering a step S7, and if the hot spot is detected, performing hot spot early warning and entering a step S8;
the hot spot detection algorithm comprises the following specific steps:
s6.1, extracting the characteristics of hot spots in the B channel, the Cr channel and the V channel;
the step S6.1 extracts the features by using a feature extraction method of the adaptive sliding window, and the specific steps include:
s6.1.1 dividing the long side H of a solar photovoltaic panel in the photovoltaic module into 3 areas, wherein the long side H with the cross is equal to H/2, and the width is equal to the wide self-adaptive sliding window of the solar photovoltaic panel;
s6.1.2, calculating an artificial feature in each of said sliding windows, wherein said artificial feature refers to a pixel ratio alpha below the B-channel exceeding 10,
the calculation formula is that
Wherein bin i The cumulative frequency of pixel values i in the statistical histogram is represented, and H and W represent the height and width of the solar photovoltaic panel;
s6.1.3, extracting other features in the sliding window with the largest artificial features, wherein the other features are features of extracting hot spots in a B channel, a Cr channel and a V channel, and forming a final input feature vector with the artificial features of the sliding window;
s6.2, identifying the extracted features;
s6.3, identifying hot spots by a hot spot identification algorithm based on the identified features;
s7, judging whether the solar photovoltaic string discrete rate scanning is finished or not, if so, entering a step S8, and if not, returning to the step S3;
s8, cleaning the photovoltaic string;
s9, ending.
2. The dust warning method based on infrared detection according to claim 1, wherein the discrete rate in step S3 is used for evaluating the overall operation condition of the group string current of the photovoltaic power station, the discrete rate=standard deviation of the group string current/average value of the group string current is 100%, and the step of discrete rate scanning includes:
s3.1, scanning the photovoltaic group strings under the same inverter, taking out the current of each photovoltaic group string, and calculating the discrete rate of the current;
s3.2, judging whether the discrete rate in the group string is abnormal according to the calculated result of the discrete rate.
3. The dust early warning method based on infrared detection according to claim 1, wherein the edge detection algorithm in step S5 is an edge detection segmentation algorithm based on a parallel Canny operator, and the specific steps include:
s5.1, filtering and denoising the infrared photo;
s5.2, carrying out image enhancement on the infrared photo taking the photovoltaic string as a foreground background; the foreground and background contrast is increased by a contrast enhancement method so as to make foreground and background segmentation in HSV space by utilizing an H value;
s5.3, performing foreground segmentation on the infrared photo processed in the step S5.2;
s5.4, converting the image obtained after the foreground segmentation in the step S5.3 into a YCrCb color space by an RGB color space, separating a Cb channel, and performing Canny operator edge detection under the Cb channel, so that edges between the photovoltaic modules can be clearly seen, and a connected domain of the independent photovoltaic modules is obtained;
and S5.5, performing contour analysis and further filtering on the connected domain obtained in the step S5.4 by utilizing a contour tracking algorithm to obtain a final segmentation image of the photovoltaic module.
4. The dust early warning method based on infrared detection according to claim 3, wherein: the filtering and denoising in the step S5.1 is performed by using a pilot filter to preserve edge information.
5. The dust early warning method based on infrared detection according to claim 3, wherein: in the step S5.2, gamma transformation is adopted to enhance contrast, and when the gamma transformation factor is greater than 1, the details of the highlight region of the image can be increased, so that the expected requirement is met.
6. The dust early warning method based on infrared detection according to claim 3, wherein: in the step S5.3, the foreground segmentation is carried out in HSV space by utilizing H values by utilizing a color image segmentation method, segmentation results are converted into R channels corresponding to RGB images, and single-threshold segmentation is carried out again to obtain a foreground binary image.
7. The dust pre-warning method based on infrared detection according to claim 6, wherein the step of performing Canny operator edge detection in step S5.4 includes:
s5.4.1, the Canny operator comprises a high Canny operator with high double threshold and a low Canny operator with low double threshold which are connected in parallel, and the edge detection graph obtained by the high Canny operator is subjected to expansion operation;
s5.4.2, performing exclusive-or operation on the result obtained by the expansion operation and the foreground binary image to obtain an exclusive-or operation intermediate result image;
s5.4.3, detecting more edges as much as possible by the low Canny operator on the premise of sacrificing a certain accuracy, performing AND operation on the edges obtained by the low Canny operator and the edge detection diagram obtained by the high Canny operator after the edges are taken out, and adding morphological operation to obtain a final result image.
8. The dust early warning method based on infrared detection according to claim 1, wherein the hot spot recognition algorithm in the step S6.3 adopts a hot spot recognition algorithm based on a support vector machine, and a zero-mean normalization is adopted to normalize a data set according to a training set: test set = 4:1 ratio random partitioning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210356868.XA CN114937001B (en) | 2022-04-06 | 2022-04-06 | Dust early warning method based on infrared detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210356868.XA CN114937001B (en) | 2022-04-06 | 2022-04-06 | Dust early warning method based on infrared detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114937001A CN114937001A (en) | 2022-08-23 |
| CN114937001B true CN114937001B (en) | 2023-06-09 |
Family
ID=82862512
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210356868.XA Active CN114937001B (en) | 2022-04-06 | 2022-04-06 | Dust early warning method based on infrared detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114937001B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115514021B (en) * | 2022-11-11 | 2023-03-28 | 云南电网能源投资有限责任公司 | Fault regulation and control method and system for distributed photovoltaic power station and computer equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200764A (en) * | 2020-09-02 | 2021-01-08 | 重庆邮电大学 | Photovoltaic power station hot spot detection and positioning method based on thermal infrared image |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107617616B (en) * | 2016-07-14 | 2019-03-08 | 北京天诚同创电气有限公司 | Cleaning device and photovoltaic array |
| CN111555716B (en) * | 2020-03-13 | 2023-07-28 | 远景智能国际私人投资有限公司 | Method, device, equipment and storage medium for determining working state of photovoltaic array |
| CN111614317B (en) * | 2020-05-12 | 2022-11-15 | 国家电投集团江西电力有限公司 | IV curve scanning-based diagnosis method for shadow shielding of photovoltaic panel |
| CN112434606A (en) * | 2020-11-24 | 2021-03-02 | 黄晟晖 | Double-light-integration-based solar cell panel RGB image recognition method and device |
-
2022
- 2022-04-06 CN CN202210356868.XA patent/CN114937001B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200764A (en) * | 2020-09-02 | 2021-01-08 | 重庆邮电大学 | Photovoltaic power station hot spot detection and positioning method based on thermal infrared image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114937001A (en) | 2022-08-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110266268B (en) | Photovoltaic module fault detection method based on image fusion recognition | |
| Niazi et al. | Hotspot diagnosis for solar photovoltaic modules using a Naive Bayes classifier | |
| Davari et al. | Intelligent diagnosis of incipient fault in power distribution lines based on corona detection in UV-visible videos | |
| CN111696075A (en) | Intelligent fan blade defect detection method based on double-spectrum image | |
| 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 | |
| CN111652857A (en) | Infrared detection method for insulator defects | |
| CN114937001B (en) | Dust early warning method based on infrared detection | |
| CN105335949A (en) | Video image rain removal method and system | |
| CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
| CN111105398A (en) | Transmission line component crack detection method based on visible light image data | |
| Prabhakaran et al. | Deep Learning-Based Model for Defect Detection and Localization on Photovoltaic Panels. | |
| CN113837994B (en) | Photovoltaic panel defect diagnosis method based on edge detection convolutional neural network | |
| Wang et al. | Effective segmentation approach for solar photovoltaic panels in uneven illuminated color infrared images | |
| Sohail et al. | Fault detection and computation of power in PV cells under faulty conditions using deep-learning | |
| Wang et al. | Efficient and refined deep convolutional features network for the crack segmentation of solar cell electroluminescence images | |
| CN112053322B (en) | Method for segmenting and detecting surface shielding of photovoltaic cell panel | |
| KR101779040B1 (en) | Method and apparatus for extracting panel area from thermal infrared images of photovoltaic array | |
| Afifah et al. | A new approach for hot spot solar cell detection based on multi-level otsu algorithm | |
| Pant et al. | Automatic Nepali number plate recognition with support vector machines | |
| CN110610474A (en) | Solar panel defect real-time detection method based on infrared image | |
| Pathak et al. | Evaluation of effect of pre-processing techniques in solar panel fault detection | |
| Wang et al. | Machine vision-based conveyor belt tear detection in a harsh environment | |
| CN109308685A (en) | Infrared photovoltaic array dividing method based on Threshold segmentation and K mean cluster | |
| Zhou et al. | Water photovoltaic plant contaminant identification using visible light images | |
| Rajapaksha et al. | Optimized Multi-Shaped Traffic Light Detection and Signal Recognition with Guided Framework |
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 |