CN118096738B

CN118096738B - New energy photovoltaic module fault detection method and system

Info

Publication number: CN118096738B
Application number: CN202410487334.XA
Authority: CN
Inventors: 胡国利
Original assignee: Yishite Intelligent System Integration Co ltd
Current assignee: Yishite Intelligent System Integration Co ltd
Filing date: 2024-04-23
Publication date: 2024-06-25
Anticipated expiration: 2044-04-23

Abstract

The invention relates to the technical field of photovoltaic system monitoring, in particular to a new energy photovoltaic module fault detection method and system. Firstly, acquiring an initial image of a component of a photovoltaic component, and determining color distances of two component blocking areas according to gray value distribution differences between the two component blocking areas, an abnormal probability value of the component blocking areas and a background gray value; acquiring the adjustment space distance of the two component block areas according to the difference of the abnormal probability values and the initial space distance of the two component block areas and the difference of the color distances between the component block areas and all the component block areas in the preset neighborhood range; and obtaining a fault detection result of the photovoltaic module according to the module saliency map. According to the invention, the single-scale significance value capable of effectively reflecting the difference between the weak hot spot area and the normal area is constructed, so that the significance detection effect of the component significance map is improved, and the accuracy of the fault detection result is improved.

Description

New energy photovoltaic module fault detection method and system

Technical Field

The invention relates to the technical field of photovoltaic system monitoring, in particular to a new energy photovoltaic module fault detection method and system.

Background

The photovoltaic module is used as a core part of a new energy power generation system, and the performance of the photovoltaic module directly influences the power generation efficiency and stability of the whole system. Through the fault detection of the photovoltaic module, potential problems can be found and solved in time, and the normal operation of the photovoltaic module is ensured, so that the expected power generation efficiency is achieved. The hot spot is a fault phenomenon possibly occurring in the operation process of the photovoltaic module, and can cause the local temperature of the module to be too high, so that the output power and the service life of the photovoltaic module are affected. In order to ensure efficient operation and long-term stability of the photovoltaic module, it is important to find hot spot areas of the photovoltaic module in time.

The temperature of a normal region on the photovoltaic module is uniformly changed, the temperature of a hot spot region is generally higher, the temperature is more severe than that of a surrounding region, the image of the hot spot region on the photovoltaic module is more obvious, and the prior art generally adopts a significance detection CA (Context-Aware) algorithm to carry out significance analysis on the image of the photovoltaic module, so that the hot spot region in the photovoltaic module is determined. The saliency detection CA algorithm can effectively detect the region with obvious hot spot performance by calculating the single-scale saliency value, but the single-scale saliency value is difficult to embody the difference between the weak hot spot region and the normal region due to the insignificant influence of the difference between the weak hot spot region and the normal region in the process of calculating the single-scale saliency value, so that the region with weaker hot spot performance is difficult to detect, and the fault detection result of the photovoltaic module is inaccurate.

Disclosure of Invention

In order to solve the technical problems that in the prior art, in the process of calculating a single-scale significance value by a significance detection CA algorithm, the single-scale significance value is difficult to embody the difference between a weak hot spot area and a normal area and the area with weak hot spot performance is difficult to detect, and the failure detection result of a photovoltaic module is inaccurate, the invention aims to provide a new energy photovoltaic module failure detection method and system, and the adopted technical scheme is as follows:

A new energy photovoltaic module fault detection method, the method comprising:

Acquiring an initial image of a component of the photovoltaic component; determining a background gray value of the initial image of the component; under different preset scales, uniformly dividing the initial image of the component into a preset number of component block areas;

Optionally selecting one preset scale as a scale to be analyzed; determining an abnormal probability value according to the fluctuation of gray values and gradient values of all pixel points in the component block area and the difference between the gray values of the pixel points and the background gray values under the scale to be analyzed; determining the color distance of the two component blocking areas according to the gray value distribution difference between the two component blocking areas, the abnormal probability value of the component blocking areas and the background gray value; acquiring the adjustment space distance of the two component block areas according to the difference of the abnormal probability values and the initial space distance of the two component block areas and the difference of the color distances between the component block areas and all the component block areas in the preset neighborhood range;

Acquiring non-similarity measurement values of the two component blocking areas according to the adjustment space distance of the two component blocking areas and the color distance of the two component blocking areas under the to-be-analyzed scale; obtaining a single-scale significance value of the component blocking area according to the difference of abnormal probability values between the component blocking area and all the component blocking areas;

Obtaining a component saliency map corresponding to the component initial image by using a saliency detection CA algorithm according to single-scale saliency values of all component blocking areas under all preset scales; and obtaining a fault detection result of the photovoltaic module according to the module saliency map.

Further, the method for acquiring the abnormal probability value includes:

obtaining an abnormal probability value according to an abnormal probability value formula, wherein the obtaining formula of the abnormal probability value comprises the following steps:

; wherein/> For/>Abnormal probability values for the component block areas; /(I)For/>Variance of gray values of all pixel points in each component block area; /(I)For/>Variance of gradient values of all pixel points in each component block area; /(I)For/>The/>, in each of the component blocking areasGray values of the individual pixels; /(I)For/>The total number of all pixel points in the component blocking area; /(I)Background gray values for the initial image of the component; /(I)Is a normalization function; /(I)As a hyperbolic tangent function.

Further, the method for obtaining the color distance comprises the following steps:

Acquiring an area gray value of the component block area according to the abnormal probability value of the component block area, the difference between the gray value of each pixel point in the component block area and the background gray value and the gray value of each pixel point in the component block area;

and calculating absolute values of differences of the gray values of the two component block areas to obtain color distances of the two component block areas under the dimension to be analyzed.

Further, the method for acquiring the regional gray value comprises the following steps:

Obtaining a regional gray value according to a regional gray value formula, wherein the regional gray value obtaining formula comprises the following steps:

; wherein/> For/>The region gray values of the component block regions; /(I)For/>Abnormal probability values for the component block areas; /(I)For/>The/>, in each of the component blocking areasGray values of the individual pixels; /(I)For/>The total number of all pixel points in the component blocking area; /(I)Background gray values for the initial image of the component; /(I)Is a normalization function.

Further, the method for acquiring the adjustment space distance comprises the following steps:

In a preset neighborhood range of the component blocking area, taking each component blocking area as each reference blocking area of the component blocking area;

Calculating the absolute value of the difference value of the color distance between the component block area and the reference block area to obtain the characteristic difference value of the component block area corresponding to the reference block area;

calculating the average value of the characteristic difference values of the component block areas corresponding to all the reference block areas, and obtaining the color significance value of the component block areas;

acquiring an adjustment spatial distance according to an adjustment spatial distance formula, wherein the acquisition formula of the adjustment spatial distance comprises:

; wherein/> For/>The component is divided into areas and the/>Adjusting the space distance between the component blocking areas; /(I)For/>Color saliency values for each of the component blocked areas; /(I)For/>The abnormal probability values for the component chunk areas; /(I)For/>The abnormal probability values for the component chunk areas; /(I)For/>The component is divided into areas and the/>The initial spatial distance between the component chunk regions; /(I)Is an exponential function based on e; /(I)As a hyperbolic tangent function.

Further, the method for obtaining the dissimilarity measure value includes:

acquiring non-similarity measurement values of the two component blocking areas according to the adjustment space distances of the two component blocking areas and the color distances of the two component blocking areas;

The adjusted spatial distance and the dissimilarity measure are in negative correlation; the color distance and the dissimilarity measure are positively correlated.

Further, the method for acquiring the single-scale saliency value comprises the following steps:

Obtaining a single-scale saliency value according to a single-scale saliency value formula, wherein the single-scale saliency value obtaining formula comprises:

; wherein/> For/>Individual said component blocking area single-scale significance values; /(I)For/>The abnormal probability values for the component chunk areas; /(I)For/>The abnormal probability values for the component chunk areas; /(I)The total number of blocked areas for all components; /(I)For/>The component is divided into areas and the/>A non-similarity measure between the component partitioned areas; /(I)Is an exponential function based on e; /(I)Is a normalization function.

Further, the method for acquiring the component saliency map comprises the following steps:

Taking the average value of single-scale significance values of the component block areas under all preset scales as the average significance value of each component block area;

And acquiring a component saliency map corresponding to the component initial image by using a saliency detection CA algorithm according to the average saliency values of all the component blocking areas.

Further, the method for obtaining the background gray value comprises the following steps:

and taking the gray value with the maximum number of corresponding pixel points in the initial image of the component as the background gray value of the initial image of the component.

The invention also provides a new energy photovoltaic module fault detection system, which comprises: a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of any one of the methods when the computer program is executed.

The invention has the following beneficial effects:

The invention needs to firstly acquire an initial image of the component of the photovoltaic component. Considering that the hot spot area is caused by local circuit short circuit, the local temperature is higher and the change is relatively more severe; the normal area is mainly caused by natural reasons such as sunlight irradiation angle, and the change is relatively gentle; the normal temperature change area and the hot spot area possibly exist in the component blocking area, the probability of containing the hot spot area is different for different component blocking areas, the abnormal probability value of the component blocking area under the to-be-analyzed scale is obtained, and the abnormal probability value can reflect the local abnormal probability of the component blocking area under the to-be-analyzed scale. In order to construct a single-scale saliency value that effectively reflects the difference between the hot spot region and the normal region, it is necessary to improve the contrast between the hot spot region and the normal region. Firstly, the color distance is obtained, and the color distance can more effectively reflect the distinguishing degree of the hot spot area and the normal area on the color characteristics. In order to enhance the distinguishing degree of the spot areas and the normal areas at the space positions, the larger the difference of abnormal probability values of the two component block areas is considered, the larger the probability distinguishing degree of the two component block areas containing hot spot parts is shown, and the fact that the hot spot areas are mainly caused by internal short circuits due to the fact that part of battery pieces are blocked and cannot work is considered, wherein the blocking parts are relatively concentrated in the image position space, and then the initial space distance of the two component block areas is adjusted, and the adjusted space distance of the two component block areas under the to-be-analyzed scale is obtained; adjusting the spatial distance can more effectively reflect the division between the hot spot region and the normal region in the spatial position. And the dissimilarity measure value can more effectively reflect the difference degree between two component blocking areas under the dimension to be analyzed, ensure that the single-dimension saliency value can reflect the saliency degree of a full weak hot spot area, improve the saliency detection effect of the component saliency map and improve the accuracy of a fault detection result.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages 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 only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for detecting a fault of a new energy photovoltaic module according to an embodiment of the present invention.

Detailed Description

In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of the new energy photovoltaic module fault detection method and system according to the invention with reference to the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

The invention provides a new energy photovoltaic module fault detection method and a system specific scheme by combining the drawings.

Referring to fig. 1, a method flowchart of a new energy photovoltaic module fault detection method according to an embodiment of the present invention is shown, and the method includes the following steps:

step S1: acquiring an initial image of a component of the photovoltaic component; determining a background gray value of an initial image of the component; and uniformly dividing the component initial image into a preset number of component block areas under different preset scales.

Hot spots in photovoltaic modules are one of the major faults in the detection process. The reason for the generation of the hot spots is that part of the battery pieces do not provide the generated power contribution and become energy-consuming loads in the assembly, meanwhile, the local temperature of the assembly is increased, the bypass diode is started, the corresponding battery strings are short-circuited, the generated energy is influenced, and even the assembly is burnt. In order to ensure efficient operation and long-term stability of the photovoltaic module, it is necessary to discover hot spot faults of the photovoltaic module. The invention needs to firstly acquire an initial image of the component of the photovoltaic component. In order to reflect the gray value of the normal panel area, a background gray value is obtained. Since the saliency of the image may be represented differently at different scales, the saliency of the image can be comprehensively captured by detecting at different preset scales. By uniformly dividing the component initial image into a preset number of component block areas, the local significance of each component block area can be analyzed, so that the characteristics of different parts of the image are captured.

Specifically, an infrared image of a photovoltaic cell panel is acquired through an unmanned aerial vehicle carrying infrared thermal imager, the acquired infrared image has noise, and the noise can influence a fault area of a photovoltaic module to be analyzed later, so that noise reduction operation is carried out on the image, and a noise reduction image is acquired. And the influence caused by noise and external interference is eliminated, and the accuracy of subsequent analysis is enhanced. The noise reduction image is subjected to graying treatment to obtain a gray image of the photovoltaic cell panel, so that a pre-trained semantic segmentation network can be adopted to extract the image only containing the photovoltaic cell panel from the gray image in order to avoid unnecessary region detection, and an initial assembly image of the photovoltaic assembly is obtained. The embodiment of the invention adopts bilateral filtering to reduce noise of the image, and an implementer can set the image according to actual conditions. The semantic segmentation network adopts a DNN neural network and is in an Encode-Decode structure, the network input is a gray image of the photovoltaic cell panel, and the network output is an initial image of a component only comprising the photovoltaic cell panel; the labeling mode is that the area only containing the photovoltaic cell panel is marked as 1, the other areas are marked as 0, and the cross entropy loss function is used as the network loss function; the image acquisition device can be adjusted according to the implementation scene, and is not limited and described herein, and the training process of the semantic segmentation network is a process well known to those skilled in the art, and is not described herein in detail.

It should be noted that, in order to facilitate the operation, all index data involved in the operation in the embodiment of the present invention is subjected to data preprocessing, so as to cancel the dimension effect. The specific means for removing the dimension influence is a technical means well known to those skilled in the art, and is not limited herein.

Considering that when the infrared thermal imager collects the infrared image of the solar cell, the surface of the solar cell is provided with glass, the infrared thermal imager indirectly detects the heat distribution of the solar cell below the glass by detecting the heat distribution of the glass on the surface of the solar cell, so that the initial image of the assembly is difficult to accurately reflect the temperature distribution of the cell panel, the difference between a hot spot area and a normal temperature area of the cell panel is small, and the saliency detection CA algorithm is difficult to detect the weak hot spot area.

The initial image of the component has a normal area with uniformly changed gray values, the hot spot area of the photovoltaic component is mainly represented as an area with higher gray values compared with the whole normal area in the initial image of the component, and compared with the normal battery area in normal operation, the hot spot area has severe temperature change and the gray value of the hot spot area in the initial image of the component has larger change. The hot spot area in the initial image of the component shows more remarkable, and the distribution characteristics of various components in the image can be quantified by a statistical method for the saliency detection, so that whether the difference among the components is remarkable or not is judged, namely whether obvious uneven distribution phenomenon exists or not is judged. Therefore, in the embodiment of the invention, the saliency detection CA algorithm is adopted to carry out the saliency detection, and the steps of the saliency detection CA algorithm can be summarized as follows:

(1) The color distance of the regions at a single scale, which reflects the differences in color characteristics between the regions, and the spatial distance of the regions at a single scale, which describes their relative positional relationship in the image, are calculated. Obtaining a dissimilarity measure value of the region under a single scale; the dissimilarity measure of the region is proportional to the color distance; inversely proportional to the spatial distance. Under each scale, obtaining a single-scale significance value of each scale according to the dissimilarity measure values of all the areas; (2) And expanding the single-scale calculation to multiple scales, and finally obtaining the linear average of the single-scale significance values of each region under all scales as an average significance value. (3) extracting the region of interest and redefining the saliency value. (4) obtaining a saliency map, and highlighting a saliency region. However, in the process of calculating the single-scale saliency value by the saliency detection CA algorithm in the step (1), the single-scale saliency value is not obviously influenced by the difference between the weak hot spot area and the normal area, so that the difference between the weak hot spot area and the normal area is difficult to embody, and the area with weak hot spot performance is difficult to detect.

Preferably, in order to reflect the gray value of the normal panel area, the method for obtaining the background gray value according to one embodiment of the present invention includes:

Specifically, different scales can reveal different salient features in image processing, small scales often show details, large scales often show overall structures or modes, some features are more remarkable on the small scales, and some features are remarkable on the large scales, and the salient degree of the image can be comprehensively captured by detecting under different preset scales. To capture the features of different parts of the image, a zonal study is also required. Therefore, in this embodiment of the present invention, different preset scales are set, and under each preset scale, the initial image is uniformly divided into a preset number of area images, and the preset number has a value of 64, and it should be noted that, in this embodiment of the present invention, the scales specifically refer to dimensions, and the preset scales are set to 100%, 80%, 50%, 30%, 10%, and the preset scales can be adjusted according to the implementation scenario, which is not limited herein. In one embodiment of the present invention, the preset number is 64, and the practitioner can set the preset number according to the implementation scenario.

Step S2, selecting a preset scale as a scale to be analyzed; determining an abnormal probability value according to the fluctuation of gray values and gradient values of all pixel points in the component block area and the difference between the gray values of the pixel points and the background gray values under the scale to be analyzed; determining the color distance of the two component block areas according to the gray value distribution difference between the two component block areas, the abnormal probability value of the component block areas and the background gray value; and acquiring the adjustment space distance of the two component block areas according to the difference of the abnormal probability values and the initial space distance of the two component block areas and the difference of the color distances between the component block areas and all the component block areas in the preset neighborhood range.

Considering that the hot spot area is caused by local circuit short circuit, the local temperature is higher and the change is relatively more severe; the normal area is mainly caused by natural reasons such as sunlight irradiation angle, and the change is relatively gentle; the normal temperature change area and the hot spot area possibly exist in the component blocking area, the probability of containing the hot spot area is different for different component blocking areas, the abnormal probability value of the component blocking area under the to-be-analyzed scale is obtained, and the abnormal probability value can reflect the local abnormal probability of the component blocking area under the to-be-analyzed scale. In order to construct a single-scale saliency value that effectively reflects the difference between the hot spot region and the normal region, it is necessary to improve the contrast between the hot spot region and the normal region. Firstly, obtaining color distances of two component blocking areas under the dimension to be analyzed; the color distance can more effectively reflect the degree of distinction between the hot spot region and the normal region in the color feature. In order to enhance the distinguishing degree of the spot areas and the normal areas at the space positions, the larger the difference of abnormal probability values of the two component block areas is considered, the larger the probability distinguishing degree of the two component block areas containing hot spot parts is shown, and the fact that the hot spot areas are mainly caused by internal short circuits due to the fact that part of battery pieces are blocked and cannot work is considered, wherein the blocking parts are relatively concentrated in the image position space, and then the initial space distance of the two component block areas is adjusted, and the adjusted space distance of the two component block areas under the to-be-analyzed scale is obtained; adjusting the spatial distance can more effectively reflect the division between the hot spot region and the normal region in the spatial position.

Preferably, in order to reflect the probability that the component block area includes a hot spot area under the scale to be analyzed, in one embodiment of the present invention, the method for acquiring the abnormal probability value includes:

; wherein/> To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)To be under the scale to be analyzed, the firstVariance of gray values of all pixel points in the individual component block areas; /(I)To be at the scale to be analyzed, the first/>Variance of gradient values of all pixel points in the individual component block areas; /(I)To be at the scale to be analyzed, the first/>First/>, in individual component chunking areasGray values of the individual pixels; /(I)To be at the scale to be analyzed, the first/>The total number of all pixel points in the individual component block area; /(I)Background gray values for the initial image of the component; /(I)Is a normalization function; /(I)As a hyperbolic tangent function.

In the formula, considering that the normal temperature change area and the hot spot area possibly exist in the component block area, the probability of containing the hot spot area is different for different component block areas, and the probability of containing the hot spot area in the component block area under the to-be-analyzed scale needs to be analyzed.Representing the variance of gray values in the component block area, reflecting the temperature variation amplitude of the component block area, when/>When the temperature change is larger, the larger the temperature change range in the component block area is, the greater the possibility that the component block area contains a hot spot area is, and the gravity center of judgment is biased to/>，/>The intensity of the temperature change of the component blocking area is reflected, and the larger the intensity is, the larger the probability that the component blocking area contains the hot spot area is. When/>Smaller, a smaller magnitude of temperature change within the component segmented region is indicated, the less likely that the component segmented region will contain hot spot regions, at this point in time, pass/>It is difficult to accurately determine the probability that the component-diced region contains hot spot regions, considering that in a panel, the temperature of the normal region is typically low and relatively concentrated,/>Can reflect the gray value of the normal region, and the center of gravity of the judgment is biased，/>The gray value difference between the hot spot area and the normal area is reflected, and the larger the difference is, the larger the probability that the component blocking area contains the hot spot area is. By/>And/>To adjust/>And/>The weight and the abnormal probability value can more accurately reflect the probability of the hot spot region included in the component block region under the scale to be analyzed.

Considering that the calculated color distance of the saliency detection CA algorithm is often in the Lab color space, the image acquired by the infrared thermal imager in the embodiment of the invention is a single-channel image, and the single-channel image cannot be converted into the Lab color space, so that the saliency detection CA algorithm in the prior art adopts a gray image in the calculated color distance, and the saliency detection CA algorithm takes the average value of gray values of the partitioned areas under a single scale as the gray value of the areas and further acquires the color distance of the two areas under the single scale according to the Euclidean distance of the gray values of the areas of the two areas. The invention reconstructs a calculation method of the color distance in order to enable the color distance to more effectively reflect the degree of distinction between the hot spot area and the normal area on the color characteristics.

Preferably, in one embodiment of the present invention, when an abnormal probability value of a component block area is larger, it is considered that a probability that a hot spot portion is included in the component block area is larger, a gray value of the hot spot portion is higher, a calculated duty ratio of a pixel point with a high gray value is amplified, an area gray value of the component block area under a scale to be analyzed is obtained, the area gray value can improve a significance of the hot spot area, so that a difference between the area gray values corresponding to the hot spot area and a normal area is larger, and an obtaining method of the area gray value includes:

; wherein/> To be at the scale to be analyzed, the first/>Regional gray values of individual component block regions; /(I)To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)To be at the scale to be analyzed, the first/>First/>, in individual component chunking areasGray values of the individual pixels; /(I)To be at the scale to be analyzed, the first/>The total number of all pixel points in the individual component block area; /(I)Background gray values for the initial image of the component; /(I)Is a normalization function.

In the formula, in order to make the region gray value of the weak hot spot region be different from the normal region, when the abnormal probability value is larger, the probability of including the hot spot region in the component block region is larger, when the region gray value is calculated, the duty ratio of the pixel point with high gray value can be properly amplified, when the abnormal probability value is smaller, the probability of including the hot spot region in the component block region is smaller, and when the region gray value is calculated, the gray average value can be calculated normally. When (when)The larger the probability that a region of a component partition in the scale to be analyzed includes a hot spot region is, the larger the gravity center of the judgment tends to be，/>The duty ratio of the pixel points of the high gray value can be amplified, so that the whole gray value of the component blocking area simultaneously makes the difference between the hot spot area and the normal area larger. When/>The smaller the probability that a region of a component partition in the scale to be analyzed includes a hot spot region is, the smaller the gravity center of the judgment tends to/>，/>The overall gray value of the component blocked area can be characterized. The region gray value characterizes the overall gray value of the component blocked region and makes the difference between the region gray value of the hot spot region and the region gray value of the normal region larger.

Preferably, in one embodiment of the present invention, in order to embody the degree of distinction between two component block areas on the color feature, the method for obtaining the color distance includes:

and calculating the absolute value of the difference value of the regional gray values of the two component blocking regions to obtain the color distance of the two component blocking regions under the dimension to be analyzed. Through the steps, the regional gray value is constructed, so that the regional gray value of the weak hot spot region is different from that of the normal region, and the color distance is beneficial to more effectively reflecting the degree of distinction between the hot spot region and the normal region on the color characteristics.

In one embodiment of the present invention, the color distance obtaining formula includes:

; wherein/> To be at the scale to be analyzed, the first/>Individual component blocking area and/>Color distance between individual component tile areas; /(I)To be at the scale to be analyzed, the first/>Regional gray values of individual component block regions; /(I)To be at the scale to be analyzed, the first/>Regional gray values for individual component tile regions.

In order to effectively reflect the degree of distinction between the hot spot region and the normal region in the spatial position, it is first necessary to obtain the initial spatial distance of the two component block regions at the scale to be analyzed. Specifically, based on a significance detection CA algorithm, the spatial distance of the two component block areas under the scale to be analyzed is obtained, and the spatial distance is taken as an initial spatial distance. It should be noted that the significance detection CA algorithm is a technical means well known to those skilled in the art, and only a brief step of obtaining the initial spatial distances of two component block areas under the scale to be analyzed is described herein: and calculating Euclidean distance of positions of the two component blocking areas, acquiring the spatial distance of the two component blocking areas under the scale to be analyzed, and taking the spatial distance as the initial spatial distance.

Preferably, the initial spatial distance of the two component block areas under the scale to be analyzed is obtained in consideration of the Euclidean distance between the two areas in the existing significance detection CA algorithm. In order to effectively reflect the distinction between the hot spot area and the normal area in the space position, the invention considers that the hot spot area is mainly caused by internal short circuit due to the fact that part of battery plates are blocked and can not work, wherein the blocking part is relatively concentrated in the space of the image position, and further adjusts the initial space distance of the two component blocking areas, and obtains the adjusted space distance of the two component blocking areas under the dimension to be analyzed. In one embodiment of the present invention, the method for obtaining the adjustment spatial distance includes:

For subsequent analysis of the surroundings of the component area, in one embodiment of the present invention, the component area is used as the target processing area, and all the component areas adjacent to the target processing area are used as the preset neighborhood range of the target processing area. And in a preset neighborhood range of the component blocking area, taking each component blocking area as each reference blocking area of the component blocking area, wherein each reference blocking area represents each surrounding area of the component blocking area.

Calculating the absolute value of the difference value of the color distance between the component block area and the reference block area to obtain the characteristic difference value of the component block area corresponding to the reference block area; calculating the average value of the characteristic difference values of the component block areas corresponding to all the reference block areas, and obtaining the color significance value of the component block areas; the component tile color saliency value reflects the degree of color difference of the component tile as compared to the surrounding.

; wherein/> To be at the scale to be analyzed, the first/>Individual component blocking area and/>Adjusting the spatial distance between the component block areas; /(I)To be at the scale to be analyzed, the first/>Color saliency values for individual component tile areas; /(I)To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)To be at the scale to be analyzed, the first/>Individual component blocking area and/>An initial spatial distance between the component chunk regions; /(I)Is an exponential function based on e; /(I)As a hyperbolic tangent function.

In the method, in the process of the invention,To be at the scale to be analyzed, the first/>The color saliency value of each component block area reflects the color difference degree of the component block area compared with the surrounding area, and the greater the difference degree, the greater the local special degree representing the component block area, the more likely the component block area comprises the hot spot area, in consideration of the fact that the pixel performance of the hot spot area on the local part of the image is usually relatively concentrated. Abnormal probability value/>, taking into account strong hot spot regionAbnormal probability value/>, of weak hot spot regionAbnormal probability value of normal region, when/>Greater than/>At the time of need for magnification/>The method is favorable for better showing the distinguishing degree of the hot spot area and the normal area.Will/>The value ranges of (2) are normalized in proportion to the value ranges of [0,2], so that when/>Greater than/>At the time/>The value is (1, 2) representing the/>Individual component chunking area ratio/>The abnormal probability of the individual component block areas is large, and the magnification/>Is a ratio of (c) to (d). So that proper/>Less than/>At the time/>The value is [0,1], and the shrinking/>Is a ratio of (c) to (d). Finally make/>The larger represents the/>The greater the likelihood that the individual component tile areas include hot spot areas. Due to/>The dissimilarity metric value is in negative correlation, through/>Modulation/>Logic relationship of (i.e./>)The greater the local degree of specificity of the representative component chunk region of the component chunk region, and the/>Individual component chunking area ratio/>The greater the probability of anomalies in individual component blocked areas, the more desirable it is to narrow/>, when computing salient parametersThe adjustment of the spatial distance can more effectively reflect the division between the hot spot region and the normal region in the spatial position.

Step S3, under the scale to be analyzed, obtaining non-similarity measurement values of the two component blocking areas according to the adjustment space distances of the two component blocking areas and the color distances of the two component blocking areas; and acquiring a single-scale significance value of the component blocking area according to the difference of the abnormal probability values between the component blocking area and all the component blocking areas.

The dissimilarity measure can reflect the degree of difference between the two component blocked areas at the scale to be analyzed. In order to ensure the detection accuracy of the weak hot spot region, in the process of calculating the single-scale saliency value of the two component block regions, the weight of the abnormal probability value which is larger than the dissimilarity metric value of the component block region in the weighted average value calculation process is reduced, so that the influence of the strong hot spot region in the single-scale saliency value calculation process is reduced, the salient difference between the self region and the normal background region is amplified, and the single-scale saliency value can reflect the salient degree of the sufficient weak hot spot region.

Preferably, in one embodiment of the present invention, the method for obtaining the dissimilarity measure value includes:

According to the adjustment space distance of the two component blocking areas and the color distance of the two component blocking areas, obtaining the dissimilarity metric values of the two component blocking areas; adjusting the spatial distance and the dissimilarity measure to be in negative correlation; the color distance and the dissimilarity measure are positively correlated.

Obtaining a dissimilarity measure according to a dissimilarity measure formula, wherein the obtaining formula of the dissimilarity measure comprises:

; wherein, the first/>, is at the scale to be analyzed Individual component blocking area and/>A non-similarity measure between the component partitioned areas; /(I)To be at the scale to be analyzed, the first/>Individual component blocking area and/>Color distance between individual component tile areas; /(I)To be at the scale to be analyzed, the first/>Individual component blocking area and/>The adjustment spatial distance between the individual component block areas.

The color distance which can better show the distinguishing degree of the two component block areas on the color characteristics and the adjustment space distance which can better show the distinguishing degree of the two component block areas on the space position are constructed through the steps, so that the dissimilarity metric value can reflect the difference degree between the two component block areas under the dimension to be analyzed.

Preferably, in one embodiment of the present invention, the method for obtaining the single-scale saliency value includes:

; wherein/> To be at the scale to be analyzed, the first/>Individual component blocking area single-scale significance values; /(I)To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)To be at the scale to be analyzed, the first/>Abnormal probability values for individual component block areas; /(I)The total number of blocked areas for all components; /(I)To be at the scale to be analyzed, the first/>Individual component blocking area and/>A non-similarity measure between the component partitioned areas; /(I)Is an exponential function based on e; /(I)Is a normalization function.

In the formula, in order to ensure the detection probability of the weak hot spot area, the detection probability is calculated byAs/>Reducing the abnormal probability value to be greater than the first/>Weights of non-similarity measurement values of individual component block areas in a weighted average calculation process, so that influence of strong hot spot areas in a single-scale significance value calculation process is reduced, and the/>The significant difference between the individual component blocking area and the normal background area ensures that the single-scale significant value can reflect the significant degree of the full weak hot spot area, thereby constructing the single-scale significant value which can effectively reflect the difference between the weak hot spot area and the normal area, and the single-scale significant value can pertinently reflect the significant degree of the weak hot spot area.

Step S4, obtaining a component saliency map corresponding to the component initial image by using a saliency detection CA algorithm according to single-scale saliency values of all component blocking areas under all preset scales; and obtaining a fault detection result of the photovoltaic module according to the module saliency map.

The single-scale saliency values of all the component blocking areas under all the preset scales can reflect the defect display saliency condition of the component blocking areas under each preset scale, and the single-scale saliency values can reflect the saliency degree of the full weak hot spot areas, so that the saliency detection effect of the component saliency map is improved, and the accuracy of fault detection results is improved.

Preferably, in one embodiment of the present invention, a single-scale saliency value capable of reflecting a saliency degree of a sufficiently weak hot spot region is constructed through the above steps, so as to improve a saliency detection effect of a component saliency map, and an acquisition method of the component saliency map includes:

taking the average value of the single-scale significance values of the component block areas under all preset scales as the average significance value of each component block area;

And acquiring a component saliency map corresponding to the component initial image by using a saliency detection CA algorithm according to the average saliency values of all component blocking areas. It should be noted that, using the saliency detection CA algorithm to obtain the component saliency map corresponding to the component initial image according to the average saliency value of all component blocking areas is a technical means well known to those skilled in the art, and will not be described in detail herein.

The component saliency map of the photovoltaic component can be obtained, and the component saliency map is a gray scale map, wherein the region saliency of the higher brightness is higher.

Specifically, according to the component saliency map, a fault detection result of the photovoltaic component is obtained. And marking the pixel points with the significant values not smaller than the preset judging threshold value as fault pixel points, and marking the rest pixel points as normal pixel points. In one embodiment of the present invention, the preset judgment threshold is 0.8. And calculating the duty ratio of the fault pixel point to all the pixel points in the component saliency map to obtain the defect parameters. When the defect parameter is smaller than the first set parameter, judging that the quality detection grade of the photovoltaic module is excellent; when the defect parameter is not smaller than the first set parameter and smaller than the second set parameter, judging that the quality detection grade of the photovoltaic module is good; when the defect parameter is not smaller than the second setting parameter and smaller than the third setting parameter, judging that the quality detection grade of the photovoltaic module is qualified; and when the defect parameter is not smaller than the third setting parameter, judging that the quality detection grade of the photovoltaic module is unqualified, and needing to be reworked. In this embodiment, the value of the first setting parameter is set to 0.15, the value of the second setting parameter is set to 0.35, and the value of the third setting parameter is set to 0.45, so that the operator can set the setting according to the implementation scenario. So far, determining the fault detection result of the photovoltaic module.

The embodiment also provides a new energy photovoltaic module fault detection system, which comprises a memory, a processor and a computer program, wherein the memory is used for storing the corresponding computer program, the processor is used for running the corresponding computer program, and the computer program can realize any one step of the new energy photovoltaic module fault detection method when running on the processor.

In summary, the embodiment of the invention firstly acquires an initial image of a component of the photovoltaic component, and determines the color distance of two component blocking areas according to the gray value distribution difference between the two component blocking areas, the abnormal probability value of the component blocking areas and the background gray value; acquiring the adjustment space distance of the two component block areas according to the difference of the abnormal probability values and the initial space distance of the two component block areas and the difference of the color distances between the component block areas and all the component block areas in the preset neighborhood range; thereby obtaining a single-scale significance value of the component blocking area; and obtaining a fault detection result of the photovoltaic module according to the module saliency map. According to the invention, the single-scale significance value capable of effectively reflecting the difference between the weak hot spot area and the normal area is constructed, so that the significance detection effect of the component significance map is improved, and the accuracy of the fault detection result is improved.

It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.

Claims

1. The new energy photovoltaic module fault detection method is characterized by comprising the following steps of:

2. The method for detecting the failure of the new energy photovoltaic module according to claim 1, wherein the method for acquiring the abnormal probability value comprises the following steps:

3. The new energy photovoltaic module fault detection method according to claim 1, wherein the color distance acquisition method comprises the following steps:

4. The method for detecting the failure of the new energy photovoltaic module according to claim 3, wherein the method for acquiring the regional gray value comprises the following steps:

5. The new energy photovoltaic module fault detection method according to claim 1, wherein the obtaining method for adjusting the spatial distance comprises the following steps:

; wherein, For/>The component is divided into areas and the/>Adjusting the space distance between the component blocking areas; /(I)For/>Color saliency values for each of the component blocked areas; /(I)For/>The abnormal probability values for the component chunk areas; /(I)For/>The abnormal probability values for the component chunk areas; /(I)For/>The component is divided into areas and the/>The initial spatial distance between the component chunk regions; /(I)Is an exponential function based on e; /(I)As a hyperbolic tangent function.

6. The method for detecting the failure of the new energy photovoltaic module according to claim 1, wherein the method for obtaining the dissimilarity measure value comprises the following steps:

7. The method for detecting the fault of the new energy photovoltaic module according to claim 1, wherein the method for acquiring the single-scale significance value comprises the following steps:

8. The method for detecting the failure of the new energy photovoltaic module according to claim 1, wherein the method for acquiring the module saliency map comprises the following steps:

9. The method for detecting the failure of the new energy photovoltaic module according to claim 1, wherein the method for acquiring the background gray value comprises the following steps:

10. A new energy photovoltaic module fault detection system comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the processor when executing the computer program performs the steps of the method according to any one of claims 1 to 9.