Digital intelligence risk detection method and equipment for high-conversion perovskite battery, and medium

Abstract

The embodiment of the invention discloses a digital intelligent risk detection method and equipment for a high-conversion perovskite battery, and a medium, belongs to the technical field of battery detection, and solves the problems of energy waste and production interruption caused by maintenance after the high-conversion perovskite battery breaks down. Carrying out batch fusion on multi-scale image layers corresponding to the high-conversion perovskite battery image, and carrying out temperature labeling on the fused image through a temperature mapping algorithm to generate a defect thermodynamic diagram; performing feature vector extraction on the defect thermodynamic diagram to determine a defect grade corresponding to the high-conversion perovskite battery based on a feature vector; under the condition that the defect grade is greater than a preset grade threshold value, converting defect area information in the high-conversion perovskite battery into a control signal, and adjusting a shooting light source and a shooting angle so as to determine new defect area information; and performing path evolution on battery defects through a digital twinborn model, and obtaining risk prediction information of the high-conversion perovskite battery based on an evolution result.

Description

Technical Field

[0001] The present application relates to the field of battery detection technology, and in particular to a digital risk detection method, equipment and medium for high-conversion peroxia batteries. Background Art

[0002] In the current energy sector, high-conversion perovskite cells (PCTs) have attracted significant attention due to their exceptional photoelectric performance, becoming a key area of development for solar cell technology. However, in their practical application and maintenance, the associated detection technologies face numerous challenges, severely restricting their efficient and stable operation.

[0003] Currently, testing of high-conversion perovskite batteries is often limited to repairs only after a noticeable fault has occurred. This post-fault testing model has significant drawbacks. Because it's difficult to detect potential risks during battery operation, early-stage faults aren't effectively addressed. As the problem worsens, it can lead to severe performance degradation or even complete failure, resulting in significant energy waste and production disruptions. Summary of the Invention

[0004] The embodiments of the present application provide a digital risk detection method, equipment and medium for high-conversion calcium titanium batteries, which are used to solve the following technical problems: the existing technology usually only performs maintenance after obvious faults occur in high-conversion calcium titanium batteries, resulting in a large amount of energy waste and production interruption.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] The present application provides a digital risk detection method for high-conversion perovskite batteries. The method comprises: performing multi-scale decomposition on infrared images and visible light images of the acquired high-conversion perovskite battery to obtain a multi-scale image layer; fusing the multi-scale image layer in batches, and temperature-labeling the fused image using a temperature mapping algorithm to generate a defect heat map; extracting feature vectors from the defect heat map, and comparing the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; when the defect level is greater than a preset level threshold, converting the defect area information in the high-conversion perovskite battery into a control signal, adjusting the shooting light source and shooting angle, and determining new defect area information based on the adjusted re-shot image; inputting the defect heat map and the new defect area information into a digital twin model of the high-conversion perovskite battery, performing path evolution on the battery defects, and obtaining risk prediction information for the high-conversion perovskite battery based on the evolution results.

[0007] The embodiment of the present application analyzes the image at different levels through a multi-scale decomposition process, which can highlight the key features in the image. Based on the adjusted and re-shot image, new defect area information is determined to form a closed-loop feedback, which can automatically optimize the detection process according to the previous detection results and continuously iterate to improve the detection accuracy. The model is used to simulate the path evolution of battery defects, which can dynamically predict the development trend of battery defects. The possible failure risk of the battery is predicted in advance to provide sufficient time for timely maintenance measures. Based on the risk prediction information obtained from the evolution results, it is possible to reasonably arrange maintenance plans, give priority to high-risk batteries, optimize resource allocation, reduce the probability of battery defects from the source, and improve the overall quality and reliability of the battery.

[0008] In one implementation of the present application, the acquired infrared image and visible light image of the high-conversion perovskite battery are subjected to multi-scale decomposition to obtain a multi-scale image layer, specifically including: matching the infrared image and the visible light image with the preset structural template image of the high-conversion perovskite battery, and dividing the infrared image and the visible light image into multiple regions based on the matching results; matching the corresponding filter kernel parameters based on the high-conversion perovskite battery structural features corresponding to each region to filter different regions separately; combining the filtered regions to obtain a filtered high-conversion perovskite battery image; subtracting the infrared image and the visible light image from the corresponding filtered high-conversion perovskite battery image to obtain the first layer of high-frequency detail image of the Laplace pyramid; repeating the filtering process and image subtraction process to construct a multi-scale Laplace pyramid corresponding to the high-conversion perovskite battery.

[0009] In one implementation of the present application, multi-scale image layers are fused in batches, specifically including: grouping the multi-scale image layers based on the layer sequence number of the multi-scale Laplacian pyramid to obtain multiple groups of images with different frequencies; based on the importance differences corresponding to different regions of the high-conversion peroxia battery, the weights corresponding to different regions of each image layer are determined by image entropy and gradient information; and based on the weights, multiple groups of images with different frequencies are fused.

[0010] In one implementation of the present application, the fused image is temperature-labeled through a temperature mapping algorithm to generate a defect heat map, specifically including: determining the temperature adjustment parameters based on the current ambient temperature and the thermal radiation characteristics of the battery; determining the temperature values corresponding to different areas of the high-conversion perotanium battery based on the thermal radiation intensity corresponding to each pixel point in the infrared image, the preset battery temperature function and the temperature adjustment parameters; clustering the pixel points in the image area corresponding to the same temperature segment, and assigning different colors to different areas based on the clustering results to generate a defect heat map.

[0011] In one implementation of the present application, the temperature adjustment parameter is determined based on the current ambient temperature and the battery thermal radiation characteristics, specifically including: based on the function:

[0012] C'=C(1+k(Tenv-T0))α;

[0013] Determine the thermal radiation constant correction value; where C' is the correction value; C is the original radiation constant; k is the battery material related correction coefficient; T env is the real-time temperature of the environment; T0 is the reference temperature value; α is the crystal structure anisotropy factor; based on the function:

[0014]

[0015] Determine a new carrier mobility compensation value; where μ is the new carrier mobility compensation value; T env is the real-time temperature of the environment; I is the light intensity variable; μ0 is the carrier mobility at room temperature; E is the activation energy; kB is the Boltzmann constant; β is the light influence coefficient; based on the function:

[0016]

[0017] Determine the thermal conductivity adjustment value; where λ is the thermal conductivity adjustment value; ω i is the distribution weight; m i is the temperature gradient correlation coefficient; n i is the correlation coefficient of ambient temperature change; λ0 i is the thermal conductivity coefficient of each phase at room temperature; γ is the correction factor; based on the thermal radiation constant correction value, the new carrier mobility compensation value and the thermal conductivity adjustment value, the temperature adjustment parameter is determined.

[0018] In one implementation of the present application, based on the thermal radiation intensity corresponding to each pixel point in the infrared image, the preset battery temperature function and the temperature adjustment parameter, the temperature values corresponding to different areas of the high conversion peroxia battery are determined, specifically including: based on the function:

[0019]

[0020] Determine the radiation temperature T of the high conversion perovskite cell surface rad ; Among them, I rad is the corrected thermal radiation intensity; ε is the cell surface emissivity; and σ is the Stefan-Boltzmann constant. The calculated function corresponding to the radiation temperature on the surface of the high-conversion perovskite cell is used as the optimized preset cell temperature function. The cell area is divided into multiple small cells, and the temperature values of different cell regions are obtained through iterative processing based on the preset cell temperature function and the temperature adjustment parameters.

[0021] In one implementation of the present application, the defect area information in the high-conversion perovskite battery is converted into a control signal, and the shooting light source and the shooting angle are adjusted to determine the new defect area information based on the adjusted re-shot image, specifically including: inputting the defect heat map into a preset defect recognition model to output the probability of the defect area corresponding to each pixel point in the defect heat map, so as to obtain the defect area information based on the defect area probability; wherein the defect area information includes at least the defect position, defect type, defect area and temperature excess value; determining the shooting angle adjustment parameters based on the current lighting information, defect position and defect area; and determining the shooting light source adjustment parameters based on the defect type and temperature excess value; re-shooting the defect area in the high-conversion perovskite battery based on the shooting angle adjustment parameters and the shooting light source adjustment parameters to obtain new defect area information.

[0022] In one implementation of the present application, the defect heat map and the new defect area information are input into the digital twin model of the high-conversion perotitanium battery, the battery defects are path-evolved, and the risk prediction information of the high-conversion perotitanium battery is obtained based on the evolution results, specifically including: mapping the temperature distribution information and the new defect area information in the defect heat map to the corresponding units of the digital twin model of the high-conversion perotitanium battery, so as to perform battery path evolution based on the digital twin model of the high-conversion perotitanium battery; extracting operating status indicators based on the battery path evolution, and constructing a battery operating status line graph based on the extracted operating status indicators; in the battery operating status line graph, marking the indicator threshold corresponding to each operating status indicator; determining the corresponding risk level based on the difference between each operating status indicator and the indicator threshold.

[0023] An embodiment of the present application provides a digital risk detection device for a high-conversion perovskite battery, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so as to enable the at least one processor to: perform multi-scale decomposition of acquired infrared images and visible light images of the high-conversion perovskite battery to obtain multi-scale image layers; fuse the multi-scale image layers in batches, and temperature-label the fused images through a temperature mapping algorithm to generate a defect heat map; extract feature vectors from the defect heat map, and compare the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; when the defect level is greater than a preset level threshold, convert the defect area information in the high-conversion perovskite battery into a control signal, adjust the shooting light source and shooting angle, and determine new defect area information based on the adjusted re-shot image; input the defect heat map and the new defect area information into a digital twin model of the high-conversion perovskite battery, perform path evolution on the battery defects, and obtain risk prediction information of the high-conversion perovskite battery based on the evolution results.

[0024] A non-volatile computer storage medium provided in an embodiment of the present application stores computer-executable instructions, and the computer-executable instructions are configured to: perform multi-scale decomposition on the acquired infrared image and visible light image of the high-conversion perovskite battery to obtain a multi-scale image layer; fuse the multi-scale image layer in batches, and temperature-label the fused image through a temperature mapping algorithm to generate a defect heat map; extract feature vectors from the defect heat map, and compare the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; when the defect level is greater than a preset level threshold, convert the defect area information in the high-conversion perovskite battery into a control signal, adjust the shooting light source and shooting angle, and determine new defect area information based on the adjusted re-shot image; input the defect heat map and the new defect area information into a digital twin model of the high-conversion perovskite battery, perform path evolution on the battery defects, and obtain risk prediction information of the high-conversion perovskite battery based on the evolution results.

[0025] At least one of the above-mentioned technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: The embodiments of the present application analyze the image at different levels through a multi-scale decomposition process, which can highlight the key features in the image. New defect area information is determined based on the re-taken image after adjustment, forming a closed-loop feedback, which can automatically optimize the detection process according to the previous detection results, and continuously iterate to improve the detection accuracy. By using the model to simulate the path evolution of battery defects, the development trend of battery defects can be dynamically predicted. The possible failure risk of the battery is predicted in advance to provide sufficient time for timely maintenance measures. Based on the risk prediction information obtained from the evolution results, it is possible to reasonably arrange maintenance plans, give priority to high-risk batteries, optimize resource allocation, reduce the probability of battery defects from the source, and improve the overall quality and reliability of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0027] Figure 1 A flow chart of a digital risk detection method for a high-conversion perotanium battery provided in an embodiment of the present application;

[0028] Figure 2 A schematic structural diagram of a digital risk detection device for high-conversion perotanium batteries provided in an embodiment of the present application.

[0029] Reference numerals:

[0030] 200: Digital risk detection equipment for high-conversion perotanium batteries, 201: Processor, 202: Memory. DETAILED DESCRIPTION

[0031] The embodiments of the present application provide a digital risk detection method, device and medium for high-conversion peroxia batteries.

[0032] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0033] The technical solutions proposed in the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0034] Figure 1 A flow chart of a digital risk detection method for a high-conversion perotanium battery provided in an embodiment of the present application is shown in FIG. Figure 1 As shown in the figure, the digital risk detection method for high-conversion perotanium batteries includes the following steps:

[0035] Step 101: Perform multi-scale decomposition on the acquired infrared image and visible light image of the high-conversion peroxia battery to obtain a multi-scale image layer.

[0036] In one implementation of the present application, the infrared image and the visible light image are matched with a preset high-conversion perovskite battery structural template image, and the infrared image and the visible light image are divided into multiple regions based on the matching results. Based on the high-conversion perovskite battery structural features corresponding to each region, the corresponding filter kernel parameters are matched to filter the different regions separately. The filtered regions are combined to obtain a filtered high-conversion perovskite battery image. The infrared image and the visible light image are subtracted from the corresponding filtered high-conversion perovskite battery image to obtain the first-level high-frequency detail image of the Laplacian pyramid. The filtering process and image subtraction process are repeated to construct a multi-scale Laplacian pyramid corresponding to the high-conversion perovskite battery.

[0037] Specifically, a preset structural template image of a high-conversion perovskite battery is obtained. The template image contains standard feature information of the structure of each part of the battery, such as the shape and position distribution of the electrode, active layer, substrate, and other regions. The collected infrared image and visible light image are matched with the structural template image respectively. Through matching, the positions in the image corresponding to each structural region in the template image are determined. Then, based on the matching results, the infrared image and visible light image are divided into multiple regions, each region corresponding to a specific structural part of the battery. For example, the corresponding electrode region in the image is divided into one region, and the active layer region is divided into another region.

[0038] Furthermore, the high-conversion perovskite battery structural features in different regions have different image characteristics. For example, the electrode region may have more metal textures and is more sensitive to high-frequency noise; while the active layer region may be more concerned with the overall temperature distribution and uniformity, and has higher requirements for retaining low-frequency information. Based on the high-conversion perovskite battery structural features corresponding to each region, the corresponding filter kernel parameters are matched for each region. For the electrode region, a high-frequency enhancement filter kernel may be selected to highlight the metal texture details and reduce noise interference; for the active layer region, a low-pass filter kernel is selected to smooth the image, retain the main temperature distribution information, and remove high-frequency noise. Then, these matched filter kernel parameters are used to perform filtering on different regions separately.

[0039] Furthermore, the filtered regions are reassembled according to the positional relationship of the previously divided regions to form a complete image. This image is the filtered image of the high-conversion perovskite battery, which retains the characteristics of each structural region of the battery while reducing noise and unnecessary high-frequency or low-frequency interference.

[0040] Furthermore, the infrared image is subtracted from the corresponding filtered high-conversion perovskite cell image to obtain the first-level high-frequency detail image of the Laplacian pyramid. Specifically, the filtered image retains low-frequency information, and after subtracting the filtered image from the original image, what remains is the high-frequency detail. The same operation is performed on the visible light image to obtain its corresponding first-level high-frequency detail image. The filtered high-conversion perovskite cell image is then downsampled and filtered again to obtain a lower-resolution filtered image. The image subtraction process is repeated to obtain the second-level high-frequency detail image. By repeating this process, a multi-scale Laplacian pyramid corresponding to the high-conversion perovskite cell can be constructed, with each level containing high-frequency detail information at a different scale.

[0041] Step 102: The multi-scale image layers are batch-fused, and the fused images are temperature-labeled using a temperature mapping algorithm to generate a defect heat map.

[0042] In one implementation of this application, multiscale image layers are grouped based on the layer sequence of a multiscale Laplacian pyramid, generating multiple sets of images of different frequencies. Based on the importance differences corresponding to different regions of a high-conversion perovskite cell, weights corresponding to different regions of each image layer are determined using image entropy and gradient information. Based on these weights, the multiple sets of images of different frequencies are fused.

[0043] Specifically, the multi-scale image layers are grouped based on their layer numbers, taking into account that adjacent layers often have similar frequency characteristics and levels of detail. Typically, several adjacent layers are grouped together. For example, the first and second layers can be grouped together, representing relatively high-frequency details; the third and fourth layers can be grouped together, representing medium-frequency information; and so on.

[0044] Furthermore, different regions of high-conversion perovskite cells have varying degrees of importance. For example, the electrode region plays a key role in the cell's electrical performance, while the active layer region directly impacts the cell's photoelectric conversion efficiency. To highlight information from important regions during image fusion, image entropy and gradient information are used to determine weights for different regions within each image layer. Image entropy reflects the amount of information in an image; higher entropy values indicate richer information within that region. Gradient information reflects the degree of pixel variation within an image; larger gradients indicate more distinct edges or details within that region. For highly important regions, such as the electrode region, if their image entropy and gradient values are large within a given image layer, this indicates that they contain critical information at that scale and should be assigned a higher weight. For less important regions, such as the cell base, if their image entropy and gradient values are small, they should be assigned a lower weight. In this way, weights reflecting their importance are assigned to different regions within each image layer. After image layer grouping and weight determination, multiple sets of images with different frequencies are fused. For each set of images with different frequencies, the pixel values of the corresponding regions within the image are weighted and summed according to the previously determined weights for each region.

[0045] In one implementation of this application, a temperature adjustment parameter is determined based on the current ambient temperature and the battery's thermal radiation characteristics. The temperature values corresponding to different regions of the high-conversion perovskite battery are determined based on the thermal radiation intensity corresponding to each pixel in the infrared image, a preset battery temperature function, and the temperature adjustment parameter. Pixels in the image regions corresponding to the same temperature range are clustered, and different regions are assigned different colors based on the clustering results to generate a defect heat map.

[0046] Specifically, based on the thermal radiation characteristics of the battery, a functional relationship between thermal radiation and temperature and environmental conditions is established. However, in actual high-conversion perovskite batteries, thermal radiation is also constrained by other factors due to the complex characteristics of the material, such as anisotropy and multiphase structure. As the ambient temperature changes, the heat conduction and carrier migration processes within the battery will also change, thereby affecting the thermal radiation characteristics.

[0047] Furthermore, the overall temperature adjustment parameter determination method of the embodiment of the present application is:

[0048] Function-based:

[0049] C'=C(1+k(Tenv-T0))α;

[0050] Determine the thermal radiation constant correction value; where C' is the correction value; C is the original radiation constant; k is the battery material related correction coefficient; T env is the real-time temperature of the environment; T0 is the reference temperature value; α is the crystal structure anisotropy factor;

[0051] Function-based:

[0052]

[0053] Determine a new carrier mobility compensation value; where μ is the new carrier mobility compensation value; T env is the real-time temperature of the environment; I is the light intensity variable; μ0 is the carrier mobility at room temperature; E is the activation energy; kB is the Boltzmann constant; β is the light influence coefficient;

[0054] Function-based:

[0055]

[0056] Determine the thermal conductivity adjustment value; where λ is the thermal conductivity adjustment value; ω i is the distribution weight; m i is the temperature gradient correlation coefficient; n i is the correlation coefficient of ambient temperature change; λ0 i is the thermal conductivity coefficient of each phase at room temperature; γ is the correction factor;

[0057] A temperature adjustment parameter is determined based on the thermal radiation constant correction value, the new carrier mobility compensation value, and the thermal conductivity adjustment value.

[0058] Furthermore, the embodiment of the present application is based on the function:

[0059]

[0060] Determine the radiation temperature T of the high conversion perovskite cell surface rad ; Among them, I rad is the corrected thermal radiation intensity; ε is the surface emissivity of the battery; σ is the Stefan-Boltzmann constant.

[0061] The calculated function corresponding to the radiation temperature on the surface of the high-conversion perovskite battery is used as the optimized preset battery temperature function. The battery area is divided into multiple small units, and based on the preset battery temperature function and temperature adjustment parameters, the temperature values of different battery areas are obtained through iterative processing.

[0062] Specifically, the present embodiment introduces a temperature adjustment parameter to correct the emissivity, taking into account the changes in the surface properties of the battery material under different ambient temperatures. The thermal radiation intensity of each pixel in the infrared image is then substituted into the corrected temperature function to calculate the temperature value corresponding to that pixel. Because the thermal radiation characteristics and structure of different regions of the high-conversion perovskite battery may vary, the temperature value calculated for each pixel represents the temperature of the corresponding region of the battery.

[0063] Furthermore, after obtaining the temperature values for different regions of the high-conversion perovskite battery, the pixels in the image regions corresponding to the same temperature range are clustered. Using a clustering algorithm, such as the K-means clustering algorithm, several temperature categories are pre-defined, and pixels with similar temperature values are grouped together. Based on the clustering results, different colors are assigned to different cluster regions, presenting the temperature distribution of different battery regions in an intuitive color format, generating a defect heat map.

[0064] Step 103: extract feature vectors from the defect heat map, and compare the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results.

[0065] In one implementation of the present application, the feature extraction method in the embodiment of the present application may include an analysis method based on image texture, shape and statistical characteristics. For texture features, a grayscale co-occurrence matrix algorithm is used. The algorithm extracts texture feature values such as contrast, correlation, energy and entropy by calculating the frequency of occurrence of different grayscale pixel pairs in the image at a specific direction and distance. For example, in a defect thermal map, the texture features of the hot spot area and the normal area are different. The hot spot area may have higher contrast and energy values, reflecting that its temperature changes are drastic and unevenly distributed. In terms of shape feature extraction, a contour detection algorithm is used to obtain the contour of the defect area, and then calculate the shape feature parameters such as the perimeter, area, and circularity of the contour. These parameters can describe the geometric shape of the defect area. For example, the contour of a crack defect may appear elongated, and its perimeter to area ratio is large. In addition, feature extraction methods based on statistical characteristics, such as calculating the mean, variance, skewness and other statistical quantities of the thermal map, can reflect the overall temperature distribution characteristics of the image. The mean can represent the average temperature level of the image, and the variance reflects the degree of discreteness of the temperature distribution. By integrating these different types of features, a multi-dimensional feature vector is constructed to comprehensively and accurately describe the characteristics of the defect heat map.

[0066] Furthermore, the pre-set database in the embodiment of the present application stores a large number of feature vectors of high-conversion perovskite batteries with different types and degrees of defects, as well as corresponding defect level annotation information. The extracted defect heat map feature vector is compared with the feature vectors in the pre-set database. Through the comparison, the feature vectors in several databases with the highest similarity to the extracted feature vector are found. The defect level corresponding to these similar feature vectors is then determined based on the defect levels corresponding to these similar feature vectors.

[0067] Step 104: When the defect level is greater than a preset level threshold, the defect area information in the high-conversion perovskite battery is converted into a control signal, and the shooting light source and shooting angle are adjusted to determine new defect area information based on the adjusted re-shot image.

[0068] In one implementation of the present application, a defect heat map is input into a preset defect recognition model to output the probability of the defect region corresponding to each pixel in the defect heat map. Defect region information is then obtained based on the defect region probability. The defect region information includes at least the defect location, defect type, defect area, and temperature excess value. Shooting angle adjustment parameters are determined based on the current lighting information, defect location, and defect area. Furthermore, shooting light source adjustment parameters are determined based on the defect type and temperature excess value. Based on the shooting angle adjustment parameters and the shooting light source adjustment parameters, the defect region in the high-conversion perovskite battery is re-photographed to obtain new defect region information.

[0069] Specifically, the preset defect recognition model in the embodiment of the present application is a convolutional neural network. During the training phase, the model uses a large number of labeled high-conversion perovskite battery defect heat map samples for learning. These samples are detailed with the defect area to which each pixel belongs, as well as the corresponding defect type, temperature exceedance value and other information. When a new defect heat map is input into the model, the model extracts and analyzes the image features through a series of network structures such as convolution layers and pooling layers, and finally outputs the probability of each pixel corresponding to the defect area. Based on these probabilities, it is determined whether each pixel belongs to the defect area by setting a threshold, etc., and then the outline of the defect area is obtained. The defect position and defect area are calculated by statistical analysis of the pixels within the outline. At the same time, according to the correspondence between the defect features and the defect type and temperature exceedance values learned during the model training process, the defect area is classified, the defect type is determined, and the temperature exceedance value is obtained.

[0070] Furthermore, the position of the defect determines the direction in which the camera needs to be adjusted, and the area of the defect affects the amplitude of the adjustment. For example, if the defect is located at the edge of the battery, the camera needs to be rotated a certain angle toward the edge; if the defect area is large, the shooting angle may need to be adjusted more significantly to ensure that the defect area is completely and clearly presented in the picture. By establishing an illumination model, a camera model, and a relationship model between the defect position and area and the shooting angle, the shooting angle adjustment parameters can be calculated. For example, based on the principle of light propagation and the geometric relationship of camera imaging, combined with the relative relationship between the current illumination direction and the defect position, the angle at which the camera needs to be rotated around the horizontal and vertical axes is calculated to achieve the best shooting effect. Different defect types have different requirements for light sources. For example, for hot spot defects, a stronger light source may be required to highlight the temperature difference; for crack defects, side light may be more conducive to highlighting the outline of the crack.

[0071] The camera is then rotated and positioned according to the adjusted angle, and the light source is set to the adjusted intensity, color, and angle. The defective area in the high-conversion perovskite cell is then re-photographed. Thanks to the optimized shooting angle and lighting conditions, the newly captured image shows the details of the defective area more clearly. The pre-set defect recognition model is then used to analyze the newly captured image again, obtaining even more accurate information about the defective area.

[0072] Step 105: Input the defect heat map and the new defect area information into the digital twin model of the high-conversion peroximate battery, perform path evolution on the battery defects, and obtain risk prediction information of the high-conversion peroximate battery based on the evolution results.

[0073] In one implementation of the present application, the temperature distribution information in the defect thermodynamic map and the new defect area information are mapped to the corresponding units of the high-conversion peroximate battery digital twin model to perform battery path evolution based on the high-conversion peroximate battery digital twin model. Based on the battery path evolution, the operating status indicators are extracted to construct a battery operating status line graph based on the extracted operating status indicators. In the battery operating status line graph, the indicator threshold corresponding to each operating status indicator is marked. Based on the difference between each operating status indicator and the indicator threshold, the corresponding risk level is determined.

[0074] Specifically, the embodiments of the present application establish a spatial correspondence between the digital twin model and the actual battery, usually based on the geometric structure and finite element division of the battery. For example, the digital twin model is divided into many tiny units, each unit corresponding to a tiny area in the actual battery. For temperature distribution information, according to the temperature value corresponding to each pixel in the defect thermodynamic map, it is assigned to the temperature attribute of the corresponding spatial position unit in the digital twin model. For defect area information, if the defect is located in a certain area, the digital twin model unit corresponding to the area is marked as a defective unit, and the corresponding physical parameter adjustment is set according to the defect type. For example, hot spot defects may increase the heat generation rate of the unit, and crack defects may change the mechanical and thermal conductivity characteristics of the unit. In this way, the defect information of the actual battery is accurately implanted into the digital twin model.

[0075] Furthermore, after obtaining the defect information, the model starts the path evolution simulation. In terms of heat conduction, the heat conduction equation is used to calculate the heat transfer inside the battery based on the adjusted heat generation rate of the defect area and the thermal conductivity coefficient between each unit. The change in temperature will affect the mobility of carriers in the electrochemical process, thereby changing the current distribution. In terms of mechanics, for crack defects, their impact on the stress distribution of the battery structure is considered. As the crack expands, it will further affect the heat conduction and electrochemical processes. During the path evolution process, multiple key operating status indicators are extracted, such as electrical performance indicators such as battery voltage, current, and power, thermal performance indicators such as battery average temperature, hot spot temperature, and temperature gradient, as well as structural performance indicators such as crack length and structural stress. These indicators can comprehensively reflect the changes in the operating status of the battery under the influence of defects.

[0076] Furthermore, with time as the horizontal axis and each operating status indicator as the vertical axis, the indicator values extracted at different time points during the path evolution are sequentially connected to form a line graph. For example, for the battery voltage indicator, connecting the voltage values at different times can clearly show the voltage fluctuation trend. At the same time, the indicator threshold is labeled for each operating status indicator. Risk weights are assigned to different operating status indicators, and different indicators have different degrees of impact on battery performance and safety. For example, battery voltage is directly related to the battery's output capacity and is assigned a higher risk weight; while the temperature gradient in a local area of the battery surface has a relatively small impact on overall performance, a lower risk weight can be assigned. Then, the difference between each indicator and the threshold is calculated, and different risk intervals are divided according to the size of the difference. For example, for the voltage indicator, if the voltage is less than 0.7V, the larger the difference, the higher the risk; for the average temperature indicator, if it exceeds 35°C, the larger the excess, the higher the risk. Finally, the overall risk level is determined by comprehensively considering the risk weights and risk intervals of each indicator. In the embodiments of this application, the risk level can be divided into low risk, medium risk, high risk, etc.

[0077] Figure 2 This is a schematic diagram of the structure of a digital risk detection device for a high conversion perotanium battery provided in an embodiment of the present application. Figure 2As shown, the digital risk detection device 200 for high-conversion peroxia batteries includes: at least one processor 201; and a memory 202 in communication with the at least one processor 201; wherein the memory 202 stores instructions that can be executed by the at least one processor 201, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to: perform multi-scale decomposition on the acquired infrared image and visible light image of the high-conversion peroxia battery to obtain a multi-scale image layer; fuse the multi-scale image layer in batches, and annotate the temperature of the fused image through a temperature mapping algorithm to generate a defect heat map; Feature vectors are extracted from the defect heat map, and the extracted feature vectors are compared with the feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; when the defect level is greater than the preset level threshold, the defect area information in the high-conversion perovskite battery is converted into a control signal, and the shooting light source and shooting angle are adjusted to determine the new defect area information based on the re-taken image after adjustment; the defect heat map and the new defect area information are input into the digital twin model of the high-conversion perovskite battery, the battery defects are path evolved, and the risk prediction information of the high-conversion perovskite battery is obtained based on the evolution results.

[0078] A non-volatile computer storage medium provided in an embodiment of the present application stores computer-executable instructions, and the computer-executable instructions are configured to: perform multi-scale decomposition on the acquired infrared image and visible light image of the high-conversion perovskite battery to obtain a multi-scale image layer; fuse the multi-scale image layer in batches, and temperature-label the fused image through a temperature mapping algorithm to generate a defect heat map; extract feature vectors from the defect heat map, and compare the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; when the defect level is greater than a preset level threshold, convert the defect area information in the high-conversion perovskite battery into a control signal, adjust the shooting light source and shooting angle, and determine new defect area information based on the adjusted re-shot image; input the defect heat map and the new defect area information into a digital twin model of the high-conversion perovskite battery, perform path evolution on the battery defects, and obtain risk prediction information of the high-conversion perovskite battery based on the evolution results.

[0079] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0080] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. It will be apparent to those skilled in the art that various modifications and variations may be made to the embodiments of the present application. However, such modifications or substitutions do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A digital risk detection method for high-conversion perotanium batteries, characterized in that: The method comprises: The acquired infrared image and visible light image of the high-conversion perovskite battery are decomposed into multi-scale layers to obtain a multi-scale image layer; The multi-scale image layers are batch-fused, and the fused images are temperature-labeled using a temperature mapping algorithm to generate a defect heat map; Extracting feature vectors from the defect heat map, and comparing the extracted feature vectors with feature vectors in a preset database to determine the defect level corresponding to the high-conversion perovskite battery based on the comparison results; When the defect level is greater than a preset level threshold, the defect area information in the high-conversion perovskite battery is converted into a control signal, and the shooting light source and shooting angle are adjusted to determine new defect area information based on the adjusted re-shot image; The defect heat map and the new defect area information are input into the digital twin model of the high-conversion peroximate battery, the battery defects are subjected to path evolution, and the risk prediction information of the high-conversion peroximate battery is obtained based on the evolution results.

2. The digital risk detection method for high-conversion perotanium batteries according to claim 1 is characterized in that: The infrared image and visible light image of the high-conversion peroxia battery are subjected to multi-scale decomposition to obtain a multi-scale image layer, specifically comprising: Matching the infrared image and the visible light image with a preset high-conversion peroxia battery structure template image respectively, and dividing the infrared image and the visible light image into multiple regions based on the matching results; Based on the high-conversion perovskite battery structural characteristics corresponding to each of the regions, the corresponding filter kernel parameters are matched to perform filtering processing on different regions respectively; The filtered regions are combined to obtain a filtered high-conversion perovskite battery image; Subtracting the infrared image and the visible light image from the corresponding filtered high-conversion perovskite battery image to obtain a first-layer high-frequency detail image of the Laplacian pyramid; The filtering process and the image subtraction process are repeated to construct a multi-scale Laplacian pyramid corresponding to the high-conversion perovskite battery.

3. The digital risk detection method for high-conversion perotanium batteries according to claim 2 is characterized in that: The fusing the multi-scale image layers in batches specifically includes: Grouping the multi-scale image layers based on the layer sequence numbers of the multi-scale Laplacian pyramid to obtain multiple groups of images with different frequencies; Based on the importance differences corresponding to different regions of the high-conversion perovskite battery, the weights corresponding to different regions of each image layer are determined by image entropy and gradient information; Based on the weights, multiple groups of images with different frequencies are fused.

4. The digital risk detection method for high-conversion peroxia batteries according to claim 2 is characterized in that: The temperature mapping algorithm is used to mark the temperature of the fused image and generate a defect heat map, which specifically includes: Determine the temperature adjustment parameters based on the current ambient temperature and the battery's thermal radiation characteristics; Determining the temperature values corresponding to different areas of the high-conversion perovskite battery based on the thermal radiation intensity corresponding to each pixel point in the infrared image, the preset battery temperature function, and the temperature adjustment parameter; The pixel points of the image area corresponding to the same temperature segment are clustered, and different colors are assigned to different areas based on the clustering results to generate the defect heat map.

5. The digital risk detection method for high-conversion perotanium batteries according to claim 4 is characterized in that: The temperature adjustment parameters are determined based on the current ambient temperature and the battery thermal radiation characteristics, specifically including: Function-based: C'=C(1+k(Tenv-T0))α; Determine the thermal radiation constant correction value; where C' is the correction value; C is the original radiation constant; k is the battery material related correction coefficient; T env is the real-time temperature of the environment; T0 is the reference temperature value; α is the crystal structure anisotropy factor; Function-based: Determine a new carrier mobility compensation value; where μ is the new carrier mobility compensation value; T env is the real-time temperature of the environment; I is the light intensity variable; μ0 is the carrier mobility at room temperature; E is the activation energy; k B is the Boltzmann constant; β is the illumination influence coefficient; Function-based: Determine the thermal conductivity adjustment value; where λ is the thermal conductivity adjustment value; ω i is the distribution weight; m i is the temperature gradient correlation coefficient; n i is the correlation coefficient of ambient temperature change; 0i is the thermal conductivity coefficient of each phase at room temperature; γ is the correction factor; The temperature adjustment parameter is determined based on the thermal radiation constant correction value, the new carrier mobility compensation value, and the thermal conductivity adjustment value.

6. The digital risk detection method for high-conversion perotanium batteries according to claim 4 is characterized in that: The determining of the temperature values corresponding to different areas of the high conversion peroxia battery based on the thermal radiation intensity corresponding to each pixel point in the infrared image, the preset battery temperature function, and the temperature adjustment parameter specifically includes: Function-based: Determine the radiation temperature T of the high conversion perovskite cell surface rad ; Among them, I rad is the corrected thermal radiation intensity; ε is the surface emissivity of the battery; σ is the Stefan-Boltzmann constant; The calculated function corresponding to the radiation temperature of the surface of the high conversion peroxia battery is used as the optimized preset battery temperature function; The battery area is divided into a plurality of small units, and the temperature values of different battery areas are obtained through iterative processing based on the preset battery temperature function and the temperature adjustment parameter.

7. The digital risk detection method for high-conversion perotanium batteries according to claim 1 is characterized in that: The step of converting the defect area information in the high-conversion perovskite battery into a control signal and adjusting the shooting light source and shooting angle to determine new defect area information based on the adjusted re-shot image specifically includes: Inputting the defect thermogram into a preset defect recognition model to output the probability of a defect region corresponding to each pixel point in the defect thermogram, thereby obtaining defect region information based on the defect region probabilities; wherein the defect region information includes at least defect location, defect type, defect area, and temperature exceedance; Determining a shooting angle adjustment parameter based on the current lighting information, the defect position, and the defect area; and, determining a shooting light source adjustment parameter based on the defect type and the temperature exceeding value; Based on the shooting angle adjustment parameter and the shooting light source adjustment parameter, the defective area in the high-conversion perovskite battery is re-photographed to obtain the new defective area information.

8. The digital risk detection method for high-conversion perotanium batteries according to claim 1 is characterized in that: The step of inputting the defect heat map and the new defect area information into the high-conversion peroximate battery digital twin model, performing path evolution on the battery defects, and obtaining risk prediction information of the high-conversion peroximate battery based on the evolution results specifically includes: Mapping the temperature distribution information in the defect thermodynamic map and the new defect area information to corresponding units of the high-conversion peroximate battery digital twin model, so as to perform battery path evolution based on the high-conversion peroximate battery digital twin model; Extracting an operating status indicator based on the battery path evolution, and constructing a battery operating status line graph based on the extracted operating status indicator; In the battery operating status line graph, mark the indicator threshold corresponding to each of the operating status indicators; Based on the difference between each of the operating status indicators and the indicator threshold, a corresponding risk level is determined.

9. A digital risk detection device for high-conversion perotanium batteries, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 8.

Images