Solid-state battery safety early warning methods, systems and electronic devices

Abstract

This invention provides a method, system, and electronic device for safety early warning of solid-state batteries, relating to the field of solid-state battery detection. The method obtains multimodal images of negative electrode dendrites by acquiring structural images, morphological images, and labeled fluorescence images, and then performs parameter fusion and feature fusion to obtain the fused image corresponding to the target solid-state battery. This achieves a cross-modal feature complementary fusion process, solving the problems of low detection accuracy and static fusion parameters in traditional single-modal methods, improving the accuracy and real-time performance of dendrite growth monitoring, and enhancing the safety early warning effect of solid-state batteries.

Description

Technical Field

[0001] This invention relates to the field of solid-state battery testing, and in particular to a solid-state battery safety early warning method, system, and electronic device. Background Technology

[0002] Solid-state batteries have become the core direction of next-generation energy storage devices due to their high energy density and excellent safety. However, the interfacial short circuit problem caused by lithium dendrite growth in the negative electrode is a key bottleneck restricting their industrialization. Therefore, monitoring dendrite growth in solid-state batteries is an important means of assessing battery safety, but the existing technology has the following problems:

[0003] 1. Limitations of a single mode: Optical coherence tomography (OCT) can achieve in-situ real-time observation, but its resolution at the micro-nano scale is insufficient; scanning electron microscopy (SEM) can provide high-resolution micro-morphology, but it cannot achieve dynamic monitoring; fluorescence microscopy can label dendrites, but it lacks three-dimensional structural information. A single mode is difficult to fully characterize dendrite growth characteristics.

[0004] 2. Staticization of fusion methods: Traditional multimodal fusion methods (such as wavelet transform) rely on fixed weight parameters, which cannot adapt to the dynamic growth process of dendrites from initiation to penetration, resulting in distortion of fused image features;

[0005] 3. Insufficient real-time performance: Existing systems mostly use offline image processing mode, which cannot be synchronized with the battery charging and discharging process, making it difficult to meet the needs of online safety early warning. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a method, system and electronic device for safety early warning of solid-state batteries. The method obtains multimodal images of negative electrode dendrites by acquiring structural images, morphological images and labeled fluorescence images, and obtains the fused image corresponding to the target solid-state battery by parameter fusion and feature fusion, thereby realizing the cross-modal feature complementary fusion process, solving the problems of low detection accuracy and static fusion parameters in traditional single-mode methods, improving the accuracy and real-time performance of dendrite growth monitoring, and improving the safety early warning effect of solid-state batteries.

[0007] In a first aspect, embodiments of the present invention provide a solid-state battery safety early warning method, the method comprising:

[0008] Image acquisition steps: Acquire the structural image, morphological image, and labeled fluorescence image corresponding to the negative electrode dendrite of the target solid-state battery, and determine the multimodal image corresponding to the target solid-state battery based on the structural image, morphological image, and labeled fluorescence image;

[0009] Preprocessing steps: Gaussian filtering for noise reduction, edge feature extraction, and 3D reconstruction are performed on the multimodal images to obtain the preprocessed images corresponding to the multimodal images;

[0010] Parameter fusion steps: Calculate the image entropy, standard deviation, and structural similarity index of the preprocessed image based on the grayscale value of the preprocessed image, and construct the fusion parameters corresponding to the target solid-state battery based on the image entropy, standard deviation, and structural similarity index;

[0011] Feature fusion steps: Obtain the structural features and detail features corresponding to the preprocessed image, perform regional energy weighted fusion calculation on the structural features using fusion parameters, and perform dynamic weighted fusion calculation on the detail features using fusion parameters to obtain the fused image corresponding to the target solid-state battery;

[0012] Real-time warning steps: Obtain the dendrite region contained in the fused image, calculate the dendrite length and growth rate in the dendrite region in real time, and provide a safety warning for the target solid-state battery based on the dendrite length and growth rate.

[0013] Optional image acquisition steps include:

[0014] Structural images were acquired using an optical coherence tomography (OCT) device corresponding to the target solid-state battery.

[0015] Topographic images were acquired using a scanning electron microscope corresponding to the target solid-state battery.

[0016] Acquire labeled fluorescence images using a fluorescence microscope corresponding to the target solid-state battery;

[0017] Multimodal images of the target solid-state battery are obtained by synchronously sampling the data channels corresponding to the structural image, morphological image, and labeled fluorescence image.

[0018] Optional preprocessing steps include:

[0019] After performing Gaussian filtering denoising on the structural image, the filtered image corresponding to the structural image is obtained;

[0020] The edge features of the topography image are calculated using the Canny operator to obtain the dendrite edge data of the target solid-state battery, and the dendrite edge region contained in the topography image is extracted using the gradient value of the dendrite edge data.

[0021] Obtain the Z-axis scanning sequence corresponding to the labeled fluorescence image, and obtain the three-dimensional structure image corresponding to the labeled fluorescence image based on the maximum intensity projection result of the Z-axis scanning sequence;

[0022] The preprocessed image corresponding to the multimodal image is determined based on the filtered image, dendrite edge region, and three-dimensional structure image.

[0023] Optionally, the filtered image is calculated using the following formula:

[0024] ;

[0025] in, For filtered images; For Gaussian kernel function, And satisfy ; For structural images;

[0026] The dendrite edge region is calculated using the following formula:

[0027] ; ;

[0028] in, For the preset high threshold, The preset low threshold; The gradient value of the dendrite edge data; ; , ; For morphological images;

[0029] The three-dimensional structure image is calculated using the following formula:

[0030] ;

[0031] in, It is a three-dimensional structural image; This is a Z-axis scan sequence. .

[0032] Optional parameter fusion steps include:

[0033] The first fusion weight, the second fusion weight, and the third fusion weight are determined for the structural image, the morphological image, and the labeled fluorescence image, respectively.

[0034] A three-dimensional honey source vector is constructed using the first fusion weight, the second fusion weight, and the third fusion weight;

[0035] The search step size is determined based on the maximum number of iterations of the hired bees corresponding to the artificial bee colony model. The new nectar source corresponding to the three-dimensional nectar source vector is determined by the step size, and the nectar source corresponding to the artificial bee colony model is updated according to the new nectar source.

[0036] Obtain the grayscale values ​​of the preprocessed image, and use the grayscale values ​​to calculate the image entropy, standard deviation, and structural similarity index of the preprocessed image;

[0037] The fitness function corresponding to the preprocessed image is determined based on image entropy, standard deviation, and structural similarity index.

[0038] Based on the roulette wheel beating strategy of the observation bees and the nectar source update strategy of the scout bees corresponding to the artificial bee colony model, the probability of nectar source selection under the fitness function is determined.

[0039] The optimal fusion parameters corresponding to the fitness function are determined by selecting probabilities, and the optimal fusion parameters are used to construct the fusion parameters corresponding to the target solid-state battery.

[0040] Optionally, the fitness function is calculated using the following formula:

[0041] ;

[0042] in, For the adaptive function; A three-dimensional honey source vector; Image entropy; Standard deviation; It is a structural similarity index; , , These are the weight values ​​corresponding to image entropy, standard deviation, and structural similarity index, respectively.

[0043] ; grayscale value The probability distribution;

[0044] ; This is the fused image corresponding to the preprocessed image obtained using grayscale values; The mean of the merged images;

[0045] ;

[0046] in and These represent the mean values ​​of the preprocessed image and the reference image, respectively, and the mean value of the fused image. For covariance, , , It is a grayscale level.

[0047] Optional feature fusion steps include:

[0048] The preprocessed image is decomposed using a non-subsampled shear wave transform to obtain the corresponding low-frequency components. and high frequency components ;

[0049] Low-frequency components are used to determine the structural features corresponding to the preprocessed image, and high-frequency components are used to determine the detailed features corresponding to the preprocessed image.

[0050] By performing region energy-weighted fusion calculation on structural features using fusion parameters, the low-frequency fusion coefficients corresponding to the preprocessed image are obtained. ;in, ; For regional energy weighting, ; For regional energy, ; For modal indexing;

[0051] By performing region energy-weighted fusion calculation on structural features using fusion parameters, the high-frequency fusion coefficients corresponding to the preprocessed image are obtained. ;in, ; The optimal weight values ​​corresponding to the fusion parameters;

[0052] The inverse transform operator used in non-subsampled shear wave transform decomposition is obtained. After performing inverse transform processing on the low-frequency and high-frequency fusion coefficients using the inverse transform operator, a fused image corresponding to the target solid-state battery is obtained; wherein, the fused image... The result is obtained through the following formula; ; It is the inverse transform operator.

[0053] Optional real-time alert steps include:

[0054] The dendritic regions contained in the fused image were determined based on the structural image, morphological image, and labeled fluorescence image corresponding to the negative electrode dendrites.

[0055] The dendrite length and growth rate in the dendrite region are calculated in real time based on the endpoints of the dendrite skeleton within the dendrite region; where, the dendrite length... ; and The starting and ending points of the dendritic framework; growth rate ; Let be the length of the t-th monitoring session. For monitoring intervals;

[0056] The system obtains preset warning threshold conditions, determines the warning level corresponding to the target solid-state battery based on the correspondence between dendrite length and growth rate and the warning threshold conditions, and uses the warning level to provide safety warnings for the target solid-state battery.

[0057] In a second aspect, the present invention provides a solid-state battery safety early warning system, the system comprising:

[0058] Image acquisition unit: used to acquire structural images, morphological images and labeled fluorescence images corresponding to the negative electrode dendrites of the target solid-state battery, and to determine the multimodal image corresponding to the target solid-state battery based on the structural images, morphological images and labeled fluorescence images;

[0059] Preprocessing unit: used to perform Gaussian filtering for noise reduction, edge feature extraction and 3D reconstruction on the multimodal image to obtain the preprocessed image corresponding to the multimodal image;

[0060] Parameter fusion unit: used to calculate the image entropy, standard deviation and structural similarity index of the preprocessed image based on the gray value of the preprocessed image, and to construct the fusion parameters corresponding to the target solid-state battery based on the image entropy, standard deviation and structural similarity index;

[0061] Feature fusion unit: used to obtain structural features and detail features corresponding to the preprocessed image, perform regional energy weighted fusion calculation on the structural features using fusion parameters, and perform dynamic weighted fusion calculation on the detail features using fusion parameters to obtain the fused image corresponding to the target solid-state battery;

[0062] Real-time early warning unit: used to acquire dendritic regions contained in the fused image, calculate dendrite length and growth rate in the dendrite region in real time, and provide safety warnings for the target solid-state battery based on dendrite length and growth rate.

[0063] Thirdly, embodiments of the present invention also provide an electronic device, including a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of the solid-state battery safety warning method provided in the first aspect.

[0064] Fourthly, embodiments of the present invention also provide a storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the steps of the solid-state battery safety warning method provided in the first aspect.

[0065] This invention provides a solid-state battery safety early warning method, system, and electronic device. In the process of providing safety early warning for negative electrode dendrites in a target solid-state battery, the method first acquires structural images, morphological images, and labeled fluorescence images corresponding to the negative electrode dendrites of the target solid-state battery, and determines the multimodal image corresponding to the target solid-state battery based on these images. Then, the multimodal image is processed by Gaussian filtering for noise reduction, edge feature extraction, and 3D reconstruction to obtain a preprocessed image corresponding to the multimodal image. Subsequently, the image entropy, standard deviation, and structural similarity index of the preprocessed image are calculated based on its grayscale values, and fusion parameters corresponding to the target solid-state battery are constructed based on these parameters. Next, structural features and detail features corresponding to the preprocessed image are acquired, and regional energy weighted fusion calculations are performed on the structural features using the fusion parameters. Dynamic weighted fusion calculations are also performed on the detail features using the fusion parameters to obtain a fused image corresponding to the target solid-state battery. Finally, the dendrite region contained in the fused image is acquired, and the dendrite length and growth rate in the dendrite region are calculated in real time. Based on the dendrite length and growth rate, a safety early warning is provided for the target solid-state battery. This method obtains multimodal images of negative electrode dendrites by acquiring structural images, morphological images, and labeled fluorescence images. After parameter fusion and feature fusion, a fused image corresponding to the target solid-state battery is obtained, thereby realizing a cross-modal feature complementary fusion process. This solves the problems of low detection accuracy and static fusion parameters in traditional single-modal methods, improves the accuracy and real-time performance of dendrite growth monitoring, and enhances the safety early warning effect of solid-state batteries.

[0066] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0067] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0068] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0069] Figure 1 A flowchart of a solid-state battery safety early warning method provided in an embodiment of the present invention;

[0070] Figure 2 A flowchart of the image acquisition step S101 of a solid-state battery safety early warning method provided in an embodiment of the present invention;

[0071] Figure 3 A flowchart of the preprocessing step S102 of a solid-state battery safety early warning method provided in an embodiment of the present invention;

[0072] Figure 4 A flowchart of parameter fusion step S103 of a solid-state battery safety early warning method provided in an embodiment of the present invention;

[0073] Figure 5 A flowchart of the feature fusion step S104 of a solid-state battery safety early warning method provided in an embodiment of the present invention;

[0074] Figure 6 A flowchart of the real-time warning step S105 of a solid-state battery safety warning method provided in an embodiment of the present invention;

[0075] Figure 7 This is a schematic diagram of a solid-state battery safety early warning system provided in an embodiment of the present invention;

[0076] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0077] icon:

[0078] 710 - Image acquisition unit; 720 - Preprocessing unit; 730 - Parameter fusion unit; 740 - Feature fusion unit; 750 - Real-time early warning unit;

[0079] 101 - Processor; 102 - Memory; 103 - Bus; 104 - Communication interface. Detailed Implementation

[0080] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] To facilitate understanding of this embodiment, a detailed description of a solid-state battery safety early warning method disclosed in this embodiment of the invention will be provided first, such as... Figure 1 As shown, the method includes:

[0082] Image acquisition step S101: Acquire the structural image, morphological image and labeled fluorescence image corresponding to the negative electrode dendrite of the target solid-state battery, and determine the multimodal image corresponding to the target solid-state battery based on the structural image, morphological image and labeled fluorescence image.

[0083] This step involves acquiring three key images of dendrites on the negative electrode of the target solid-state battery: structural images (reflecting the macroscopic spatial distribution of dendrites), morphological images (presenting the microscopic morphological details of dendrites), and labeled fluorescence images (achieving dendrite-specific identification). Then, through image registration technology (ensuring that the spatial positions of the three types of images correspond), they are integrated to form a multimodal image that can comprehensively characterize the dendrite state.

[0084] Preprocessing step S102: Gaussian filtering for noise reduction, edge feature extraction, and 3D reconstruction are performed on the multimodal images to obtain the preprocessed images corresponding to the multimodal images.

[0085] This step is a preprocessing step. For the multimodal image obtained by S101, three key processes are performed respectively: Gaussian filtering for noise reduction (eliminating electronic noise, background fluorescence and other interference during the acquisition process), edge feature extraction (highlighting the boundary contours between dendrites and negative electrode / electrolyte), and three-dimensional reconstruction (converting the two-dimensional image into a three-dimensional dendrite model). Finally, a preprocessed image with low noise, clear features and spatial information is obtained.

[0086] Parameter fusion step S103: Calculate the image entropy, standard deviation and structural similarity index of the preprocessed image based on the gray value of the preprocessed image, and construct the fusion parameters corresponding to the target solid-state battery based on the image entropy, standard deviation and structural similarity index.

[0087] This step calculates three types of quantification metrics based on the pixel grayscale values ​​of the preprocessed image: image entropy (reflecting the richness of image information, with higher entropy values ​​in dendrite regions), standard deviation (reflecting grayscale dispersion, with higher standard deviations for greater differences between dendrites and the background), and structural similarity index (SSIM) (verifying the structural consistency of images of different modalities and eliminating interference from impurities). Then, the three types of metrics are integrated through a dynamic weighting algorithm to construct a fusion parameter that quantifies the "existence, reliability, and complexity" of dendrites.

[0088] Feature fusion step S104: Obtain the structural features and detail features corresponding to the preprocessed image, perform regional energy weighted fusion calculation on the structural features using fusion parameters, and perform dynamic weighted fusion calculation on the detail features using fusion parameters to obtain the fused image corresponding to the target solid-state battery.

[0089] This step first extracts two types of core features from the preprocessed image: structural features (macroscopic dendrite distribution, trunk growth direction, etc.) and detail features (dendrite tip curvature, branch density, etc.). Then, based on the fusion parameters of S103, regional energy-weighted fusion is performed on the structural features (highlighting high-risk areas, such as dense dendrite areas), and dynamic weighted fusion is performed on the detail features (adjusting weights according to feature reliability to avoid errors). Finally, the two types of optimized features are integrated to obtain a fused image that takes into account both macroscopic structure and microscopic details.

[0090] Real-time warning step S105: Obtain the dendrite region contained in the fused image, calculate the dendrite length and growth rate in the dendrite region in real time, and provide a safety warning for the target solid-state battery based on the dendrite length and growth rate.

[0091] This step first uses a relevant segmentation algorithm (to automatically identify and locate dendrite regions in the fused image), and then calculates two core risk parameters in real time: dendrite length (the straight-line distance from the negative electrode to the farthest dendrite to determine whether it is close to the electrolyte thickness) and dendrite growth rate (calculated by the length difference at different time points to determine whether the growth is accelerated); then the two parameters are compared with preset safety thresholds, and corresponding safety warnings are triggered for the target solid-state battery according to the risk level (low / medium / high).

[0092] Optionally, the image acquisition step S101, such as Figure 2 As shown, it includes:

[0093] Step S201: Acquire structural images using an optical coherence tomography (OCT) device corresponding to the target solid-state battery.

[0094] First, an optical coherence tomography (OCT) scanner adapted to the target solid-state battery is used to scan the negative electrode region inside the battery, generating a structural image that reflects the macroscopic spatial state of the dendrites. The OCT scanner offers advantages such as non-invasiveness, high resolution (micrometer-level), and rapid imaging. It can penetrate the electrolyte and casing without disassembling the battery, clearly showing the growth path of the dendrites from the negative electrode (e.g., whether it extends along electrolyte gaps or moves towards the positive electrode), providing a basis for subsequent judgment of the macroscopic growth trend of the dendrites. For example, it can directly observe whether the dendrites have approached the middle layer of the electrolyte, providing a preliminary assessment of the short-circuit risk.

[0095] Step S202: Acquire morphological images using a scanning electron microscope corresponding to the target solid-state battery.

[0096] Using a scanning electron microscope (SEM) corresponding to the target solid-state battery, the dendrites on the surface and shallow region of the negative electrode are scanned at a microscopic level to capture morphological images that reveal the microscopic morphology of the dendrites. The core advantages of SEM are its high magnification (up to hundreds of thousands of times) and nanometer-level resolution, which can clearly present the microscopic details of the dendrites: such as the diameter of the dendrite trunk, the number and distribution of branches, and the sharpness of the tips (the sharper the tip, the easier it is to break through the interface barrier of the solid electrolyte). These microscopic morphological features are directly related to the growth activity of the dendrites and are the key basis for subsequent judgment on whether the dendrites are accelerating their growth.

[0097] Step S203: Acquire labeled fluorescence images using a fluorescence microscope corresponding to the target solid-state battery.

[0098] This step first involves injecting a fluorescent probe (such as a fluorescent molecule containing a lithium-affinity functional group) that specifically binds to lithium dendrites into the target solid-state battery. Then, a fluorescence microscope is used to observe the negative electrode region, acquiring labeled fluorescent images showing fluorescence signals emitted only in the dendrite areas. The fluorescence microscope, through specific fluorescence recognition, effectively eliminates interference from impurities within the battery (such as electrolyte decomposition products and electrode powder). Only lithium dendrites bound to the fluorescent probe emit light; other areas show no fluorescence signal. This allows for precise localization of the true dendrite region, avoiding subsequent misclassification of impurities as dendrites and resulting in warning errors.

[0099] Step S204: After synchronously sampling the data channels corresponding to the structural image, morphological image and labeled fluorescence image, a multimodal image corresponding to the target solid-state battery is obtained.

[0100] The structural image acquired by S201, the morphological image acquired by S202, and the labeled fluorescence image acquired by S203 are sampled temporally and spatially from their respective device data channels (e.g., the optical signal channel of OCT, the electronic signal channel of SEM, and the fluorescence signal channel of fluorescence microscopy). This ensures that the pixel positions and acquisition times of the three types of images correspond one-to-one, and the images are ultimately integrated to obtain a multimodal image. In practical scenarios, the multi-channel synchronous sampling process can be time-aligned using a 16-bit AD conversion chip, with a sampling interval ≤10ms.

[0101] Optionally, preprocessing step S102, such as Figure 3 As shown, it includes:

[0102] Step S301: After performing Gaussian filtering to denoise the structure image, the filtered image corresponding to the structure image is obtained.

[0103] First, Gaussian filtering is applied to the structural image for noise reduction, based on the original OCT image. (M×N is the size of the image) Two-dimensional Gaussian filtering is used to suppress noise.

[0104] The filtered image is calculated using the following formula:

[0105] ;

[0106] in, For filtered images; For Gaussian kernel function, And satisfy ; This is a structural image.

[0107] Step S302: Use the Canny operator to calculate the edge features of the topography image to obtain the dendrite edge data of the target solid-state battery, and use the gradient value of the dendrite edge data to extract the dendrite edge region contained in the topography image.

[0108] Edge features are extracted from the topographic images acquired by SEM, and the Canny operator is used to process the topographic images corresponding to the SEM. Extract the dendrite edges, and then calculate the gradient using the Sobel operator:

[0109] , ;

[0110] The asterisk (*) indicates convolution calculation and gradient magnitude. and direction The calculation formula is:

[0111] , ;

[0112] The edge preservation condition is: or Furthermore, the domain contains high threshold edges. For the preset high threshold, This is a preset low threshold.

[0113] Step S303: Obtain the Z-axis scanning sequence corresponding to the labeled fluorescence image, and obtain the three-dimensional structure image corresponding to the labeled fluorescence image based on the maximum intensity projection result of the Z-axis scanning sequence.

[0114] This step involves three-dimensional reconstruction of the fluorescence image, based on the Z-axis scanning sequence of the fluorescence microscope. Three-dimensional structural images are generated using maximum intensity projection (MIP): , where Z is the number of scan layers along the Z-axis.

[0115] Step S304: Determine the preprocessed image corresponding to the multimodal image based on the filtered image, dendrite edge region, and three-dimensional structure image.

[0116] Optionally, the parameter fusion step S103 can determine the fusion parameters based on the Artificial Bee Colony (ABC) algorithm, such as... Figure 4 As shown, it includes:

[0117] Step S401: Determine the first fusion weight, second fusion weight, and third fusion weight corresponding to the structural image, morphological image, and labeled fluorescence image, respectively.

[0118] This step corresponds to the honey source encoding and initialization process. The first fusion weight, second fusion weight, and third fusion weight corresponding to the structural image, morphological image, and labeled fluorescence image are respectively... , , .

[0119] Step S402: Construct a three-dimensional honey source vector using the first fusion weight, the second fusion weight, and the third fusion weight.

[0120] Specifically, the high-frequency fusion weights of OCT, SEM, and fluorescence modalities are encoded into a three-dimensional honey source vector. ,satisfy and (m is the modality index). The initial population size is 50, and the formula for randomly generating nectar sources is:

[0121] ;

[0122] rand(0,1) is a uniformly random number in the interval (0,1), and the superscript (0) indicates the initial iteration.

[0123] Step S403: Determine the search step size based on the maximum number of iterations of the hired bees corresponding to the artificial bee colony model, use the step size to determine the new nectar source corresponding to the three-dimensional nectar source vector, and update the nectar source corresponding to the artificial bee colony model according to the new nectar source.

[0124] This step is the territory search process for hired bees, during which they generate new nectar sources. The calculation formula is:

[0125] ;

[0126] in The weight dimension is k, where k≠i represents other randomly selected honey sources, and the search step size is k. Dynamic decay with iteration number t: ;in This represents the maximum number of iterations.

[0127] Step S404: Obtain the grayscale value of the preprocessed image, and use the grayscale value to calculate the image entropy, standard deviation and structural similarity index of the preprocessed image.

[0128] Specifically, Image entropy; Standard deviation; It is a structural similarity index;

[0129] ; grayscale value The probability distribution;

[0130] ; This is the fused image corresponding to the preprocessed image obtained using grayscale values; The mean of the merged images;

[0131] ;in and These represent the mean values ​​of the preprocessed image and the reference image, respectively, and the mean value of the fused image. For covariance, , , It is a grayscale level.

[0132] Step S405: Determine the fitness function corresponding to the preprocessed image based on the image entropy, standard deviation, and structural similarity index.

[0133] The fitness function is calculated using the following formula:

[0134] ;

[0135] in, For the adaptive function; A three-dimensional honey source vector; , , These are the weight values ​​corresponding to image entropy, standard deviation, and structural similarity index, respectively.

[0136] Step S406: Based on the roulette wheel beating strategy of the observation bees and the nectar source update strategy of the scout bees corresponding to the artificial bee colony model, determine the selection probability of nectar sources under the fitness function.

[0137] This step involves the selection of observation bees and the updating of scout bees. Observation bees select nectar sources according to a roulette wheel strategy, with selection probabilities... For honey sources that have not been optimized in 20 consecutive iterations, a global random reset is performed: ,in Use standard normally distributed random numbers to avoid the algorithm getting trapped in local optima.

[0138] Step S407: Determine the optimized fusion parameters corresponding to the fitness function by selecting the probability, and use the optimized fusion parameters to construct the fusion parameters corresponding to the target solid-state battery.

[0139] The above process is based on the optimization of fusion parameters of the artificial bee colony (ABC) algorithm. The fusion weights of the multi-scale transform domain are encoded as nectar sources. The fusion parameters are optimized by employing bee domain search, observation bee roulette wheel selection and scout bee global update strategy, with image entropy, standard deviation and structural similarity index as fitness functions.

[0140] Optionally, feature fusion step S104, such as Figure 5 As shown, it includes:

[0141] Step S501: Perform non-subsampled shear wave transform decomposition on the preprocessed image to obtain the low-frequency components corresponding to the preprocessed image. and high frequency components .

[0142] The preprocessed image is decomposed using non-subsampled shear wave transform (NSCT). Four-level NSCT decomposition is then performed on each modal image after preprocessing to obtain the low-frequency components. and high frequency components (l=1,…4 are the decomposition levels), the decomposition process satisfies: ,in This is the inverse NSCT transform operator.

[0143] Step S502: Use low-frequency components to determine the structural features corresponding to the preprocessed image, and use high-frequency components to determine the detailed features corresponding to the preprocessed image.

[0144] Step S503: Perform region energy weighted fusion calculation on the structural features using fusion parameters to obtain the low-frequency fusion coefficients corresponding to the preprocessed image. ;

[0145] This step is the low-frequency component fusion process, which uses a regional energy weighting rule to fuse low-frequency components. The low-frequency fusion coefficient at a certain location (x,y) is... for:

[0146] ;

[0147] Among them, regional energy weight Regional energy .

[0148] Step S504: Perform region energy weighted fusion calculation on the structural features using fusion parameters to obtain the high-frequency fusion coefficients corresponding to the preprocessed image. ;

[0149] This step is the high-frequency component fusion process, which uses ABC-optimized weights to fuse high-frequency components, and the high-frequency fusion coefficients... ,in The optimal weights are obtained by optimizing the ABC algorithm.

[0150] Step S505: Obtain the inverse transform operator used in non-subsampled shear wave transform decomposition. After performing inverse transform processing on the low-frequency fusion coefficient and the high-frequency fusion coefficient using the inverse transform operator, obtain the fusion image corresponding to the target solid-state battery.

[0151] fused images The result is obtained through the following formula; ; It is the inverse transform operator.

[0152] Optional, real-time alert step S105, such as Figure 6 As shown, it includes:

[0153] Step S601: Determine the dendrite regions contained in the fused image based on the structural image, morphological image and labeled fluorescence image corresponding to the negative electrode dendrites;

[0154] Step S602: Calculate the dendrite length and growth rate in the dendrite region in real time based on the endpoints of the dendrite skeleton in the dendrite region;

[0155] Step S603: Obtain the preset warning threshold conditions, determine the warning level corresponding to the target solid-state battery based on the correspondence between dendrite length and growth rate and the warning threshold conditions, and use the warning level to provide a safety warning for the target solid-state battery.

[0156] Specifically, the above process includes a dendrite feature extraction step and an early warning pre-judgment step. During the dendrite feature extraction process, the dendrite length... ; and The starting and ending points of the dendritic framework; growth rate ; Let be the length of the t-th monitoring session. For monitoring intervals.

[0157] During the early warning threshold determination process, corresponding early warning threshold conditions can be preset. Specifically: Level 1 early warning is... and Cycle; Level II warning is And the tip curvature (curvature) A Level 3 warning is triggered when dendrites are detected penetrating the electrolyte interface (determined by OCT depth information).

[0158] As can be seen from the above solid-state battery safety early warning method, this method simultaneously acquires multimodal images of negative electrode dendrites using optical coherence tomography, scanning electron microscopy, and fluorescence microscopy. It dynamically optimizes the multi-scale transform domain fusion parameters using an improved artificial bee colony algorithm to achieve cross-modal feature complementary fusion. Combined with a real-time monitoring module, it completes dynamic tracking and early warning of dendrite growth, thereby solving the problems of low detection accuracy and static fusion parameters in traditional single-mode methods. This improves the accuracy and real-time performance of dendrite growth monitoring and enhances the safety early warning effect of solid-state batteries.

[0159] Corresponding to the above embodiments of the solid-state battery safety warning method, this invention also provides a solid-state battery safety warning system, such as... Figure 7 As shown, the system includes:

[0160] Image acquisition unit 710: used to acquire structural images, morphological images and labeled fluorescence images corresponding to the negative electrode dendrites of the target solid-state battery, and to determine the multimodal image corresponding to the target solid-state battery based on the structural images, morphological images and labeled fluorescence images;

[0161] Preprocessing unit 720: is used to perform Gaussian filtering for noise reduction, edge feature extraction and 3D reconstruction on the multimodal image to obtain the preprocessed image corresponding to the multimodal image;

[0162] Parameter fusion unit 730: used to calculate the image entropy, standard deviation and structural similarity index of the preprocessed image based on the gray value of the preprocessed image, and to construct the fusion parameters corresponding to the target solid-state battery based on the image entropy, standard deviation and structural similarity index;

[0163] Feature fusion unit 740: used to acquire structural features and detail features corresponding to the preprocessed image, perform regional energy weighted fusion calculation on the structural features using fusion parameters, and perform dynamic weighted fusion calculation on the detail features using fusion parameters to obtain the fused image corresponding to the target solid-state battery;

[0164] Real-time warning unit 750: used to acquire dendritic regions contained in the fused image, calculate dendrite length and growth rate in the dendrite region in real time, and provide safety warnings for the target solid-state battery based on dendrite length and growth rate.

[0165] As can be seen from the above solid-state battery safety early warning system, the system obtains multimodal images of negative electrode dendrites by acquiring structural images, morphological images and labeled fluorescence images, and then performs parameter fusion and feature fusion to obtain the fused image corresponding to the target solid-state battery. This realizes the cross-modal feature complementary fusion process, solves the problems of low detection accuracy and static fusion parameters in traditional single-mode systems, improves the accuracy and real-time performance of dendrite growth monitoring, and enhances the safety early warning effect of solid-state batteries.

[0166] The solid-state battery safety warning system provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned solid-state battery safety warning method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned solid-state battery safety warning method embodiment.

[0167] This embodiment also provides an electronic device, the structural schematic diagram of which is shown below. Figure 8 As shown, the device includes a processor 101 and a memory 102; wherein, the memory 102 is used to store one or more computer instructions, which are executed by the processor to implement the steps of the above-described solid-state battery safety warning method.

[0168] Figure 8 The electronic device shown also includes a bus 103 and a communication interface 104, with the processor 101, communication interface 104 and memory 102 connected via the bus 103.

[0169] The memory 102 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. The bus 103 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0170] The communication interface 104 is used to connect to at least one user terminal and other network units through a network interface, and to send encapsulated IPv4 packets or IPv4 packets to the user terminal through the network interface.

[0171] Processor 101 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 101 or by instructions in software form. The processor 101 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 102. The processor 101 reads the information in memory 102 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0172] This invention also provides a storage medium storing a computer program, which, when run by a processor, executes the steps of the solid-state battery safety warning method described in the foregoing embodiments.

[0173] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, devices, and methods can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0174] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0175] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0176] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0177] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A solid-state battery safety early warning method, characterized by, The method comprises: An image acquisition step: acquiring a structure image, a morphology image and a labeled fluorescence image corresponding to a negative electrode dendrite of a target solid-state battery, and determining a multi-modal image corresponding to the target solid-state battery according to the structure image, the morphology image and the labeled fluorescence image; A preprocessing step: after respectively performing Gaussian filter denoising, edge feature extraction and three-dimensional reconstruction processing on the multi-modal image, a preprocessed image corresponding to the multi-modal image is obtained; A parameter fusion step: calculating image entropy, standard deviation and structural similarity index of the preprocessed image based on the gray value of the preprocessed image, and constructing a fusion parameter corresponding to the target solid-state battery based on the image entropy, the standard deviation and the structural similarity index; A feature fusion step: obtaining structure features and detail features corresponding to the preprocessed image, performing regional energy weighted fusion calculation on the structure features by using the fusion parameter, and performing dynamic weight fusion calculation on the detail features by using the fusion parameter, to obtain a fusion image corresponding to the target solid-state battery; A real-time early warning step: acquiring a dendrite region contained in the fusion image, calculating dendrite length and growth rate in the dendrite region in real time, and performing safety warning on the target solid-state battery based on the dendrite length and the growth rate; The parameter fusion step comprises: respectively determining first fusion weight, second fusion weight and third fusion weight corresponding to the structure image, the morphology image and the labeled fluorescence image; constructing a three-dimensional honey source vector by using the first fusion weight, the second fusion weight and the third fusion weight; determining a search step by using the maximum number of iterations of the artificial bee colony model corresponding to the hired bee, determining a new honey source corresponding to the three-dimensional honey source vector by using the step, and updating the honey source of the artificial bee colony model according to the new honey source; calculating image entropy, standard deviation and structural similarity index of the preprocessed image by using the gray value of the preprocessed image; determining an adaptive function corresponding to the preprocessed image according to the image entropy, the standard deviation and the structural similarity index; determining the selection probability of the honey source under the adaptive function based on the roulette strategy of the observer bee and the honey source update strategy of the scout bee of the artificial bee colony model; determining an optimized fusion parameter corresponding to the adaptive function by using the selection probability, and constructing the fusion parameter corresponding to the target solid-state battery by using the optimized fusion parameter; The feature fusion step comprises: performing non-subsampled shearlet transform decomposition on the preprocessed image to obtain a low-frequency component and a high-frequency component corresponding to the preprocessed image and high-frequency components ​ determining the structure features corresponding to the preprocessed image by using the low-frequency component, and determining the detail features corresponding to the preprocessed image by using the high-frequency component; The structure features are subjected to zone energy weighted fusion calculation through the fusion parameters, to obtain low-frequency fusion coefficients corresponding to the preprocessed image ; wherein, ; is a zone energy weight, ; is a zone energy, ; is a modal index; The detail features are calculated by dynamic weight fusion according to the fusion parameters to obtain high-frequency fusion coefficients corresponding to the preprocessed image ; wherein ; is an optimal weight value corresponding to the fusion parameter An inverse transform operator used in a non-subsampled shearlet transform decomposition is obtained, and the low-frequency fusion coefficients and the high-frequency fusion coefficients are inversely transformed using the inverse transform operator to obtain a fusion image corresponding to the target solid-state battery; wherein the fusion image is calculated by the following formula; is the inverse transform operator.​ 2. The solid-state battery safety warning method of claim 1, wherein, The image acquisition step comprises: acquiring the structure image by using an optical coherence tomography device corresponding to the target solid-state battery; acquiring the morphology image by using a scanning electron microscope corresponding to the target solid-state battery; acquiring the labeled fluorescence image by using a fluorescence microscope corresponding to the target solid-state battery; Synchronously sampling data channels corresponding to the structural image, the topographic image, and the labeled fluorescence image to obtain the multi-modal image corresponding to the target solid-state battery.

3. The solid-state battery safety warning method of claim 1, wherein, The preprocessing step comprises: After Gaussian filtering and denoising calculation of the structural image, a filtered image corresponding to the structural image is obtained; Edge feature calculation of the topographic image is performed using a Canny operator to obtain dendrite edge data of the target solid-state battery, and a gradient value of the dendrite edge data is used to extract a dendrite edge region contained in the topographic image; A Z-axis scanning sequence corresponding to the labeled fluorescence image is obtained, and a three-dimensional structural image corresponding to the labeled fluorescence image is obtained based on maximum intensity projection results of the Z-axis scanning sequence; The preprocessed image corresponding to the multi-modal image is determined based on the filtered image, the dendrite edge region, and the three-dimensional structural image.

4. The solid-state battery safety warning method of claim 3, wherein, The filtered image is calculated by the following formula: ； wherein is the filtered image; is a Gaussian kernel function, , and satisfies ; is the structure image; The dendrite edge region is calculated by the following formula: ； ； wherein, is a pre-set high threshold value, is a pre-set low threshold value; is the gradient value of the dendrite edge data; ; , ; is the topographic image; The three-dimensional structural image is calculated by the following formula: ； wherein is the three-dimensional structure image; is the Z-axis scan sequence, .

5. The solid-state battery safety warning method of claim 1, wherein, The adaptive function is calculated by the following formula: ； wherein, is the adaptation function; is the three-dimensional nectar source vector; is the image entropy; is the standard deviation; is the structural similarity index; , , are weight values corresponding to the image entropy, the standard deviation and the structural similarity index, respectively. ; is a probability distribution of gray values ; ; a fusion image corresponding to the pre-processed image calculated using the gray values; is a mean value of the fusion image; ； wherein and are the mean of the corresponding reference image and the mean of the fused image respectively for the pre-processed image, is the covariance, , , is the gray level.

6. The solid-state battery safety warning method of claim 1, wherein, The real-time early warning step comprises: A dendrite region contained in the fusion image is determined based on the structural image, the topographic image, and the labeled fluorescence image corresponding to the negative dendrite; calculating the dendrite length and growth rate in the dendrite region in real time according to the end points of the dendrite skeleton in the dendrite region; wherein the dendrite length ; and is the starting end point and the ending end point of the dendrite skeleton; the growth rate ; is the length of the tth monitoring, is the monitoring interval; A preset early warning threshold condition is obtained, a warning level corresponding to the target solid-state battery is determined based on a corresponding relationship between the dendrite length and the growth rate and the early warning threshold condition, and the target solid-state battery is subjected to safety warning using the warning level.

7. A solid-state battery safety warning system, characterized by, The system comprises: An image acquisition unit for acquiring a structural image, a topographic image, and a labeled fluorescence image corresponding to a negative dendrite of a target solid-state battery, and determining a multi-modal image corresponding to the target solid-state battery based on the structural image, the topographic image, and the labeled fluorescence image; A preprocessing unit for obtaining a preprocessed image corresponding to the multi-modal image after Gaussian filtering and denoising, edge feature extraction, and three-dimensional reconstruction processing of the multi-modal image, respectively; A parameter fusion unit for calculating an image entropy, a standard deviation, and a structural similarity index of the preprocessed image based on a gray value of the preprocessed image, and constructing a fusion parameter corresponding to the target solid-state battery based on the image entropy, the standard deviation, and the structural similarity index; A feature fusion unit for obtaining a structural feature and a detail feature corresponding to the preprocessed image, performing region energy weighted fusion calculation on the structural feature using the fusion parameter, and performing dynamic weight fusion calculation on the detail feature using the fusion parameter to obtain a fusion image corresponding to the target solid-state battery; A real-time early warning unit for obtaining a dendrite region contained in the fusion image, calculating a dendrite length and a growth rate in the dendrite region in real time, and performing safety warning on the target solid-state battery based on the dendrite length and the growth rate. The parameter fusion unit is further configured to determine first, second and third fusion weights corresponding to the structure image, the topography image and the labeled fluorescence image respectively, construct a three-dimensional honey source vector using the first, second and third fusion weights, determine a search step length based on a maximum iteration number of an artificial bee colony model corresponding to a scout bee, determine a new honey source corresponding to the three-dimensional honey source vector using the step length, update a honey source corresponding to the artificial bee colony model according to the new honey source, obtain a gray value of the preprocessed image, calculate an image entropy, a standard deviation and a structural similarity index of the preprocessed image using the gray value, determine an adaptive function corresponding to the preprocessed image according to the image entropy, the standard deviation and the structural similarity index, determine a selection probability of the honey source under the adaptive function based on a roulette strategy of an observer bee and a honey source update strategy of a scout bee of the artificial bee colony model, determine an optimized fusion parameter corresponding to the adaptive function through the selection probability, and construct the fusion parameter corresponding to the target solid-state battery using the optimized fusion parameter. The feature fusion unit is further configured to: perform non-subsampled shear wave transform decomposition on the preprocessed image to obtain the low-frequency component corresponding to the preprocessed image. and high frequency components The low-frequency components are used to determine the structural features corresponding to the preprocessed image, and the high-frequency components are used to determine the detailed features corresponding to the preprocessed image. The structural features are then subjected to region energy weighted fusion calculation using the fusion parameters to obtain the low-frequency fusion coefficients corresponding to the preprocessed image. ;in, ; For regional energy weighting, ; For regional energy, ; The modality index is used; dynamic weight fusion calculation is performed on the detail features using the fusion parameters to obtain the high-frequency fusion coefficients corresponding to the preprocessed image. ;in, ; The optimal weight values ​​corresponding to the fusion parameters are determined; the inverse transform operator used in non-subsampled shear wave transform decomposition is obtained, and the low-frequency fusion coefficients and the high-frequency fusion coefficients are inversely transformed using the inverse transform operator to obtain the fused image corresponding to the target solid-state battery; wherein, the fused image The result is obtained through the following formula; ; Let be the inverse transform operator.

8. An electronic device, comprising: The solid-state battery safety warning method comprises a processor and a memory, the memory stores computer executable instructions capable of being executed by the processor, and the processor executes the computer executable instructions to implement the steps of the solid-state battery safety warning method in any one of claims 1 to 6.

Images