Performance analysis method and system for solid-state lithium-ion batteries

Abstract

The present invention provides a performance analysis method and system for solid-state lithium-ion batteries, relating to the field of artificial intelligence technology. First, the target manufacturing process data and target manufacturing process images of the battery electrode are acquired. Second, semantic mining is performed on the target manufacturing process data and target manufacturing process images, respectively, to form a manufacturing process latent vector and a manufacturing process latent vector. Then, the manufacturing process latent vector and the manufacturing process latent vector are aggregated at multiple levels to form a process semantic vector. Finally, analysis is performed based on the process semantic vector to obtain target performance analysis data. Based on the above, the relatively low reliability and efficiency of performance analysis in existing technologies can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a performance analysis method and system for solid-state lithium-ion batteries. Background Art

[0002] The energy density of traditional lithium-ion batteries is difficult to break through further, and its safety is low, and there are risks such as flammability and explosion. Solid-state electrolytes can greatly improve the safety of lithium-ion batteries by replacing traditional electrolytes. However, the problem of poor ion conduction in thick electrodes and the high interface contact impedance between electrodes and solid-state electrolytes can also lead to low performance of all-solid-state lithium-ion batteries, such as relatively low energy density. Based on this, in the prior art, a method for preparing electrodes specifically for all-solid-state lithium-ion batteries is provided. The method adopts a wet process to add materials such as succinonitrile, lithium salts, polymer electrolytes and lithium iron phosphate or lithium nickel cobalt manganese oxide for compounding, and coating on a current collector to prepare an electrode sheet. The method can effectively solve the problem of poor ion conduction in thick electrodes and the high interface contact impedance between electrodes and solid-state electrolytes. The electrode can be applied to all-solid-state lithium-ion batteries to obtain a higher energy density. However, the actual performance of all-solid-state lithium-ion batteries requires actual measurement to determine, which leads to relatively low efficiency of performance analysis. Moreover, during the actual measurement process, it is easily affected by the measurement personnel, making the reliability of performance analysis relatively low. Summary of the Invention

[0003] In view of this, an object of the present invention is to provide a performance analysis method and system for solid-state lithium-ion batteries, so as to improve the problems of relatively low reliability and efficiency of performance analysis in the prior art.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A performance analysis method for a solid-state lithium-ion battery, comprising:

[0006] Acquiring target manufacturing process data and target manufacturing process images of a battery electrode of a target battery, wherein the target battery is an all-solid-state lithium-ion battery;

[0007] Semantic mining is performed on the target manufacturing process data and the target manufacturing process image respectively to form corresponding manufacturing process latent vectors and manufacturing process latent vectors, wherein the size of the manufacturing process latent vector is larger than the size of the manufacturing process latent vector;

[0008] Performing multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector, wherein the process semantic vector is used to represent the global semantic information of the manufacturing process latent vector and the manufacturing process latent vector;

[0009] An analysis is performed based on the process semantic vector to obtain target performance analysis data, wherein the target performance analysis data is used to characterize the performance of the target battery related to energy density.

[0010] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing semantic mining on the target manufacturing process data and the target manufacturing process image to form corresponding manufacturing process latent vectors and manufacturing process latent vectors includes:

[0011] Performing an embedding operation on the target manufacturing process data to form a manufacturing process embedding vector, and obtaining a manufacturing process latent vector based on the manufacturing process embedding vector;

[0012] A first convolution operation is performed on each image frame included in the target production process image to form an image convolution vector corresponding to each image frame, and a semantic extraction operation and a second convolution operation are performed on the image convolution vector corresponding to each image frame to form an image convolution depth vector corresponding to each image frame, and the image convolution depth vectors corresponding to each image frame are spliced ​​to form a production process latent vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0013] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of embedding the target manufacturing process data to form a manufacturing process embedding vector, and obtaining a manufacturing process latent vector based on the manufacturing process embedding vector includes:

[0014] Performing an embedding operation on the target manufacturing process data to form a manufacturing process embedding vector;

[0015] Decomposing the manufacturing process embedding vector to form a first process decomposition vector and a second process decomposition vector, and, based on a plurality of predetermined depth information, performing depth mining on the first process decomposition vector and the second process decomposition vector to form a plurality of corresponding first process depth vectors and a plurality of second process depth vectors;

[0016] Merging the plurality of first process depth vectors and the second process depth vector according to corresponding depth information to form a plurality of corresponding merged process depth vectors;

[0017] Obtaining a manufacturing process characterization vector formed by training based on sample manufacturing data and corresponding performance labels, and optimizing the manufacturing process characterization vectors in sequence based on the multiple merged process depth vectors to form an updated manufacturing process characterization vector;

[0018] Based on the updated manufacturing process representation vector, a manufacturing process latent vector is obtained.

[0019] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector includes:

[0020] Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector to form a first production aggregated vector;

[0021] Performing semantic extraction on the production process latent vector and the production technique latent vector and then aggregating them to form a second production aggregated vector;

[0022] Performing semantic adaptation on the first production aggregation vector and the second production aggregation vector and then bidirectionally aggregating them to form a third production aggregation vector and a fourth production aggregation vector;

[0023] Performing semantic expansion on the third production aggregation vector and the fourth production aggregation vector and then bidirectionally aggregating them to form a fifth production aggregation vector and a sixth production aggregation vector;

[0024] Aggregation is performed based on the fifth production aggregation vector and the sixth production aggregation vector to form a process semantic vector.

[0025] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing a semantic extraction operation on the manufacturing process latent vector and aggregating it with the manufacturing process latent vector to form a first manufacturing aggregated vector includes:

[0026] performing a semantic extraction operation on the production process latent vector and aggregating the resulting vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector, wherein the semantic extraction operation includes convolution and / or pooling to extract semantic information;

[0027] Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector;

[0028] The first aggregated convolution vector is subjected to a semantic extraction operation and then aggregated with the second aggregated convolution vector, and a first convolution operation is performed on the aggregated vector to form a first aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0029] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing semantic extraction on the manufacturing process latent vector and the manufacturing technique latent vector and then aggregating them to form a second manufacturing aggregated vector includes:

[0030] Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector;

[0031] Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector;

[0032] The first aggregated convolution vector and the second aggregated convolution vector are subjected to a semantic extraction operation and then aggregated, and a second convolution operation is performed on the aggregated vector to form a second aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0033] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing a semantic adaptation operation on the first production aggregation vector and the second production aggregation vector and then bidirectionally aggregating them to form a third production aggregation vector and a fourth production aggregation vector includes:

[0034] performing a semantic expansion operation on the second produced aggregate vector and then aggregating the vector with the first produced aggregate vector, and performing a first convolution operation on the aggregated vector to form a third produced aggregate vector;

[0035] The first aggregated vector is subjected to a semantic extraction operation and then aggregated with the second aggregated vector, and a second convolution operation is performed on the aggregated vector to form a fourth aggregated vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information, the semantic extraction operation includes convolution and / or pooling to achieve extraction of semantic information, and the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0036] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing a semantic expansion operation on the third production aggregation vector and the fourth production aggregation vector and then performing bidirectional aggregation to form a fifth production aggregation vector and a sixth production aggregation vector includes:

[0037] performing a semantic expansion operation on the third aggregate vector and the fourth aggregate vector respectively and then aggregating them to form a fifth aggregate vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information;

[0038] The fourth production aggregation vector is subjected to a semantic expansion operation and then aggregated with the third production aggregation vector to form a sixth production aggregation vector.

[0039] In a preferred embodiment of the present invention, in the above-mentioned solid-state lithium-ion battery performance analysis method, the step of performing a semantic expansion operation on the fourth production aggregation vector and then aggregating it with the third production aggregation vector to form a sixth production aggregation vector includes:

[0040] Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector;

[0041] Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector;

[0042] Performing semantic expansion operations on the fourth aggregated vector and the third aggregated vector respectively and then aggregating them, adding the aggregated vector and the first aggregated convolution vector, and performing a first convolution operation on the added vector to form a first added convolution vector;

[0043] Performing a semantic expansion operation on the fourth produced aggregate vector and then aggregating the resulting aggregate vector with the third produced aggregate vector, adding the aggregated vector and the second aggregate convolution vector, and performing a second convolution operation on the added vector to form a second added convolution vector;

[0044] The second additive convolution vector is semantically expanded and then aggregated with the first additive convolution vector, and the aggregated vector and the manufacturing process latent vector are added together. Furthermore, a second convolution operation is performed on the added vector to form a sixth manufacturing aggregated vector.

[0045] On the basis of the above, the present invention also provides a performance analysis system for solid-state lithium-ion batteries, comprising:

[0046] memory for storing computer programs;

[0047] The processor connected to the memory is used to execute the computer program stored in the memory to implement the above-mentioned performance analysis method of the solid-state lithium-ion battery.

[0048] The performance analysis method and system of solid-state lithium-ion batteries provided by the present invention first obtain the target manufacturing process data and target manufacturing process images of the battery electrode; secondly, semantic mining is performed on the target manufacturing process data and target manufacturing process images respectively to form a manufacturing process latent vector and a manufacturing process latent vector; then, the manufacturing process latent vector and the manufacturing process latent vector are multi-level aggregated to form a process semantic vector; finally, analysis is performed based on the process semantic vector to obtain target performance analysis data. Based on the above content, on the one hand, semantic mining can be performed to obtain corresponding latent semantic information, so that performance analysis can be performed based on the latent semantic information. In this way, compared with the conventional actual testing scheme, it can have higher efficiency and is not easily interfered with by the tester, so the reliability can also be higher. In addition, since not only the potential semantic information in the process data is mined, but also the semantic information in the production process, after multi-level aggregation, a process semantic vector with richer semantic information can be obtained. Moreover, since different process data and actual production processes have an impact on the performance of the formed battery electrode, the semantic representation accuracy of the process semantic vector that combines the two aspects of semantic information can also be guaranteed, thereby ensuring the reliability of the target performance analysis data obtained, and then improving the problem of relatively low reliability and efficiency of performance analysis in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0050] Figure 1 This is a structural block diagram of a solid-state lithium-ion battery performance analysis system provided by an embodiment of the present invention.

[0051] Figure 2 A schematic flow chart of a method for analyzing the performance of a solid-state lithium-ion battery according to an embodiment of the present invention.

[0052] Figure 3 This is a capacity diagram at a rate of 0.2-2C provided by an embodiment of the present invention.

[0053] Figure 4 This is a cycle performance diagram at a 0.5C rate provided by an embodiment of the present invention.

[0054] Figure 5 A schematic diagram of forming a first aggregation vector provided by an embodiment of the present invention.

[0055] Figure 6 A schematic diagram of forming the fifth production aggregation vector and the sixth production aggregation vector provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0057] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0058] like Figure 1 As shown, an embodiment of the present invention provides a solid-state lithium-ion battery performance analysis system, wherein the solid-state lithium-ion battery performance analysis system may include a memory, a processor, and a solid-state lithium-ion battery performance analysis device.

[0059] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, the memory and the processor can be electrically connected through one or more communication buses or signal lines. The performance analysis device of the solid-state lithium-ion battery includes at least one software function module stored in the memory in the form of software or firmware. The processor is used to execute an executable computer program stored in the memory, for example, the software function module and computer program included in the performance analysis device of the solid-state lithium-ion battery, to implement the performance analysis method of the solid-state lithium-ion battery provided in the embodiment of the present invention.

[0060] Optionally, the memory may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. Furthermore, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on a chip (SoC), 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.

[0061] Optionally, the performance analysis device for a solid-state lithium-ion battery, as a software functional module, may include a first module, a second module, a third module, and a fourth module.

[0062] The first module is used to obtain target manufacturing process data and target manufacturing process images of the battery electrodes of the target battery, wherein the target battery is an all-solid-state lithium-ion battery. The second module is used to perform semantic mining on the target manufacturing process data and the target manufacturing process images, respectively, to form corresponding manufacturing process latent vectors and manufacturing process latent vectors, wherein the size of the manufacturing process latent vector is larger than the size of the manufacturing process latent vector. The third module is used to perform multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector, wherein the process semantic vector is used to represent the global semantic information possessed by the manufacturing process latent vector and the manufacturing process latent vector. The fourth module is used to perform analysis based on the process semantic vector to obtain target performance analysis data, wherein the target performance analysis data is used to represent the performance level of the target battery related to energy density.

[0063] It is understandable that Figure 1 The structure shown is for illustration only. The performance analysis system of the solid-state lithium-ion battery may also include a comparison Figure 1 More or fewer components than shown, or with Figure 1The different configurations shown, for example, may further include a communication unit for exchanging information with other devices, such as for acquiring target manufacturing process data and target manufacturing process images.

[0064] Combine Figure 2 The embodiment of the present invention further provides a solid-state lithium-ion battery performance analysis method applicable to the above-mentioned solid-state lithium-ion battery performance analysis system. The method steps defined in the process related to the solid-state lithium-ion battery performance analysis method can be implemented by the solid-state lithium-ion battery performance analysis system (hereinafter referred to as the performance analysis system). Figure 2 The specific process shown is explained in detail.

[0065] Step S110 , obtaining target manufacturing process data and target manufacturing process images of a battery electrode of a target battery.

[0066] In an embodiment of the present invention, the performance analysis system can obtain target manufacturing process data (which can be used to characterize manufacturing process parameters) and target manufacturing process images (which can be used to characterize the manufacturing process, for example, by using an image acquisition device to capture images of the manufacturing process, i.e., reflecting the actual situation during the manufacturing process, thereby providing a reliable basis for performance analysis) of the battery electrodes of the target battery. The target battery is an all-solid-state lithium-ion battery. For example, the target manufacturing process data can be:

[0067] Dissolve 10 g of PVDF-TrFE (polyvinylidene fluoride-trifluoroethylene) in 100 mL of N-methylpyrrolidone and stir evenly at 50 °C. Then add 0.5 g of SN (succinonitrile), 4 g of LiTFSI (lithium bis(trifluoromethanesulfonyl imide)), and 1 g of LLZTO (lithium lanthanum zirconium oxide) in sequence. Ensure that the solution is stirred evenly before each addition to obtain a mixed solution 1.

[0068] 1 g of conductive agent SP (Super P) was added to Solution 1 and stirred for 4 hours. Then, 38.5 g of lithium iron phosphate material was added and stirred for 3 hours to obtain a uniformly mixed electrode slurry with a viscosity of 6000 mPa·s.

[0069] The electrode slurry is coated on the aluminum foil with a coating thickness of 250um. After the electrode is dried, it is rolled at 70℃ to a thickness of 90%.

[0070] Step S120 , performing semantic mining on the target manufacturing process data and the target manufacturing process image respectively to form corresponding manufacturing process latent vectors and manufacturing process latent vectors.

[0071] In an embodiment of the present invention, after acquiring the target manufacturing process data and the target manufacturing process image, the performance analysis system may perform semantic mining on the target manufacturing process data and the target manufacturing process image, respectively, to form corresponding manufacturing process latent vectors and manufacturing process latent vectors. This mining extracts the underlying semantic information and represents it as a vector. The size of the manufacturing process latent vector may be larger than the size of the manufacturing process latent vector (the target manufacturing process image represents detailed information about the manufacturing process and may include multiple image frames, and therefore can be represented by a larger latent vector).

[0072] Step S130 , performing multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector.

[0073] In an embodiment of the present invention, after mining the manufacturing process latent vector and the manufacturing process latent vector, the performance analysis system can perform multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector. The process semantic vector is used to characterize the global semantic information possessed by the manufacturing process latent vector and the manufacturing process latent vector. It should be noted that although the manufacturing process latent vector and the manufacturing process latent vector can both represent the performance-related semantic information of the formed battery electrode to a certain extent, for different manufacturing processes, the detailed information of the corresponding manufacturing process will also be different, that is, there is a correlation between the semantic information corresponding to the two latent vectors. Therefore, aggregation can be performed so that the semantic information can be mutually reinforced, thereby improving the accuracy of the semantic representation. Moreover, since multi-level aggregation is performed, the aggregation can also be made more sufficient.

[0074] Step S140 , performing analysis based on the process semantic vector to obtain target performance analysis data.

[0075] In an embodiment of the present invention, after forming the process semantic vector, the performance analysis system can perform analysis based on the process semantic vector to obtain target performance analysis data. The target performance analysis data is used to characterize the performance of the target battery in relation to energy density. For example, after forming the process semantic vector, the process semantic vector can be fully connected to form a corresponding fully connected vector, wherein the size of the fully connected vector can be 1*1, and then, the fully connected vector can be processed by a linear regression function (such as identity mapping, etc.) to obtain the corresponding target performance analysis data, such as a value between 1-10, where a larger value represents a higher performance of the target battery in relation to energy density.

[0076] Based on the above content, on the one hand, semantic mining can be performed to obtain corresponding latent semantic information, so that performance analysis can be performed based on the latent semantic information. In this way, compared with the conventional actual testing scheme, it can have higher efficiency and is not easily interfered with by testers, so the reliability can also be higher. In addition, since not only the latent semantic information in the process data is mined, but also the semantic information in the production process, after multi-level aggregation, a process semantic vector with richer semantic information can be obtained. Moreover, since different process data and actual production processes have an impact on the performance of the formed battery electrode, the semantic representation accuracy of the process semantic vector that combines the two aspects of semantic information can also be guaranteed, thereby ensuring the reliability of the target performance analysis data obtained, and thus improving the problem of relatively low reliability and efficiency of performance analysis in the existing technology.

[0077] In addition, after the corresponding battery electrodes are manufactured based on the above-mentioned target manufacturing process data, they can be assembled with polymer solid electrolytes and lithium sheets into button batteries, thereby obtaining the target battery.

[0078] It should also be noted that with respect to the above-mentioned step S120, the specific process of performing semantic mining on the target manufacturing process data and the target manufacturing process image is not limited and can be selected and configured accordingly according to actual conditions.

[0079] For example, in a specific implementation, the target manufacturing process data may be embedded to obtain a corresponding manufacturing process latent vector, and the target manufacturing process image may be convolved to obtain a corresponding manufacturing process latent vector.

[0080] For example, in another specific embodiment, in order to improve the accuracy of semantic mining and make the obtained manufacturing process latent vector and manufacturing process latent vector more reliable, the above-mentioned step S120 can further include step S121 and step S122. The specific implementation process of each step is as follows.

[0081] Step S121 : performing an embedding operation on the target manufacturing process data to form a manufacturing process embedding vector, and obtaining a manufacturing process latent vector based on the manufacturing process embedding vector.

[0082] In an embodiment of the present invention, an embedding operation can be performed on the target manufacturing process data to form a manufacturing process embedding vector, and a manufacturing process latent vector can be obtained based on the manufacturing process embedding vector. Exemplarily, the embedding operation can be implemented through word embedding processing, so that the target manufacturing process data, which is text data, can be mapped into a vector space.

[0083] In step S122, a first convolution operation is performed on each image frame included in the target production process image to form an image convolution vector corresponding to each image frame, and a semantic extraction operation and a second convolution operation are performed on the image convolution vector corresponding to each image frame to form an image convolution depth vector corresponding to each image frame. The image convolution depth vectors corresponding to each image frame are spliced ​​to form a production process latent vector.

[0084] In an embodiment of the present invention, a first convolution operation can be performed on each image frame included in the target production process image to form an image convolution vector corresponding to each image frame, and a semantic extraction operation and a second convolution operation can be performed on the image convolution vector corresponding to each image frame (i.e., a semantic extraction operation is first performed to obtain a corresponding extraction vector, and then a second convolution operation is performed on the extraction vector. In addition, the specific implementation process of the semantic extraction operation is described below) to form an image convolution depth vector corresponding to each image frame, and the image convolution depth vectors corresponding to each image frame are spliced ​​to form a production process latent vector. The second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap. The first convolution operation can be a conventional convolution process. In addition, the size of the image frame can be the same as the size of the corresponding image convolution vector. Moreover, through the semantic extraction operation and the second convolution operation, the size of the vector can be reduced. For example, when the size of the image convolution vector is a*b, the size of the image convolution depth vector can be a / 2*b / 2, or a / 4*b / 4, or a / 8*b / 8, etc. The specific size can be selected according to actual needs. For example, when the number of image frames is large, the size of the image convolution depth vector can be smaller, and when the number of image frames is small, the size of the image convolution depth vector can be larger.

[0085] Optionally, in step S121 above, the specific process of obtaining the manufacturing process latent vector is not limited. For example, in a specific embodiment, considering that the target manufacturing process data includes processes of multiple stages and the semantic information of processes of different stages may be correlated, in order to mine such related semantic information, step S121 above may further include the following achievable content:

[0086] In the first step, embedding operations can be performed on the target manufacturing process data to form a manufacturing process embedding vector. For example, this can be achieved through a corresponding word embedding model. Specifically, the target manufacturing process data can be segmented. For example, the word group "coating, thickness, is, 250, um" can be obtained. Then, word embedding processing (WordEmbedding) can be performed on each word to obtain the word embedding vector corresponding to "coating", the word embedding vector corresponding to "thickness", the word embedding vector corresponding to "is", the word embedding vector corresponding to "250", and the word embedding vector corresponding to "um". Then, the word embedding vectors of each word can be concatenated to form the corresponding manufacturing process embedding vector. Exemplarily, the word embedding vector corresponding to "coating" can be [0.12, -0.43, 0.76, 0.55, -0.21, 0.34, -0.51, 0.89, -0.22, 0.01, -0.67, 0.33,......, 0.44], the word embedding vector corresponding to "thickness" can be [0.25, 0.78, -0.41, 0.10, 0.67, -0.28, 0.18, -0.55, 0.33, -0.11, 0.99, -0.02,......, 0.07], the word embedding vector corresponding to "is" can be [0.43, 0.56, -0.02, 0.10, -0.33, 0.76, 0.11, -0.48, 0.61, 0.22, 0.54, -0.89,......, 0.09], the word embedding vector corresponding to "250" can be [-0.12, 0.04, 0.15, -0.26, 0.87, 0.01, 0.32, -0.45, 0.71, 0.26, -0.18, 0.15,......, 0.23], and the word embedding vector corresponding to "um" can be [0.34, -0.56, 0.12, -0.77, 0.65, 0.98, -0.34, 0.29, -0.11, 0.44, 0.56, -0.03,......, -0.22];

[0087] In the second step, the manufacturing process embedding vector is decomposed to form a first process decomposition vector and a second process decomposition vector, and based on a plurality of predetermined depth information, the first process decomposition vector and the second process decomposition vector are respectively deeply mined to form a plurality of first process depth vectors and a plurality of second process depth vectors. For example, the manufacturing process embedding vector can be subjected to different pooling processes, such as mean pooling and maximum pooling, to capture different important semantic features in the manufacturing process embedding vector, thereby achieving corresponding decomposition, i.e., obtaining the corresponding first process decomposition vector and the second process decomposition vector. In addition, During mean pooling and maximum pooling, the stride can be 1 and edge padding is performed to ensure that the size of the first process decomposition vector and the size of the second process decomposition vector are the same as the size of the manufacturing process embedding vector. In addition, deep mining can refer to convolution processing. For example, the size of the first process depth vector of the first depth is 1 / 4 of the size of the first process decomposition vector (rows and columns are 1 / 2 respectively), and the size of the first depth vector of the second depth is 1 / 16 of the size of the first process decomposition vector (rows and columns are 1 / 4 respectively). In this way, convolutions of different depths can be used to capture high-level, abstract semantic features at different levels in the process decomposition vector.

[0088] In a third step, the multiple first process depth vectors and the second process depth vectors may be merged according to the corresponding depth information to form a corresponding multiple merged process depth vectors. For example, the first process depth vector of the first depth and the second process depth vector of the first depth may be merged to obtain a first merged process depth vector. In addition, the merging method may not be limited. For example, the first process depth vector of the first depth and the second process depth vector of the first depth may be added or averaged to obtain the first merged process depth vector.

[0089] In the fourth step, a manufacturing process characterization vector formed by training based on the sample manufacturing data and the corresponding performance label can be obtained, and the manufacturing process characterization vector can be optimized in turn based on the multiple merged process depth vectors to form an updated manufacturing process characterization vector; that is, the manufacturing process characterization vector actually represents the corresponding semantic information of the sample manufacturing data, and has a certain correlation with the semantic information of the above-mentioned target manufacturing process data, but there may be certain differences. In this way, the manufacturing process characterization vector can be optimized in turn based on the multiple merged process depth vectors, so that irrelevant semantic information in the manufacturing process characterization vector is discarded, and the relevant semantic information in the target manufacturing process data is captured; in addition, the sample manufacturing data can include sample manufacturing process data and sample manufacturing process images (that is, corresponding to the above-mentioned target manufacturing process data and target manufacturing process images). Based on this, the sample manufacturing process data and the sample manufacturing process images can be processed as described in steps S120-S140 above to obtain the corresponding sample performance analysis data, and then , the error (such as mean square error, etc.) between the sample performance analysis data and the performance label can be calculated, and then, based on the error, the corresponding parameters (such as the manufacturing process characterization vector, and in the initial stage, the manufacturing process characterization vector can be a randomly generated vector) are updated until the error converges; in addition, the optimization process can be as follows: the multiple merged process depth vectors are traversed in the direction from large to small to form the currently traversed merged process depth vector, and then, based on the currently traversed merged process depth vector, the output vector of the previous stage can be cross-attention processed (when the currently traversed merged process depth vector is the first traversed merged process depth vector, the manufacturing process characterization vector can be cross-attention processed) to form the cross-attention vector of the current stage, and the cross-attention vector of the current stage is convolved (the size is compressed to be consistent with the size of the next traversed merged process depth vector) to obtain the output vector of the current stage. In this way, the output vector of the last stage, that is, the updated manufacturing process characterization vector, can be obtained;

[0090] In a fifth step, a manufacturing process latent vector may be obtained based on the updated manufacturing process representation vector; and the updated manufacturing process representation vector may be used as the manufacturing process latent vector.

[0091] In addition, it should be noted that for the above performance labels, after the corresponding battery electrodes are manufactured based on the corresponding sample manufacturing process data, they can be further assembled to form batteries, and then the batteries can be electrochemically tested. Then, the test results can be evaluated by experts to obtain corresponding performance data. The test rate of the electrochemical performance test can be 0.2-2C (the results are as follows Figure 3 ), and carried out 200 charge and discharge life tests at 0.5C (the results are shown in Figure 4 shown).

[0092] It should also be noted that with respect to the above-mentioned step S130, the specific process of multi-level aggregation of the manufacturing process latent vector and the manufacturing process latent vector is not limited and can be selected and configured accordingly according to actual conditions.

[0093] For example, in a specific embodiment, the production process latent vector and the production process latent vector can be subjected to multiple semantic extraction operations respectively, and then the corresponding semantic extraction results can be added. Finally, the results of the addition can be spliced ​​to obtain the corresponding process semantic vector.

[0094] For example, in another specific embodiment, in order to improve the accuracy of multi-level aggregation and make the semantic representation ability of the formed process semantic vector better, the above-mentioned step S130 can further include the following steps S131, S132, S133, S134 and S135. The specific implementation process of each step is as follows.

[0095] Step S131 : performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector to form a first production aggregated vector.

[0096] In an embodiment of the present invention, the production process latent vector can be subjected to a semantic extraction operation and then aggregated with the production process latent vector to form a first production aggregate vector. In other words, key semantic information can be first extracted from the production process latent vector, and then aggregated with the semantic information in the production process latent vector to obtain a first production aggregate vector that carries semantic information from both the process and process dimensions. On the one hand, since the size of the production process latent vector is larger than that of the production process latent vector, the size of the production process latent vector can be reduced by performing a semantic extraction operation on the production process latent vector so that it can be adapted when aggregated with the production process latent vector. In addition, key information can also be captured through the semantic extraction operation.

[0097] Step S132 : performing semantic extraction on the production process latent vector and the production technique latent vector and then aggregating them to form a second production aggregated vector.

[0098] In an embodiment of the present invention, the production process latent vector and the production technique latent vector can be subjected to semantic extraction and then aggregated to form a second aggregated production vector. In other words, key semantic information can be extracted from both the production process latent vector and the production technique latent vector before being aggregated, thereby achieving aggregation of key semantic information. Based on this, steps S131 and S132 can aggregate key semantic information at different depths, thereby capturing more key semantic information and improving the richness of the represented semantic information.

[0099] Step S133 : performing a semantic adaptation operation on the first production aggregation vector and the second production aggregation vector and then performing bidirectional aggregation to form a third production aggregation vector and a fourth production aggregation vector.

[0100] In an embodiment of the present invention, after forming the first production aggregation vector and the second production aggregation vector, the first production aggregation vector and the second production aggregation vector can be semantically adapted and then bidirectionally aggregated to form a third production aggregation vector and a fourth production aggregation vector. As previously described, the first production aggregation vector and the second production aggregation vector actually have different sizes. For example, the size of the first production aggregation vector is larger than the size of the second production aggregation vector. In this way, semantic adaptation can be performed at each of the two sizes, and corresponding aggregation can be performed separately to achieve aggregation at different sizes, thereby enriching semantic information in the process of capturing key semantic information.

[0101] Step S134 : performing a semantic expansion operation on the third production aggregation vector and the fourth production aggregation vector and then performing bidirectional aggregation to form a fifth production aggregation vector and a sixth production aggregation vector.

[0102] In an embodiment of the present invention, after forming the third and fourth production aggregation vectors, the third and fourth production aggregation vectors can be semantically expanded and then bidirectionally aggregated to form fifth and sixth production aggregation vectors. As previously described, the third and fourth production aggregation vectors actually have different sizes. For example, the size of the third production aggregation vector is larger than the size of the fourth production aggregation vector. In this way, semantic adaptation can be performed at each of the two sizes, and corresponding aggregation can be performed separately to achieve aggregation at different sizes, thereby enriching semantic information while capturing key semantic information.

[0103] Step S135 : performing aggregation based on the fifth production aggregation vector and the sixth production aggregation vector to form a process semantic vector.

[0104] In an embodiment of the present invention, after forming the fifth and sixth production aggregation vectors, aggregation can be performed based on the fifth and sixth production aggregation vectors to form a process semantic vector. As described above, the third and fourth production aggregation vectors actually have different sizes. For example, the size of the third production aggregation vector is larger than that of the fourth production aggregation vector. In this way, a semantic extraction operation can be performed on the fifth production aggregation vector to compress the semantic information (the size can be consistent with the size of the sixth production aggregation vector), thereby capturing more advanced abstract semantic features. In this way, the compressed fifth production aggregation vector can be subjected to cross-attention processing based on the sixth production aggregation vector to form a process semantic vector. In this way, the association and fusion of semantic information in the two dimensions of process and process can be achieved. In other embodiments, the sixth production aggregation vector can also be subjected to cross-attention processing based on the compressed fifth production aggregation vector to form a process semantic vector. Alternatively, the results of the two cross-attention processing operations can be added together to obtain the process semantic vector.

[0105] Alternatively, in the above step S131, the specific process of performing semantic extraction on the production process latent vector and aggregating it with the production process latent vector is not limited. For example, in a specific embodiment, in order to improve the semantic representation ability of the obtained first production aggregate vector, the above step S131 may further include the following achievable content (combined with Figure 5 ):

[0106] In the first step, the production process latent vector may be subjected to a semantic extraction operation and then aggregated (e.g., concatenated) with the production technique latent vector. A first convolution operation may be performed on the aggregated vector to form a first aggregated convolution vector. The semantic extraction operation may include convolution and / or pooling to extract semantic information (and after the semantic extraction operation, the size of the production process latent vector may be compressed to the size of the production technique latent vector).

[0107] In the second step, the production process latent vector and the production process latent vector can be subjected to semantic extraction operations respectively and then aggregated (i.e., the production process latent vector after the semantic extraction operation and the production process latent vector after the semantic extraction operation are spliced, and the sizes of the two-dimensional vectors after the semantic extraction operation can be the same), and a second convolution operation is performed on the aggregated vector to form a second aggregated convolution vector;

[0108] In the third step, the first aggregated convolution vector can be subjected to a semantic extraction operation and then aggregated with the second aggregated convolution vector, and a first convolution operation can be performed on the aggregated vector to form a first aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap, so that the convolution operation can cover a larger area of ​​input data without increasing the number of convolution kernel parameters, thereby capturing a wider range of contextual information. The first convolution operation can be a conventional convolution, which can achieve size compression and capture high-level semantic information.

[0109] Optionally, in step S132, the specific process of performing semantic extraction on the production process latent vector and the production technique latent vector and then aggregating them is not limited. For example, in a specific embodiment, to improve the semantic representation capability of the obtained second production aggregated vector, step S132 may further include the following achievable content:

[0110] In the first step, the production process latent vector may be subjected to a semantic extraction operation and then aggregated with the production technique latent vector, and a first convolution operation may be performed on the aggregated vector to form a first aggregated convolution vector, as described above.

[0111] In the second step, the production process latent vector and the production technique latent vector may be subjected to semantic extraction operations respectively and then aggregated, and a second convolution operation may be performed on the aggregated vectors to form a second aggregated convolution vector, as described above.

[0112] In the third step, the first aggregated convolution vector and the second aggregated convolution vector can be subjected to a semantic extraction operation and then aggregated (that is, while capturing high-level semantic information through the semantic extraction operation, the size is compressed to be consistent, and then splicing can be performed to achieve aggregation of semantic information), and a second convolution operation is performed on the aggregated vector to form a second aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0113] Optionally, in step S133, the specific process of performing the semantic adaptation operation on the first production aggregation vector and the second production aggregation vector and then performing bidirectional aggregation is not limited. For example, in a specific embodiment, in order to improve the semantic representation capabilities of the obtained third production aggregation vector and the fourth production aggregation vector, step S133 may further include the following achievable content:

[0114] In the first step, the second aggregated vector can be semantically expanded and then aggregated with the first aggregated vector. The aggregated vector can then be subjected to a first convolution operation to form a third aggregated vector. In other words, the second aggregated vector can first be semantically expanded to have the same size as the first aggregated vector. The vectors can then be concatenated to achieve aggregation. Finally, further convolution can be performed to capture high-level semantic information from the aggregated vector.

[0115] In the second step, the first production aggregation vector can be subjected to a semantic extraction operation and then aggregated with the second production aggregation vector, and a second convolution operation can be performed on the aggregated vector to form a fourth production aggregation vector. That is, the first production aggregation vector can be subjected to a semantic extraction operation first so that its size is adjusted to the same size as the second production aggregation vector, and then splicing is performed to achieve aggregation. Finally, through further convolution, more contextual information can be captured, that is, semantic association mining is achieved, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve the expansion of semantic information, the semantic extraction operation includes convolution and / or pooling to achieve the extraction of semantic information, and the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

[0116] Optionally, in the above-mentioned step S134, the specific process of performing a semantic expansion operation on the third production aggregation vector and the fourth production aggregation vector and then performing bidirectional aggregation is not limited. For example, in a specific embodiment, in order to improve the semantic representation capabilities of the obtained fifth production aggregation vector and the sixth production aggregation vector, the above-mentioned step S134 may further include the following steps S134a and S134b. The specific implementation process of each step is as follows.

[0117] Step S134a: Performing semantic expansion operations on the third aggregated vector and the fourth aggregated vector respectively and then aggregating them to form a fifth aggregated vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information.

[0118] In an embodiment of the present invention, the third production aggregation vector and the fourth production aggregation vector can be subjected to semantic expansion operations respectively and then aggregated (i.e., by performing the semantic expansion operations separately, the sizes of the third production aggregation vector and the fourth production aggregation vector are adjusted to be consistent, and then they can be spliced) to form a fifth production aggregation vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information (i.e., the size of the vector can be increased).

[0119] Step S134b: Performing a semantic expansion operation on the fourth production aggregation vector and then aggregating the resultant vector with the third production aggregation vector to form a sixth production aggregation vector.

[0120] In an embodiment of the present invention, the fourth aggregated vector can be semantically expanded and then aggregated with the third aggregated vector to form a sixth aggregated vector. In other words, the smaller fourth aggregated vector can be semantically expanded to increase its size to match that of the third aggregated vector, and then the vectors can be concatenated.

[0121] Optionally, in the above step S134b, the specific process of performing a semantic expansion operation on the fourth aggregated vector and then aggregating it with the third aggregated vector is not limited. For example, in a specific embodiment, in order to avoid the problem of semantic distortion caused by the increase in the depth of aggregation in multi-level aggregation, so as to improve the semantic representation accuracy of the sixth aggregated vector, the above step S134b may further include the following achievable content (combined with Figure 6 shown):

[0122] In the first step, the production process latent vector may be subjected to a semantic extraction operation and then aggregated with the production technique latent vector, and a first convolution operation may be performed on the aggregated vector to form a first aggregated convolution vector, as described above.

[0123] In the second step, the production process latent vector and the production technique latent vector may be subjected to semantic extraction operations respectively and then aggregated, and a second convolution operation may be performed on the aggregated vectors to form a second aggregated convolution vector, as described above.

[0124] In a third step, the fourth aggregated vector and the third aggregated vector may be semantically expanded and then aggregated, and the aggregated vector and the first aggregated convolution vector may be added together (it is understood that the sizes of the aggregated vector and the first aggregated convolution vector may be the same or different. If they are different, the sizes may be adjusted to be the same through convolution or other processing, and then the vectors may be added together, i.e., the vector parameters at corresponding positions are added together). Furthermore, a first convolution operation may be performed on the added vector to form a first added convolution vector.

[0125] In the fourth step, the fourth aggregated vector can be semantically expanded and then aggregated with the third aggregated vector, and the aggregated vector and the second aggregated convolution vector are added (it can be understood that the sizes of the aggregated vector and the second aggregated convolution vector can be the same or different. When they are different, the sizes can be adjusted to the same through convolution or other processing, and then added, that is, the vector parameters at corresponding positions are added), and a second convolution operation is performed on the added vector to form a second added convolution vector;

[0126] In the fifth step, the first added convolution vector and the second added convolution vector can be semantically expanded and then aggregated, and the aggregated vector and the production process latent vector can be added together (it can be understood that the sizes of the aggregated vector and the production process latent vector can be the same or different. When they are different, the sizes can be adjusted to the same through convolution or other processing, and then added together, that is, the vector parameters at the corresponding positions are added), and the first convolution operation is performed on the added vector to form a fifth production aggregate vector; based on this, the direct fusion of the semantic information in the production process latent vector can be achieved, avoiding the problem of corresponding semantic information loss due to excessive depth in the aforementioned multi-level aggregation operation; in addition, it should be noted that the fifth step is an optional step, that is, the fifth production aggregate vector can be formed in the aforementioned manner or in the fifth step, and can be selected according to actual needs;

[0127] In the sixth step, the second added convolution vector can be semantically expanded and then aggregated with the first added convolution vector, and the aggregated vector and the manufacturing process latent vector can be added (it can be understood that the sizes of the aggregated vector and the manufacturing process latent vector can be the same or different. When they are different, the sizes can be adjusted to the same through convolution and other processing, and then added, that is, the vector parameters at the corresponding positions are added), and the vector formed by the addition is subjected to a second convolution operation to form a sixth manufacturing aggregate vector; based on this, the direct fusion of the semantic information in the manufacturing process latent vector can be achieved, avoiding the problem of corresponding semantic information loss due to excessive depth in the aforementioned multi-level aggregation operation.

[0128] In summary, the performance analysis method and system for solid-state lithium-ion batteries provided by the present invention first obtain the target manufacturing process data and target manufacturing process images of the battery electrode; secondly, semantic mining is performed on the target manufacturing process data and target manufacturing process images respectively to form a manufacturing process latent vector and a manufacturing process latent vector; then, the manufacturing process latent vector and the manufacturing process latent vector are multi-level aggregated to form a process semantic vector; finally, analysis is performed based on the process semantic vector to obtain target performance analysis data. Based on the above content, on the one hand, semantic mining can be performed to obtain corresponding latent semantic information, so that performance analysis can be performed based on the latent semantic information. In this way, compared with the conventional actual testing scheme, it can have higher efficiency and is not easily interfered with by the tester, so the reliability can also be higher. In addition, since not only the potential semantic information in the process data is mined, but also the semantic information in the production process, after multi-level aggregation, a process semantic vector with richer semantic information can be obtained. Moreover, since different process data and actual production processes have an impact on the performance of the formed battery electrode, the semantic representation accuracy of the process semantic vector that combines the two aspects of semantic information can also be guaranteed, thereby ensuring the reliability of the target performance analysis data obtained, and then improving the problem of relatively low reliability and efficiency of performance analysis in the existing technology.

[0129] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0130] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0131] If the functions are implemented in the form of software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion 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 for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprise," "include," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or device that includes a series of elements includes not only those elements but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0132] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for analyzing the performance of a solid-state lithium-ion battery, characterized in that: include: Acquiring target manufacturing process data and target manufacturing process images of a battery electrode of a target battery, wherein the target battery is an all-solid-state lithium-ion battery; Performing an embedding operation on the target manufacturing process data to form a manufacturing process embedding vector, and obtaining a manufacturing process latent vector based on the manufacturing process embedding vector; performing a first convolution operation on each image frame included in the target manufacturing process image to form an image convolution vector corresponding to each image frame, and performing a semantic extraction operation and a second convolution operation on the image convolution vector corresponding to each image frame to form an image convolution depth vector corresponding to each image frame, and splicing the image convolution depth vectors corresponding to each image frame to form a manufacturing process latent vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap, wherein the size of the manufacturing process latent vector is larger than the size of the manufacturing process latent vector; Performing multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector, wherein the process semantic vector is used to represent the global semantic information of the manufacturing process latent vector and the manufacturing process latent vector; An analysis is performed based on the process semantic vector to obtain target performance analysis data, wherein the target performance analysis data is used to characterize the performance of the target battery related to energy density.

2. The performance analysis method of a solid-state lithium-ion battery according to claim 1, characterized in that: The step of embedding the target manufacturing process data to form a manufacturing process embedding vector, and obtaining a manufacturing process latent vector based on the manufacturing process embedding vector includes: Performing an embedding operation on the target manufacturing process data to form a manufacturing process embedding vector; Decomposing the manufacturing process embedding vector to form a first process decomposition vector and a second process decomposition vector, and, based on a plurality of predetermined depth information, performing depth mining on the first process decomposition vector and the second process decomposition vector to form a plurality of corresponding first process depth vectors and a plurality of second process depth vectors; Merging the plurality of first process depth vectors and the second process depth vector according to corresponding depth information to form a plurality of corresponding merged process depth vectors; Obtaining a manufacturing process characterization vector formed by training based on sample manufacturing data and corresponding performance labels, and optimizing the manufacturing process characterization vectors in sequence based on the multiple merged process depth vectors to form an updated manufacturing process characterization vector; Based on the updated manufacturing process representation vector, a manufacturing process latent vector is obtained.

3. The method for analyzing the performance of a solid-state lithium-ion battery according to any one of claims 1 to 2, wherein: The step of performing multi-level aggregation on the manufacturing process latent vector and the manufacturing process latent vector to form a process semantic vector includes: Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector to form a first production aggregated vector; Performing semantic extraction on the production process latent vector and the production technique latent vector and then aggregating them to form a second production aggregated vector; Performing semantic adaptation on the first production aggregation vector and the second production aggregation vector and then bidirectionally aggregating them to form a third production aggregation vector and a fourth production aggregation vector; Performing semantic expansion on the third production aggregation vector and the fourth production aggregation vector and then bidirectionally aggregating them to form a fifth production aggregation vector and a sixth production aggregation vector; Aggregation is performed based on the fifth production aggregation vector and the sixth production aggregation vector to form a process semantic vector.

4. The performance analysis method of a solid-state lithium-ion battery according to claim 3, characterized in that: The step of performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector to form a first production aggregated vector includes: performing a semantic extraction operation on the production process latent vector and aggregating the resulting vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector, wherein the semantic extraction operation includes convolution and / or pooling to extract semantic information; Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector; The first aggregated convolution vector is subjected to a semantic extraction operation and then aggregated with the second aggregated convolution vector, and a first convolution operation is performed on the aggregated vector to form a first aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

5. The performance analysis method of a solid-state lithium-ion battery according to claim 3, characterized in that: The step of performing a semantic extraction operation on the production process latent vector and the production technique latent vector and then aggregating them to form a second production aggregated vector includes: Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector; Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector; The first aggregated convolution vector and the second aggregated convolution vector are subjected to a semantic extraction operation and then aggregated, and a second convolution operation is performed on the aggregated vector to form a second aggregated vector, wherein the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

6. The performance analysis method of a solid-state lithium-ion battery according to claim 3, characterized in that: The step of performing a semantic adaptation operation on the first production aggregation vector and the second production aggregation vector and then bidirectionally aggregating the vectors to form a third production aggregation vector and a fourth production aggregation vector includes: performing a semantic expansion operation on the second produced aggregate vector and then aggregating the vector with the first produced aggregate vector, and performing a first convolution operation on the aggregated vector to form a third produced aggregate vector; The first aggregated vector is subjected to a semantic extraction operation and then aggregated with the second aggregated vector, and a second convolution operation is performed on the aggregated vector to form a fourth aggregated vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information, the semantic extraction operation includes convolution and / or pooling to achieve extraction of semantic information, and the second convolution operation is different from the first convolution operation, and in the second convolution operation, the receptive field is increased by increasing the convolution gap.

7. The performance analysis method of a solid-state lithium-ion battery according to claim 3, characterized in that: The step of performing a semantic expansion operation on the third production aggregation vector and the fourth production aggregation vector and then performing bidirectional aggregation to form a fifth production aggregation vector and a sixth production aggregation vector includes: performing a semantic expansion operation on the third aggregate vector and the fourth aggregate vector respectively and then aggregating them to form a fifth aggregate vector, wherein the semantic expansion operation includes transposed convolution and / or interpolation to achieve expansion of semantic information; The fourth production aggregation vector is subjected to a semantic expansion operation and then aggregated with the third production aggregation vector to form a sixth production aggregation vector.

8. The performance analysis method of a solid-state lithium-ion battery according to claim 7, characterized in that: The step of performing a semantic expansion operation on the fourth production aggregate vector and then aggregating the resultant vector with the third production aggregate vector to form a sixth production aggregate vector includes: Performing a semantic extraction operation on the production process latent vector and aggregating the vector with the production technique latent vector, and performing a first convolution operation on the aggregated vector to form a first aggregated convolution vector; Performing semantic extraction operations on the production process latent vector and the production technique latent vector respectively and then aggregating them, and performing a second convolution operation on the aggregated vectors to form a second aggregated convolution vector; Performing semantic expansion operations on the fourth aggregated vector and the third aggregated vector respectively and then aggregating them, adding the aggregated vector and the first aggregated convolution vector, and performing a first convolution operation on the added vector to form a first added convolution vector; Performing a semantic expansion operation on the fourth produced aggregate vector and then aggregating the resulting aggregate vector with the third produced aggregate vector, adding the aggregated vector and the second aggregate convolution vector, and performing a second convolution operation on the added vector to form a second added convolution vector; The second additive convolution vector is semantically expanded and then aggregated with the first additive convolution vector, and the aggregated vector and the manufacturing process latent vector are added together. Furthermore, a second convolution operation is performed on the added vector to form a sixth manufacturing aggregated vector.

9. A solid-state lithium-ion battery performance analysis system, characterized in that: include: memory for storing computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the performance analysis method of the solid-state lithium-ion battery according to any one of claims 1 to 8.

Images