Lithium-ion battery automatic testing system and method

Abstract

This application discloses an automatic lithium-ion battery testing system and method, which relates to the field of intelligent testing. The system collects voltage and capacity data from the negative electrode of a lithium-ion battery during a charging test, then performs differential processing on this data to generate a capacity-voltage differential curve. This differential curve is then densely sampled to obtain a series of detailed capacity-voltage data points. Finally, these dense data points are used to determine whether lithium plating is occurring within the battery. This allows for accurate monitoring and assessment of the internal state of a lithium-ion battery during charging, particularly lithium plating, which is crucial for battery performance evaluation, safety testing, and life prediction.

Description

Technical Field

[0001] The present application relates to the field of intelligent testing, and more specifically, to a lithium-ion battery automatic testing system and method thereof. Background Art

[0002] With the rapid development of electric vehicles (EVs) and renewable energy storage systems, lithium-ion batteries are becoming increasingly widely used as efficient energy storage devices. However, during use, lithium-ion batteries may experience lithium plating, the deposition of metallic lithium on the surface of the anode. This phenomenon not only degrades battery performance and lifespan but can also lead to safety hazards such as thermal runaway and short circuits. Therefore, accurately detecting and predicting the occurrence of lithium plating is crucial to ensuring battery safety and reliability.

[0003] Patent CN118980945A discloses a method and system for testing the critical point of lithium deposition in lithium-ion batteries. Using lithium iron phosphate (LFP) as the positive electrode material, the method accurately collects negative electrode voltage and capacity data through constant current charging. This data is then differentiated to generate a capacity-voltage differential curve, clearly identifying changes in the lithium deposition characteristics. Finally, the onset of lithium deposition is accurately determined based on the specific inflection point in the differential curve, ensuring high accuracy and reliability of the test results. This method is suitable for long-term monitoring of battery status at different stages of aging.

[0004] This patent determines the starting point (critical point) of lithium precipitation manually or by embedding a judgment algorithm in the system. However, manual analysis requires a lot of time to review data and charts, which is extremely inefficient, especially when dealing with large-scale data sets. In addition, this method is highly dependent on the experience and technical level of the operator. Different people may come to different conclusions, lacking consistency and objectivity. Although the judgment algorithm embedded in the system improves efficiency, it is usually based on preset thresholds or simple mathematical models. These settings may not be flexible enough to capture complex changes in electrochemical dynamic behavior. As the battery state changes, the fixed algorithm may not accurately reflect the actual situation and requires frequent adjustments to maintain accuracy.

[0005] Therefore, an optimized automatic testing solution for lithium-ion batteries is desired. Summary of the Invention

[0006] In order to solve the above technical problems, this application is proposed.

[0007] According to one aspect of the present application, a method for automatically testing a lithium-ion battery is provided, comprising: S1: collecting voltage and capacity data of a negative electrode of a tested lithium-ion battery during a charging test; S2: performing differential processing on the collected voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; S3: densely sampling the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; S4: judging the internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points;

[0008] Wherein, said S4 includes:

[0009] Embedding and encoding each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features;

[0010] Performing multi-scale semantic implicit coding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features;

[0011] Performing capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregate features, including: calculating an aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; performing dynamic aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features based on the aggregation center representation to obtain the capacity-voltage semantic significant aggregate features;

[0012] According to the capacity-voltage semantically significant aggregation feature, a critical point at which lithium deposition begins to occur in the tested lithium-ion battery is obtained.

[0013] According to another aspect of the present application, a lithium-ion battery automatic testing system is provided, comprising: a charging data acquisition module for acquiring voltage and capacity data of a negative electrode of a tested lithium-ion battery during a charging test; a voltage-capacity differential processing module for performing differential processing on the acquired voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; a differential curve dense sampling module for performing dense sampling on the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; and a battery internal lithium deposition state determination module for determining the battery internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points. The battery internal lithium deposition state determination module comprises:

[0014] a capacity-voltage embedding coding unit, configured to embed and code each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedding features;

[0015] a capacity-voltage semantic multi-scale encoding unit, configured to perform multi-scale semantic implicit encoding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit encoding features;

[0016] A capacity-voltage semantic aggregation unit is configured to perform capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregate features. The capacity-voltage semantic aggregation unit includes: a capacity-voltage semantic feature aggregation center representation subunit, configured to calculate the aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; and a capacity-voltage semantic dynamic aggregation subunit, configured to perform dynamic aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features based on the aggregation center representation to obtain the capacity-voltage semantic significant aggregate features.

[0017] The lithium deposition critical point judgment unit is used to obtain the lithium deposition critical point at which lithium deposition begins to occur in the tested lithium-ion battery according to the capacity-voltage semantic significant aggregation.

[0018] Compared with the prior art, the present application provides a lithium-ion battery automatic testing system and method thereof, which utilizes data analysis and coding technology based on deep learning to embed and encode a dense sequence of each capacity-voltage point, and then performs multi-scale implicit coding on each capacity-voltage embedded feature, thereby intelligently identifying the lithium deposition critical point where lithium deposition begins to occur in the tested lithium-ion battery based on the dynamic pressure semantic significant aggregation representation between the multi-scale implicit coding features of each capacity-voltage semantic. In this way, full automation of data processing and critical point identification is achieved without manual intervention, ensuring the consistency and objectivity of the results. It can also capture complex electrochemical dynamic behavior changes, rather than relying solely on preset thresholds or simple mathematical models, making the model more flexible and able to accurately reflect the true state inside the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 Flowchart of a lithium-ion battery automatic testing method according to an embodiment of the present application.

[0021] Figure 2Schematic diagram of data flow in step S4 of the automatic testing method for lithium-ion batteries according to an embodiment of the present application.

[0022] Figure 3 4 is a flow chart of step S43 in the lithium-ion battery automatic testing method according to an embodiment of the present application.

[0023] Figure 4 4 is a flow chart of step S431 in the lithium-ion battery automatic testing method according to an embodiment of the present application.

[0024] Figure 5 FIG. 4 is a block diagram of a lithium-ion battery automatic testing system according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. While the drawings illustrate certain embodiments of the present disclosure, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0026] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in a different order and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0027] It should be noted that the modifications of "one" and "multiple" mentioned in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0028] Therefore, in response to the problems in the above background technology, the present application proposes a lithium-ion battery automatic testing method. Figure 1 Flowchart of the automatic testing method for lithium-ion batteries according to an embodiment of the present application. Figure 1 As shown, according to an embodiment of the present application, the automatic testing method for lithium-ion batteries includes: S1: collecting voltage and capacity data of the negative electrode of the tested lithium-ion battery during a charging test; S2: performing differential processing on the collected voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; S3: densely sampling the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; S4: judging the internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points.

[0029] Specifically, in step S1, the voltage and capacity data of the negative electrode of the tested lithium-ion battery are collected during the charging test. It should be understood that the negative electrode voltage data refers to the change in potential of the negative electrode relative to the positive electrode or relative to a reference point (for example, the battery casing in some cases) during the charging process. As lithium ions are released from the positive electrode and embedded in the negative electrode material, the potential of the negative electrode changes accordingly. These changes can provide clues as to whether lithium deposition (i.e., lithium plating) has begun on the negative electrode surface. Capacity refers to the amount of charge a battery can store and release under specific conditions, typically expressed in milliampere-hours (mAh). During the charging process, capacity data reflects how much lithium ions have been successfully embedded in the negative electrode material. When lithium plating occurs, some lithium is no longer reversibly embedded in the negative electrode material and is instead deposited as metallic lithium, resulting in a reduction in available capacity. It is worth noting that lithium plating typically occurs at the negative electrode. When the battery is overcharged, at low temperature or charged quickly, lithium ions may not be able to be embedded in the negative electrode material (such as graphite) in time, but instead are deposited as metallic lithium on the negative electrode surface. This phenomenon can be identified by monitoring the voltage changes of the negative electrode, because lithium plating will cause a voltage platform or abnormal changes in a specific voltage range.

[0030] Specifically, before any actual data acquisition begins, it's crucial to ensure that the equipment and environmental conditions used meet rigorous experimental standards. This includes selecting high-precision, high-stability measuring instruments, such as digital multimeters or multi-channel data loggers, that can capture voltage and current changes in real time at a sufficiently high frequency (typically milliseconds) and with sufficient resolution to capture subtle differences. Furthermore, a controlled experimental environment must be established to eliminate the effects of external factors on the measurement results, such as temperature fluctuations or electromagnetic interference.

[0031] Next, start collecting data. In order to obtain reliable negative electrode voltage data, a probe is usually set between the negative and positive electrodes of the battery or at the negative electrode relative to a reference point (such as the battery casing). As lithium ions are extracted from the positive electrode and embedded in the negative electrode material, the potential of the negative electrode will change accordingly. These changes can provide clues as to whether lithium deposition (i.e., lithium plating) on the negative electrode surface has begun to occur. At the same time, in order to obtain capacity data, it is necessary to continuously monitor the current and time flowing into the battery during the charging process, and obtain the total power input by integral calculation to determine the amount of power that the battery can store and release under specific conditions.

[0032] Throughout the charging test, the system continuously collects the negative electrode's voltage readings and corresponding capacity values. To improve data quality, the sampling interval must be minimized to more accurately record every moment's status. Furthermore, data transmission stability and integrity must be ensured to avoid data loss or errors due to network failures or other reasons. To this end, redundant designs, such as dual-channel simultaneous recording or local caching mechanisms, can be employed to ensure that critical information is fully preserved even in emergencies.

[0033] Finally, all collected data will be stored in a specially constructed database for subsequent analysis. Given the importance of lithium-ion battery performance evaluation, this database must not only have efficient data storage capabilities but also support complex query operations, allowing researchers to easily retrieve the complete historical records of a group of batteries within a specific time period. This will provide solid data support for the safety and reliability of lithium-ion batteries.

[0034] Specifically, in step S2, the collected voltage and capacity data of the negative electrode are differentiated to obtain a capacity-voltage differential curve. Accordingly, considering that during the charging process, when lithium deposition begins, the electrochemical behavior of the negative electrode surface will change significantly. This change may not be obvious on the original voltage-capacity curve, but by calculating the rate of change of capacity with voltage (i.e., dQ / dV), these subtle changes can be amplified. Therefore, the inflection point where lithium deposition occurs will be more clearly visible on the capacity-voltage differential curve, which helps to more accurately determine the critical point where lithium deposition begins. In particular, the traditional reference electrode-based method is easily affected by internal polarization of the battery, resulting in errors in the test results. The dQ / dV curve obtained through differential processing, whose identification criterion is based on the rate of change of voltage rather than the absolute voltage value, can better eliminate the interference caused by polarization and provide more reliable test results.

[0035] In practice, once the negative electrode voltage and capacity data during charging are obtained, a numerical differentiation algorithm is used to calculate the ratio of the capacity change to the voltage change at each time point. The capacity here refers to the amount of electricity that the battery can store and release under specific conditions, usually expressed in milliampere hours (mAh); while the voltage refers to the change in potential of the negative electrode relative to the positive electrode or a reference point. When lithium plating begins, the electrochemical behavior of the negative electrode surface will change significantly. This change may not be obvious on the original voltage-capacity curve, but after differentiation, these inflection points will be more clearly visible on the capacity-voltage differential curve, which helps to more accurately determine the critical point where lithium plating begins.

[0036] To achieve this, a series of mathematical tools and techniques are used. For example, the difference quotient method can be applied to approximate the derivative, or a more complex finite difference formula can be used to improve the accuracy. In addition, considering that noise may interfere with the results, smoothing techniques such as moving average filters can be introduced to reduce the impact of random fluctuations and ensure that the final differential curve is smooth and undistorted. At the same time, since the traditional reference electrode-based method is easily affected by the internal polarization of the battery, resulting in errors in the test results, the dQ / dV curve obtained by differential processing is based on the voltage change rate rather than the absolute voltage value, which better eliminates the interference caused by polarization and provides more reliable test results.

[0037] In this way, by performing differential processing on the collected negative electrode voltage and capacity data, not only can the subtle changes that may exist during the charging process be amplified, but also a reliable measurement method that is not affected by polarization can be provided, laying a solid foundation for the subsequent judgment of the critical point of lithium plating.

[0038] Specifically, in step S3, the capacity-voltage differential curve is densely sampled to obtain a dense sequence of capacity-voltage points. It should be understood that, considering that there are different change points between the various data points in the capacity-voltage differential curve, in order to be able to understand and analyze the numerical information in each sampling point more carefully and accurately, so as to better understand the change points on the curve, in the technical solution of the present application, the capacity-voltage differential curve is densely sampled to obtain a dense sequence of capacity-voltage points. That is, by dense sampling, the number of data points can be significantly increased, thereby increasing the resolution of the curve, which helps to capture subtle changes or feature points that may exist on the capacity-voltage differential curve, such as spikes, valleys or other discontinuities. These features are crucial for identifying specific electrochemical events (such as lithium precipitation), and by dense sampling, more data points can be provided for analysis, making it more accurate in determining the starting point or end point of important electrochemical reactions. For example, when detecting lithium precipitation, dense data points can help more accurately locate the critical point where lithium precipitation begins.

[0039] The core of dense sampling lies in choosing the right sampling interval and strategy. The ideal sampling scheme should ensure that all important features across the entire voltage range are covered without overloading the computational burden due to too many data points. To this end, the sampling parameters are set based on the specific characteristics of the battery and the intended analysis objectives. For example, in areas where voltage changes are more drastic, a shorter sampling interval may be used to capture rapid changes; in relatively flat areas, the interval can be appropriately relaxed to reduce redundant information. In addition, considering the important electrochemical reactions that may occur near certain key voltage points, special attention will be paid to these locations to ensure sufficient data density for subsequent analysis.

[0040] After dense sampling, the result is a dense sequence of numerous discrete points, each containing information about capacity change at a specific voltage. This sequence provides rich material for subsequent deep learning models and feature extraction. Especially when dealing with complex electrochemical behaviors, such as lithium plating, dense data points provide more detail, helping to more accurately pinpoint the critical point where plating begins.

[0041] Specifically, in step S4, the internal lithium deposition state of the tested lithium-ion battery is judged based on the dense sequence of the capacity-voltage points. Accordingly, the technical concept of the present application is to use data analysis and coding technology based on deep learning to embed the dense sequence of each capacity-voltage point, and then perform multi-scale implicit coding on each capacity-voltage embedded feature, so as to intelligently identify the lithium deposition critical point where lithium deposition begins to occur in the tested lithium-ion battery based on the dynamic pressure semantic significant aggregation representation between the multi-scale implicit coding features of each capacity-voltage semantic. In this way, full automation of data processing and critical point identification is achieved without manual intervention, ensuring the consistency and objectivity of the results. It can also capture complex changes in electrochemical dynamic behavior, rather than just relying on preset thresholds or simple mathematical models, making the model more flexible and able to accurately reflect the true state inside the battery.

[0042] Figure 2 Flowchart of step S4 in the automatic testing method for lithium-ion batteries according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 2 As shown, the S4 includes: S41, embedding and encoding each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features; S42, performing multi-scale semantic implicit encoding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit encoding features; S43, performing capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of capacity-voltage semantic multi-scale implicit encoding features to obtain capacity-voltage semantic significant aggregation features; S44, obtaining the critical point of lithium plating at which lithium plating begins to occur in the tested lithium-ion battery according to the capacity-voltage semantic significant aggregation features.

[0043] In step S41, embedding coding is performed on each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features. Specifically, in an embodiment of the present application, embedding coding is performed on each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features, including: embedding coding each capacity-voltage point in the dense sequence of capacity-voltage points using a capacity-voltage embedding coding matrix to obtain a dense sequence of capacity-voltage embedded coding vectors as the dense sequence of capacity-voltage embedded features.

[0044] It should be understood that, considering that each capacity-voltage point contains the semantic association relationship between different capacities and voltages, in order to better represent its inherent characteristic information, in the technical solution of the present application, a capacity-voltage embedding coding matrix is used to embed the individual capacity-voltage points in the dense sequence of the capacity-voltage points to reduce the originally sparse or redundant data dimensions to a smaller space that can still capture key information, thereby obtaining a dense sequence of capacity-voltage embedded coding vectors. It is worth mentioning that the capacity-voltage embedding coding matrix converts the original capacity-voltage point into a representation in a low-dimensional vector space. Embedding coding through the capacity-voltage embedding coding matrix can capture the complex patterns and inherent structures in different capacities and voltages, so that the battery characteristics under different states can be effectively distinguished in a low-dimensional space. In particular, the capacity-voltage embedding coding matrix is constructed by collecting a large amount of historical voltage and capacity data.

[0045] The following is a specific process of embedding and encoding each capacity-voltage point in the dense sequence of capacity-voltage points using the capacity-voltage embedding encoding matrix:

[0046] First, before implementing embedded coding, a capacity-voltage embedding matrix must be constructed. This matrix is obtained by collecting a large amount of historical voltage and capacity data. It converts the original capacity-voltage points into a representation in a low-dimensional vector space. Constructing such an encoding matrix requires extensive domain knowledge and strong machine learning capabilities to ensure that the final model accurately reflects the characteristics of the existing data while also exhibiting good generalization performance and being applicable to unseen new data.

[0047] The next step is to apply this to a dense sequence of actual capacity-voltage points. At this stage, each capacity-voltage point is treated as a multidimensional vector, containing the voltage and cumulative capacity values at a specific moment. The embedding encoding process essentially transforms these multidimensional vectors, mapping them into a low-dimensional space defined by the encoding matrix. Specifically, for each capacity-voltage point, the corresponding row vector (if matrix multiplication is used) or weight coefficient (if weighted summation is used) is found and applied based on the corresponding voltage and capacity values to generate a new low-dimensional vector. This new vector, the capacity-voltage embedding encoding vector, not only retains the key information of the original data but also enhances the comparability and correlation between different data points.

[0048] It's worth noting that embedding coding isn't a simple one-to-one mapping; instead, it attempts to reveal the true relationships hidden within high-dimensional data. For example, two seemingly different capacity-voltage points may be very close in the low-dimensional embedding space, indicating that they actually represent similar electrochemical states. Conversely, some seemingly similar data points may be far apart in the embedding space, suggesting important differences between them. Therefore, through embedding coding, researchers can gain a more intuitive and insightful understanding of the complex dynamic behavior within batteries.

[0049] In step S42, multi-scale semantic implicit coding is performed on the dense sequence of the capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features. Specifically, in an embodiment of the present application, multi-scale semantic implicit coding is performed on the dense sequence of the capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features, including: inputting the dense sequence of the capacity-voltage embedded coding vectors into a sequence encoder based on the LSTM-RNN hybrid model to obtain a dense sequence of the first capacity-voltage semantic implicit coding vector and a dense sequence of the second capacity-voltage semantic implicit coding vector; cascading the first capacity-voltage semantic implicit coding vector and the second capacity-voltage semantic implicit coding vector corresponding to each group in the dense sequence of the first capacity-voltage semantic implicit coding vector and the dense sequence of the second capacity-voltage semantic implicit coding vector to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding vectors as the dense sequence of the capacity-voltage semantic multi-scale implicit coding features.

[0050] Accordingly, considering that the dense sequence of capacity-voltage embedded code vectors contains characteristic information at different time scales during the charging test, such as short-term capacity-voltage fluctuations and periodic trends and patterns throughout the entire charging cycle, RNNs can process sequential data but are prone to vanishing or exploding gradients when processing long sequences, making it difficult to capture long-term dependencies. However, they have better perception and capture capabilities for data with shorter time intervals. LSTMs, by introducing a special gating mechanism (input gate, forget gate, and output gate), can selectively remember or forget information, effectively addressing the gradient problem of traditional RNNs and making them ideally suited for processing time series data with long-time interval dependencies. Based on this, in the technical solution of this application, the dense sequence of capacity-voltage embedded code vectors is input into a sequence encoder based on an LSTM-RNN hybrid model. This allows the RNN to focus on capturing local features and short-term dynamics, helping to identify rapid changes or abnormal events occurring within a short period of time. Simultaneously, the LSTM is used to capture long-term dependencies in the sequence, ensuring that the model can understand trends and changes throughout the entire charging and discharging cycle. This results in a dense sequence of first capacity-voltage semantic implicit code vectors and a dense sequence of second capacity-voltage semantic implicit code vectors.

[0051] It should be understood that, considering that the first capacity-voltage semantic implicit coding feature represents a local feature, such as a rapid change during the charging process, a transient phenomenon, or an event occurring in a short period of time. The second capacity-voltage semantic implicit coding feature represents a global feature, such as a trend throughout the entire charge and discharge cycle, a long-term change pattern, or a behavior in a stable state. Therefore, in order to integrate feature information at different levels or angles, thereby obtaining a richer, more comprehensive, and more representative data representation, in the technical solution of the present application, each corresponding first capacity-voltage semantic implicit coding vector and second capacity-voltage semantic implicit coding vector in the dense sequence of the first capacity-voltage semantic implicit coding vector and the dense sequence of the second capacity-voltage semantic implicit coding vector are cascaded to integrate feature information at different levels, thereby obtaining a dense sequence of capacity-voltage semantic multi-scale implicit coding vectors.

[0052] In step S43, the dense sequence of the capacity-voltage semantic multi-scale implicit coding features is subjected to capacity-voltage semantic dynamic feature significant aggregation to obtain capacity-voltage semantic significant aggregation features. Accordingly, considering that the internal electrochemical reactions at different stages (such as the initial, middle and final stages of charging) will have different characteristic manifestations, that is, each capacity-voltage semantic multi-scale implicit coding feature contains internal complex behavior patterns and key significant feature information, and these features have dynamic behaviors that continue to change as charging progresses, which is very important for identifying specific events (such as the occurrence of lithium plating) because these events are often accompanied by specific dynamic features. Therefore, in order to capture the dynamic change patterns in the data and generate feature representations with stronger expressive power and significance, in the technical solution of the present application, a capacity-voltage semantic dynamic feature significant aggregation mechanism is used to process the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregation features.

[0053] Figure 3 Flowchart of step S43 in the automatic testing method for lithium-ion batteries according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3 As shown, in step S43, the dense sequence of the capacity-voltage semantic multi-scale implicit coding features is subjected to capacity-voltage semantic dynamic feature significant aggregation to obtain capacity-voltage semantic significant aggregation features, including: S431, calculating the aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; S432, based on the aggregation center representation, the dense sequence of the capacity-voltage semantic multi-scale implicit coding features is subjected to dynamic aggregation to obtain the capacity-voltage semantic significant aggregation features.

[0054] Figure 4 FIG. 4 is a flow chart of step S431 in the automatic testing method for lithium-ion batteries according to an embodiment of the present application. More specifically, in the embodiment of the present application, Figure 4 As shown, in step S431, the aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features is calculated, including: S4311, calculating the static potential weight factor of each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential weight factors; S4312, using the dense sequence of the capacity-voltage semantic multi-scale feature static potential weight factors as a weighted sequence, calculating the position-weighted sum of the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector to obtain a capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector as the aggregation center representation.

[0055] Specifically, in the embodiment of the present application, in step S4311, the static potential weight factor of each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector is calculated to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential weight factors, including:

[0056] Each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector is subjected to characteristic static potential energy measurement to obtain a dense sequence of capacity-voltage semantic multi-scale characteristic static potential energy measurement coefficients. This process can be expressed as follows: ;in, is a dense sequence of the capacity-voltage semantic multi-scale implicit encoding vectors, and are the first, second, and third in the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector. and capacity-voltage semantic multi-scale implicit encoding vector, It is The eigenvalues of each position in the capacity-voltage semantic multi-scale implicit encoding vector, It is The mean of the capacity-voltage semantic multi-scale implicit encoding vectors, It is The standard deviation of the capacity-voltage semantic multi-scale implicit encoding vector, It is The number of eigenvalues in the capacity-voltage semantic multi-scale implicit encoding vector, is the first in the dense sequence of static potential energy measurement coefficients of capacity-voltage semantic multi-scale features. The static potential energy metric coefficient of the capacity-voltage semantic multi-scale feature;

[0057] The dense sequence of the capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is subjected to feature energy level gating screening to obtain a dense sequence of the capacity-voltage semantic multi-scale feature static potential energy weight factors.

[0058] It should be understandable that in order to quantify the feature stability and its potential energy state at each time point, so as to reveal the intrinsic connection between the features and their correlation with the target results, each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector is subjected to feature static potential energy measurement to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential energy measurement coefficients. That is, this measurement helps to more deeply understand the electrochemical behavior inside the battery and implies possible energy changes in the future, which is of great significance for predicting the occurrence of events such as lithium plating.

[0059] Accordingly, in order to further optimize and refine the feature representation and highlight the most representative and influential features under specific conditions, the dense sequence of static potential energy measurement coefficients of the capacity-voltage semantic multi-scale features is subjected to feature energy level gating screening to screen out those features with higher static potential energy measurement coefficients by setting a specific threshold, thereby obtaining a dense sequence of static potential energy weight factors of the capacity-voltage semantic multi-scale features.

[0060] More specifically, in an embodiment of the present application, the dense sequence of capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is subjected to feature energy level gating screening to obtain the dense sequence of capacity-voltage semantic multi-scale feature static potential energy weight factors, including:

[0061] The sigmoid function is used to standardize each capacity-voltage semantic multi-scale feature static potential energy measurement coefficient in the dense sequence of the capacity-voltage semantic multi-scale feature static potential energy measurement coefficient to obtain a dense sequence of standardized capacity-voltage semantic multi-scale feature static potential energy measurement coefficient. This process can be expressed as follows: ;in, is the first in the dense sequence of static potential energy measurement coefficients of capacity-voltage semantic multi-scale features. The static potential energy measurement coefficient of the capacity-voltage semantic multi-scale feature, yes function, is the first in the dense sequence of the static potential energy metric coefficients of the standardized capacity-voltage semantic multi-scale feature A standardized capacity-voltage semantic multi-scale feature static potential energy measurement coefficient;

[0062] The dense sequence of the normalized capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is input into a gated screening module based on a mask function to obtain a dense sequence of the capacity-voltage semantic multi-scale feature static potential energy weight factors. This process can be expressed as follows: ;in, is the first in the dense sequence of the static potential energy metric coefficients of the standardized capacity-voltage semantic multi-scale feature Normalized capacity-voltage semantic multi-scale feature static potential energy measurement coefficient, is the preset threshold, is a masking operation, is the first in the dense sequence of static potential weight factors of capacity-voltage semantic multi-scale features. The static potential energy weight factor of the capacity-voltage semantic multi-scale feature.

[0063] Specifically, in step S4312, the dense sequence of the capacity-voltage semantic multi-scale feature static potential weight factors is used as a weight sequence, and the position-weighted sum of the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector is calculated to obtain the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector as the aggregation center representation. This process can be expressed by the formula: ;in, is the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector capacity-voltage semantic multi-scale implicit encoding vector, is the first in the dense sequence of static potential weight factors of capacity-voltage semantic multi-scale features. The static potential energy weight factor of the capacity-voltage semantic multi-scale feature, is the number of vectors in the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector, It is the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector.

[0064] It should be understandable that the dense sequence of capacity-voltage semantic multi-scale feature static potential weight factors is used as weights to calculate the position-weighted sum of the dense sequence of capacity-voltage semantic multi-scale implicit encoding vectors to note how each feature vector is affected by the center point, and based on this, the evolution trajectory of the feature vector in the feature space is inferred to obtain a more representative and comprehensive feature representation of the capacity-voltage semantic multi-scale pseudo-anchored aggregation center representation vector.

[0065] Specifically, in step S432, based on the aggregation center representation, dynamic aggregation is performed on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain the capacity-voltage semantic significant aggregation features, including:

[0066] Calculate the aggregate movement direction of each capacity-voltage semantic multi-scale implicit encoding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector relative to the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector to obtain a dense sequence of capacity-voltage semantic multi-scale aggregate running directions. This process can be expressed as follows: ;in, is the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector capacity-voltage semantic multi-scale implicit encoding vector, is the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector, is the length of the vector, is the inverse cosine function, is the first in the dense sequence of capacity-voltage semantic multi-scale aggregation running direction Capacity-voltage semantic multi-scale aggregation operation direction;

[0067] Based on the dense sequence of the capacity-voltage semantic multi-scale aggregation running direction, the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector is dynamically aggregated toward the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector to obtain the capacity-voltage semantic salient aggregation encoding vector as the capacity-voltage semantic salient aggregation feature. This process can be expressed as follows: ;in, is the first in the dense sequence of capacity-voltage semantic multi-scale aggregation running direction Capacity-voltage semantic multi-scale aggregation operation direction, is the first in the dense sequence of static potential energy measurement coefficients of capacity-voltage semantic multi-scale features. The static potential energy measurement coefficient of the capacity-voltage semantic multi-scale feature, is the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector capacity-voltage semantic multi-scale implicit encoding vector, is the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector, is the number of vectors in the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector, is the sine function, is the first in the sequence of weight matrices A weight matrix, is the first in the sequence of bias vectors A bias vector, is the capacity-voltage semantically significant aggregated encoding vector.

[0068] Accordingly, the aggregation movement direction of each capacity-voltage semantic multi-scale implicit encoding vector relative to the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector is calculated to obtain a dense sequence of capacity-voltage semantic multi-scale aggregation running directions. The goal of this process is to measure the dynamic change trend of each capacity-voltage semantic multi-scale implicit encoding vector relative to the aggregation center representation. These trends reveal whether the feature vector is moving towards or away from the center point, mapping the mutual influence and change pattern between vectors in the feature space, which is crucial for understanding the relative position and overall dynamic behavior of feature vectors. In other words, the aggregation trend towards the center helps to identify feature changes that have a significant impact on specific events. For example, the occurrence of lithium plating is often accompanied by specific directional changes.

[0069] Finally, based on the running direction of each capacity-voltage semantic multi-scale aggregation, the dense sequence of capacity-voltage semantic multi-scale implicit coding vectors is dynamically aggregated toward the central representation vector to obtain the capacity-voltage semantic salient aggregation coding vector. In other words, based on the previously determined aggregation movement direction and the static potential energy measurement coefficient of the capacity-voltage semantic multi-scale implicit coding vector, the imitation dynamic model is applied to adjust and optimize these feature vectors to make them closer to the aggregation center vector. In this way, the dynamic characteristics of capacity-voltage can be more effectively captured, and the spatial layout of the feature vectors can be dynamically adjusted and optimized to create a more compact, representative and recognizable feature representation, which helps to improve the model's subsequent judgment accuracy of the critical point of lithium deposition.

[0070] In step S44, the critical point of lithium deposition at the beginning of lithium deposition of the tested lithium-ion battery is obtained based on the capacity-voltage semantically significant aggregation features. Specifically, in an embodiment of the present application, the critical point of lithium deposition at the beginning of lithium deposition of the tested lithium-ion battery is obtained based on the capacity-voltage semantically significant aggregation features, including: inputting the capacity-voltage semantically significant aggregation coding vector into a lithium deposition critical point identifier based on a decoder to obtain the critical point of lithium deposition at the beginning of lithium deposition of the tested lithium-ion battery. That is, the capacity-voltage semantically significant aggregation features obtained by semantic dynamic aggregation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding vectors are decoded and processed to intelligently identify the critical point of lithium deposition at the beginning of lithium deposition of the tested lithium-ion battery. In this way, full automation of data processing and critical point identification is achieved without manual intervention, ensuring the consistency and objectivity of the results. It can also capture complex electrochemical dynamic behavior changes, rather than relying solely on preset thresholds or simple mathematical models, making the model more flexible and able to accurately reflect the true state inside the battery.

[0071] In particular, in a specific embodiment of the present application, the specific processing process of inputting the capacity-voltage semantically significant aggregated coding vector into a decoder-based lithium deposition critical point identifier to obtain the lithium deposition critical point at which lithium deposition begins to occur in the tested lithium-ion battery is as follows:

[0072] First, choose a suitable decoder architecture, such as recurrent neural networks (RNNs) such as LSTM and GRU, which are suitable for processing time series data and can capture long-term dependencies; convolutional neural networks (CNNs), which are very effective for local feature extraction, especially when spatial structure needs to be considered; and transformers, which use attention mechanisms to capture global and local features at the same time and are particularly suitable for processing long sequence data.

[0073] Then, the model is trained using a large amount of historical data, including data samples with known lithium plating critical points. During training, the model learns how to identify characteristic patterns associated with lithium plating from the capacity-voltage semantically significant aggregate encoding vector. An appropriate loss function, such as cross-entropy loss or mean squared error, is defined to measure the difference between the predicted value and the true label and guide the optimization of the model parameters.

[0074] Next, the decoder model identifies characteristic patterns highly correlated with lithium deposition based on the input capacity-voltage semantically significant aggregated encoding vector. These patterns typically manifest as unusual changes or trends at specific time points. By analyzing the critical points for lithium deposition in the training data, a reasonable threshold or rule is set to determine when the critical state for lithium deposition has been reached. For example, when a certain feature value exceeds a preset threshold, that time point is considered to be the critical point for lithium deposition. By combining feature information at different time scales, the accuracy and reliability of the recognition results are ensured. For example, both short-term rapid changes and long-term trends are considered.

[0075] Subsequently, the decoder model outputs the probability of whether the lithium deposition critical point is reached at each time point or directly gives the specific critical point location.

[0076] For example, consider a trained LSTM-based decoder model for identifying lithium plating critical points. The identification process is as follows: A capacity-voltage semantically significant aggregated encoding vector is input into the decoder model. Based on the input encoding vector, the model calculates the probability of lithium plating at each time point, or directly outputs the critical point location. A threshold is set; when the probability of lithium plating at a given time point exceeds this threshold, the time point is considered a lithium plating critical point. For example, if the threshold is set to 0.5, a critical point is identified when the probability of lithium plating exceeds 50%. Finally, the decoder model outputs the lithium plating critical point for the tested lithium-ion battery.

[0077] In one example, considering that the dense sequence of the first capacity-voltage semantic implicit coding vector and the dense sequence of the second capacity-voltage semantic implicit coding vector respectively represent the implicit coding features of the capacity-voltage point dense sequence, when performing corresponding sequence cascade and feature aggregation based on feature dynamics analysis, due to the difference in the fairness of the dynamic analysis level caused by the implicit pattern differences and sequence distribution differences of the feature population attributes, the capacity-voltage semantic significant aggregation coding vector will have a lack of feature distribution aggregation inclusiveness, thereby affecting the accuracy of the lithium plating critical point at which lithium plating begins to occur in the tested lithium-ion battery obtained by inputting the decoder-based lithium plating critical point identifier.

[0078] Therefore, in this example, when the capacity-voltage semantically significant aggregated coding vector is input into the decoder-based lithium deposition critical point identifier, the capacity-voltage semantically significant aggregated coding vector is optimized. The specific optimization process is:

[0079] Determining an eigenvalue mean and an eigenvalue standard deviation corresponding to the capacity-voltage semantically significant aggregated encoding vector;

[0080] Multiplying the point-wise subtraction vector of the capacity-voltage semantically significant aggregated coding vector and the eigenvalue mean by the eigenvalue standard deviation to obtain a first capacity-voltage semantically significant aggregated coding intermediate vector, and multiplying the point-wise subtraction vector of the capacity-voltage semantically significant aggregated coding vector and the eigenvalue standard deviation by the eigenvalue mean to obtain a second capacity-voltage semantically significant aggregated coding intermediate vector;

[0081] After performing a dot multiplication of the bit-by-bit inverse of the second capacity-voltage semantically significant aggregate coding intermediate vector and the first capacity-voltage semantically significant aggregate coding intermediate vector, a bit-by-bit logarithm with base 2 is calculated to obtain a capacity-voltage semantically significant aggregate coding correction vector;

[0082] The square root of the quotient of the eigenvalue mean divided by the eigenvalue standard deviation is multiplied by the weight hyperparameter, and then added to the capacity-voltage semantically significant aggregation coding correction vector to obtain an optimized capacity-voltage semantically significant aggregation coding vector.

[0083] That is, taking into account the attribute level fairness differences of the data population corresponding to the fusion features of the capacity-voltage semantically significant aggregation coding vector, in order to improve the aggregation inclusiveness of the capacity-voltage semantically significant aggregation coding vector under the feature distribution diversity, the cross probability value constraint based on the capacity-voltage semantically significant aggregation coding vector is used as the interactive fairness target representation to correct the group feature information interactive propagation of the capacity-voltage semantically significant aggregation coding vector, and the unified statistical feature response interaction based on the capacity-voltage semantically significant aggregation coding vector is used as the feature distribution multi-level fairness target bias to achieve a robust distribution fairness unified representation of the capacity-voltage semantically significant aggregation coding vector, forming a fair collaboration paradigm under the feature distribution framework of the capacity-voltage semantically significant aggregation coding vector, and improving the accuracy of the lithium plating critical point at which lithium plating begins to occur in the tested lithium-ion battery obtained by its input decoder-based lithium plating critical point identifier.

[0084] In summary, step S4 is explained, which uses data analysis and encoding technology based on deep learning to embed the dense sequence of each capacity-voltage point, and then performs multi-scale implicit encoding on each capacity-voltage embedded feature, so as to intelligently identify the critical point of lithium plating of the tested lithium-ion battery where lithium plating begins to occur based on the dynamic pressure semantic significant aggregation representation between the multi-scale implicit coding features of each capacity-voltage semantic. In this way, full automation of data processing and critical point identification is achieved without manual intervention, ensuring the consistency and objectivity of the results. It can also capture complex changes in electrochemical dynamic behavior, rather than relying solely on preset thresholds or simple mathematical models, making the model more flexible and able to accurately reflect the true state inside the battery.

[0085] In summary, the automatic testing method for lithium-ion batteries based on the embodiments of the present application is described. The method collects the voltage and capacity data of the negative electrode of the lithium-ion battery during the charging test, then performs differential processing on this data to generate a capacity-voltage differential curve. The differential curve is then densely sampled to obtain a series of detailed capacity-voltage data points. Finally, these dense data points are used to determine whether lithium plating occurs inside the battery. In this way, the internal state of the lithium-ion battery during the charging process, especially the lithium plating phenomenon, can be accurately monitored and evaluated, which is crucial for battery performance evaluation, safety testing, and life prediction.

[0086] Figure 5 FIG. 1 is a block diagram of a lithium-ion battery automatic testing system according to an embodiment of the present application. Figure 5As shown, the lithium-ion battery automatic testing system 100 according to the embodiment of the present application includes: a charging data acquisition module 110, which is used to collect the voltage and capacity data of the negative electrode of the tested lithium-ion battery during the charging test; a voltage-capacity differential processing module 120, which is used to perform differential processing on the collected voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; a differential curve dense sampling module 130, which is used to densely sample the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; a battery internal lithium deposition state judgment module 140, which is used to judge the battery internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points; wherein the battery internal lithium deposition state judgment module includes: a capacity-voltage embedded coding unit, which is used to embed and code each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features; a capacity-voltage semantic multi-scale coding unit, which is used to Multi-scale semantic implicit coding is performed on the dense sequence of the capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features; a capacity-voltage semantic aggregation unit is used to perform capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregation features, and the capacity-voltage semantic aggregation unit includes: a capacity-voltage semantic feature aggregation center representation subunit, which is used to calculate the aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; a capacity-voltage semantic dynamic aggregation subunit, which is used to perform dynamic aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features based on the aggregation center representation to obtain the capacity-voltage semantic significant aggregation features; a lithium plating critical point judgment unit is used to obtain the lithium plating critical point at which lithium plating begins to occur in the tested lithium-ion battery according to the capacity-voltage semantic significant aggregation.

[0087] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned lithium-ion battery automatic testing system have been described in detail above. Figures 1 to 4 The above description has been described in detail in the description of the automatic testing method for lithium-ion batteries, and therefore, its repeated description will be omitted.

[0088] As described above, the lithium-ion battery automatic testing system 100 according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server equipped with a lithium-ion battery automatic testing algorithm. In one possible implementation, the lithium-ion battery automatic testing system 100 according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or hardware module. For example, the lithium-ion battery automatic testing system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the lithium-ion battery automatic testing system 100 can also be one of the many hardware modules of the wireless terminal.

[0089] Alternatively, in another example, the lithium-ion battery automatic testing system 100 and the wireless terminal may be separate devices, and the lithium-ion battery automatic testing system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in a predetermined data format.

[0090] The foregoing is merely an example of the principles of the present disclosure, and various modifications may be made by those skilled in the art without departing from the scope of the present disclosure. The above embodiments are presented for the purpose of illustration and not limitation. The present disclosure may also take many forms other than those explicitly described herein.

Claims

1. A lithium-ion battery automatic testing method, characterized in that: include: S1: Collect the voltage and capacity data of the negative electrode of the tested lithium-ion battery during the charging test; S2: performing differential processing on the collected voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; S3: densely sampling the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; S4: judging the internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points; Wherein, said S4 includes: Embedding and encoding each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features; Performing multi-scale semantic implicit coding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features; Performing capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregate features, including: calculating an aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; performing dynamic aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features based on the aggregation center representation to obtain the capacity-voltage semantic significant aggregate features; According to the capacity-voltage semantically significant aggregation feature, a critical point at which lithium deposition begins to occur in the tested lithium-ion battery is obtained.

2. The automatic testing method for lithium-ion batteries according to claim 1, characterized in that: Embedding and encoding each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded features, including: using a capacity-voltage embedding coding matrix to embed and encode each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedded coding vectors as the dense sequence of capacity-voltage embedded features.

3. The automatic testing method for lithium-ion batteries according to claim 2, characterized in that: Performing multi-scale semantic implicit coding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding features, including: Inputting the dense sequence of capacity-voltage embedded coding vectors into a sequence encoder based on an LSTM-RNN hybrid model to obtain a dense sequence of first capacity-voltage semantic implicit coding vectors and a dense sequence of second capacity-voltage semantic implicit coding vectors; Each corresponding group of the first capacity-voltage semantic implicit coding vector and the second capacity-voltage semantic implicit coding vector in the dense sequence of the first capacity-voltage semantic implicit coding vector and the dense sequence of the second capacity-voltage semantic implicit coding vector is cascaded to obtain a dense sequence of capacity-voltage semantic multi-scale implicit coding vectors as the dense sequence of the capacity-voltage semantic multi-scale implicit coding features.

4. The automatic testing method for lithium-ion batteries according to claim 3, characterized in that: Calculating the aggregate center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit encoding features, including: Calculating the static potential energy weight factor of each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential energy weight factors; Taking the dense sequence of the static potential energy weight factors of the capacity-voltage semantic multi-scale features as a weighted sequence, the position-weighted sum of the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vectors is calculated to obtain a capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector as the aggregation center representation.

5. The automatic testing method for lithium-ion batteries according to claim 4, characterized in that: Calculating the static potential energy weight factor of each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential energy weight factors, including: Performing feature static potential energy measurement on each capacity-voltage semantic multi-scale implicit coding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit coding vector to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential energy measurement coefficients; The dense sequence of the capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is subjected to feature energy level gating screening to obtain a dense sequence of the capacity-voltage semantic multi-scale feature static potential energy weight factors.

6. The automatic testing method for lithium-ion batteries according to claim 5, characterized in that: Performing feature energy level gating screening on the dense sequence of capacity-voltage semantic multi-scale feature static potential energy measurement coefficients to obtain a dense sequence of capacity-voltage semantic multi-scale feature static potential energy weight factors, including: Using a sigmoid function, each capacity-voltage semantic multi-scale feature static potential energy measurement coefficient in the dense sequence of capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is standardized to obtain a dense sequence of standardized capacity-voltage semantic multi-scale feature static potential energy measurement coefficients; The dense sequence of the standardized capacity-voltage semantic multi-scale feature static potential energy measurement coefficients is input into a gated screening module based on a mask function to obtain a dense sequence of the capacity-voltage semantic multi-scale feature static potential energy weight factors.

7. The lithium-ion battery automatic testing method according to claim 6, characterized in that: Based on the aggregation center representation, dynamic aggregation is performed on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain the capacity-voltage semantic significant aggregation features, including: Calculating the aggregate movement direction of each capacity-voltage semantic multi-scale implicit encoding vector in the dense sequence of the capacity-voltage semantic multi-scale implicit encoding vector relative to the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector to obtain a dense sequence of capacity-voltage semantic multi-scale aggregate running directions; Based on the dense sequence of the capacity-voltage semantic multi-scale aggregation running direction, the dense sequence of the capacity-voltage semantic multi-scale implicit coding vectors is dynamically aggregated toward the capacity-voltage semantic multi-scale pseudo-anchor aggregation center representation vector to obtain a capacity-voltage semantic salient aggregation coding vector as the capacity-voltage semantic salient aggregation feature.

8. The automatic testing method for lithium-ion batteries according to claim 7, characterized in that: According to the capacity-voltage semantically significant aggregation feature, a lithium deposition critical point at which lithium deposition begins to occur in the tested lithium-ion battery is obtained, including: inputting the capacity-voltage semantically significant aggregation coding vector into a decoder-based lithium deposition critical point identifier to obtain the lithium deposition critical point at which lithium deposition begins to occur in the tested lithium-ion battery.

9. A lithium-ion battery automatic testing system, characterized in that: include: A charging data acquisition module is used to collect the voltage and capacity data of the negative electrode of the tested lithium-ion battery during the charging test; a voltage-capacity differential processing module is used to perform differential processing on the collected voltage and capacity data of the negative electrode to obtain a capacity-voltage differential curve; a differential curve dense sampling module is used to perform dense sampling on the capacity-voltage differential curve to obtain a dense sequence of capacity-voltage points; a battery internal lithium deposition state judgment module is used to judge the battery internal lithium deposition state of the tested lithium-ion battery based on the dense sequence of capacity-voltage points; wherein the battery internal lithium deposition state judgment module includes: a capacity-voltage embedding coding unit, configured to embed and code each capacity-voltage point in the dense sequence of capacity-voltage points to obtain a dense sequence of capacity-voltage embedding features; A capacity-voltage semantic multi-scale encoding unit, configured to perform multi-scale semantic implicit encoding on the dense sequence of capacity-voltage embedded features to obtain a dense sequence of capacity-voltage semantic multi-scale implicit encoding features; A capacity-voltage semantic aggregation unit is configured to perform capacity-voltage semantic dynamic feature significant aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features to obtain capacity-voltage semantic significant aggregate features. The capacity-voltage semantic aggregation unit includes: a capacity-voltage semantic feature aggregation center representation subunit, configured to calculate the aggregation center representation of the dense sequence of the capacity-voltage semantic multi-scale implicit coding features; and a capacity-voltage semantic dynamic aggregation subunit, configured to perform dynamic aggregation on the dense sequence of the capacity-voltage semantic multi-scale implicit coding features based on the aggregation center representation to obtain the capacity-voltage semantic significant aggregate features. The lithium deposition critical point judgment unit is used to obtain the lithium deposition critical point at which lithium deposition begins to occur in the tested lithium-ion battery according to the capacity-voltage semantic significant aggregation.

Images