A method, device, medium and product for obtaining an incremental capacity curve of a lithium ion battery

Abstract

The application discloses a lithium ion battery incremental capacity curve acquisition method, equipment, medium and product, relates to the technical field of lithium ion battery state evaluation and nondestructive diagnosis, and the method comprises the steps of constructing an experimental test matrix under a multi-stage constant-current-constant-voltage fast charging condition; based on a charging and discharging equipment, using the experimental test matrix, acquiring time sequence data corresponding to different experimental conditions of a lithium ion battery and an incremental capacity curve corresponding to the time sequence data; extracting characteristic parameters of the incremental capacity curve based on a semi-mechanism semi-experience model; determining an incremental capacity data set according to the time sequence data, the incremental capacity curve and the characteristic parameters; determining a reconstruction model according to the incremental capacity data set and a joint loss function in combination with a deep learning model; and inputting time sequence data of a lithium ion battery to be tested into the reconstruction model to determine an incremental capacity curve of the lithium ion battery to be tested. The application can quickly and highly faithfully reconstruct the corresponding incremental capacity curve of the lithium ion battery.

Description

Technical Field

[0001] This application relates to the field of lithium-ion battery state assessment and non-destructive diagnostic technology, and in particular to a method, device, medium and product for obtaining incremental capacity curves of lithium-ion batteries. Background Technology

[0002] Incremental Capacity Analysis (ICA), as a highly efficient non-destructive diagnostic technique, transforms the voltage plateau region during the charging and discharging process of lithium-ion batteries into characteristic incremental capacity peaks. This allows for a direct mapping of internal degradation mechanisms such as loss of lithium-ion inventory (LLI) and loss of active material (LAM). In recent years, this technology has played an irreplaceable role in academic and engineering fields such as battery aging mechanism analysis and online health status assessment.

[0003] However, the traditional direct measurement method for incremental capacity curves has significant engineering limitations. This method typically requires the battery to undergo a long period of full-cycle charge-discharge at an extremely low constant current rate (such as 1 / 20C or lower) to obtain stable data approximating the open circuit voltage (OCV) state. But in real-world applications, batteries are often in multi-stage constant current-constant voltage (CCCV) fast charging conditions, making it extremely difficult to meet quasi-static, long-term full-cycle testing conditions. This results in excessively high time costs for traditional direct measurement methods, making them completely unsuitable for online real-time diagnostics in actual vehicles or energy storage systems.

[0004] To overcome the bottleneck of long-duration, low-rate testing, many related technologies attempt to indirectly extract incremental capacity curves based on partial charging segments. However, most of these methods still strictly rely on constant current or constant power conditions. For real-world scenarios with multi-stage CCCV fast charging and dynamic discharging, which exhibit strong nonlinearity and spatiotemporal coupling characteristics, conventional algorithms struggle to eliminate high-frequency noise and polarization interference. Furthermore, existing pure data-driven networks lack the ability to deeply mine local multi-view features and contextual dependencies in long-sequence dynamic data, and are further constrained by underlying physical laws. This results in significant peak distortion in incremental capacity curve prediction under dynamic conditions, leading to poor generalization ability and robustness. Summary of the Invention

[0005] The purpose of this application is to provide a method for obtaining the incremental capacity curve of a lithium-ion battery, which helps to solve the problem that the incremental capacity curve of a lithium-ion battery cannot be quickly and faithfully reconstructed in the prior art.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for obtaining incremental capacity curves of lithium-ion batteries, comprising: constructing an experimental test matrix under multi-stage constant current-constant voltage fast charging conditions; performing charge-discharge cycle tests on lithium-ion batteries using the experimental test matrix based on a charge-discharge device to obtain time-series data of the lithium-ion batteries and incremental capacity curves corresponding to the time-series data; the time-series data including current-capacity sequences and voltage-capacity sequences; extracting characteristic parameters of the incremental capacity curves based on a semi-mechanistic semi-empirical model; wherein the characteristic parameters include the peak height, peak position, and half-peak width of each peak in the incremental capacity curves; determining an incremental capacity dataset based on the time-series data, the incremental capacity curves, and the characteristic parameters; determining a reconstruction model based on the incremental capacity datasets and a joint loss function, combined with a deep learning model; the deep learning model being determined based on the fusion of a residual convolutional neural network and a Transformer model; and inputting the time-series data of the lithium-ion battery under test into the reconstruction model to determine the incremental capacity curve of the lithium-ion battery under test.

[0007] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for obtaining the incremental capacity curve of a lithium-ion battery.

[0008] Thirdly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for obtaining the incremental capacity curve of a lithium-ion battery.

[0009] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for obtaining the incremental capacity curve of a lithium-ion battery.

[0010] Based on the specific embodiments provided in this application, the following technical effects are disclosed.

[0011] This application first constructs an experimental test matrix for multi-stage constant current-constant voltage fast charging. Then, based on charging and discharging equipment, it uses the experimental test matrix to conduct charge-discharge cycle tests on lithium-ion batteries, obtaining time-series data of the lithium-ion batteries and the corresponding incremental capacity curves. This effectively overcomes the problems of traditional analytical acquisition of incremental capacity curves, which heavily relies on extremely small constant current and extremely long test times, and can directly adapt to the multi-stage constant current-constant voltage fast charging and dynamic discharge conditions commonly encountered in complex practical applications. Furthermore, based on a semi-mechanistic semi-empirical model, the characteristic parameters of the incremental capacity curves are extracted and the incremental capacity dataset is determined. Then, a deep learning model is combined to determine the reconstruction model. The deep learning model is determined by fusing residual convolutional neural networks and Transformer models, solving the problems of existing pure data-driven networks lacking the ability to deeply mine local multi-view features and contextual dependencies in long-sequence dynamic data, and lacking constraints from underlying physical laws. This leads to peak distortion, poor generalization ability, and poor robustness in the prediction of incremental capacity curves under dynamic conditions. This application innovatively integrates the characteristic parameters (peak height, peak position, and full width at half maximum) derived from a semi-mechanistic, semi-empirical model into a joint loss function to construct a reconstruction model. This effectively suppresses local extrema or non-physical distortions (such as peak shape distortion) that are prone to occur in pure data-driven algorithms under complex dynamic conditions. This not only accelerates the convergence speed of the reconstruction model but also improves the physical consistency and interpretability of the incremental capacity curve of the lithium-ion battery predicted by the reconstruction model. Ultimately, it rapidly and faithfully reconstructs the incremental capacity curve of the lithium-ion battery. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a method for obtaining incremental capacity curves of lithium-ion batteries provided in this application.

[0014] Figure 2 This is a schematic diagram of a multi-stage constant current-constant voltage fast charging process.

[0015] Figure 3 The figure shows the fitting results of the semi-mechanistic, semi-empirical model for the incremental capacity curve.

[0016] Figure 4 This is a graph showing the evolution of the peak voltage value and incremental capacity curves as the battery ages; where, Figure 4 (a) in the figure is the curve of the peak position shift of each peak as the battery ages; Figure 4 (b) in the figure is the evolution curve of the incremental capacity curve as the battery ages.

[0017] Figure 5 This is a graph showing the predicted results of the battery incremental capacity curve. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] To make the objectives, features, and advantages of this application more apparent and understandable, the following detailed description of this application is provided in conjunction with the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown, this application provides a method for obtaining the incremental capacity curve of a lithium-ion battery, which includes steps 101-106.

[0021] Step 101: Construct an experimental test matrix under multi-stage constant current-constant voltage fast charging conditions.

[0022] Specifically, step 101 includes: constructing an experimental test matrix for multi-stage constant current-constant voltage fast charging based on ambient temperature, the number of constant current-constant voltage charging stages, the charging current rate of stage I, and the switching points of the battery state of charge corresponding to multi-stage constant current-constant voltage.

[0023] Step 102: Based on the charging and discharging equipment, use the experimental test matrix to perform charge-discharge cycle tests on the lithium-ion battery, and obtain the time-series data of the lithium-ion battery and the incremental capacity curve corresponding to the time-series data; the time-series data includes current-capacity sequence and voltage-capacity sequence.

[0024] Step 102 includes: based on the charging and discharging equipment, traversing all experimental conditions in the experimental test matrix, and collecting time-series data corresponding to different experimental conditions of the lithium-ion battery; according to the preset interval test number and capacity calibration experiment, performing low-rate charge and discharge tests on the lithium-ion battery, and determining the current maximum usable capacity of the battery and the incremental capacity curve corresponding to the time-series data.

[0025] Step 103: Based on a semi-mechanistic, semi-empirical model, extract the characteristic parameters of the incremental capacity curve; wherein, the characteristic parameters include the peak height, peak position, and half-peak width of each peak in the incremental capacity curve.

[0026] Specifically, step 103 includes: fitting the semi-mechanistic semi-empirical model using the least squares method to extract the characteristic parameters of the incremental capacity curve.

[0027] Step 104: Determine the incremental capacity dataset based on the time series data, the incremental capacity curve, and the characteristic parameters.

[0028] Step 105: Based on the incremental capacity dataset and joint loss function, and combined with the deep learning model, determine the reconstruction model; the deep learning model is determined by fusing residual convolutional neural networks and Transformer models.

[0029] Step 105 specifically includes: inputting the incremental capacity dataset into a residual convolutional neural network to output a deep feature sequence; inputting the deep feature sequence into a Transformer model to output a predicted incremental capacity curve and predicted characteristic parameters; determining the incremental capacity loss based on the predicted incremental capacity curve and the true incremental capacity curve; determining the characteristic parameter loss based on the predicted characteristic parameters and the true characteristic parameters; fusing the incremental capacity loss and the characteristic parameter loss to determine a joint loss function; training the residual convolutional neural network and the Transformer model with the goal of minimizing the joint loss function, and determining the trained deep learning model as the reconstruction model.

[0030] The joint loss function is: .

[0031] in, ; The characteristic parameter loss weight, For the total loss, For incremental capacity curve loss, To predict the incremental capacity curve; This represents the actual incremental capacity curve; Mean squared error; α is the mean absolute error; The weights; ; For characteristic parameter loss; K This represents the total number of characteristic parameters. For predicting characteristic parameters, These are the actual characteristic parameters; For the index of the characteristic parameter; .

[0032] Step 106: Input the time series data of the lithium-ion battery under test into the reconstruction model to determine the incremental capacity curve of the lithium-ion battery under test.

[0033] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the methods described above.

[0034] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the methods described above.

[0035] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the methods described above.

[0036] In practical applications, the incremental capacity curve acquisition method for lithium-ion batteries provided in this application is based on low-rate capacity calibration data. It identifies parameters of a semi-mechanistic, semi-empirical model of the incremental capacity curve, extracting electrochemical characteristic parameters such as peak height, peak position, and full width at half maximum (FWHM). Subsequently, using voltage-capacity and voltage-capacity sequences obtained during dynamic charge-discharge processes as model inputs, and the corresponding incremental capacity curves and their characteristic parameters as labels, a training dataset is constructed, and a reconstruction model of the incremental capacity curve based on a Physics-Informed Neural Networks (PINN) architecture is trained. This incremental capacity curve reconstruction model is the aforementioned reconstruction model. Finally, by inputting the dynamic IQ and VQ sequences of the lithium-ion battery under test, i.e., time-series data including current-capacity and voltage-capacity sequences, the corresponding incremental capacity curve can be quickly and faithfully reconstructed.

[0037] The main method flow of the embodiments of this application is as follows: Figure 1 As shown, the specific steps include:

[0038] S1 uses ambient temperature, the number of constant current-constant voltage charging stages, the I-stage charging current rate, and the SOC nodes with different constant current-constant voltage charging switching as variables to construct an experimental test matrix covering various aging stresses.

[0039] Among them, constant current-constant voltage (CCCV) is the most typical and widely used standard charging strategy for secondary batteries such as lithium-ion batteries.

[0040] In this embodiment, the test matrix parameters for the multi-stage CCCV fast charging condition (such as...) Figure 2 The specific configuration is as follows (as shown): Temperature: Temperature is a key environmental factor affecting battery polarization characteristics and aging rate. To comprehensively analyze the perturbation of incremental capacity curves by temperature coupling effects, the test temperatures were set at three gradients: 5℃, 25℃, and 45℃. This range fully covers the operating temperature range of electric vehicles / energy storage systems under typical real-world scenarios.

[0041] CCCV charging stage number: The stage division directly determines the evolution trajectory of fast charging strategies and the internal electrochemical behavior of the battery. This embodiment sets up two CCCV charging modes, one stage and two stages, to compare and evaluate the differences in polarization accumulation and incremental capacity characteristics under different segmentation strategies. For the strategy including two-stage CCCV charging, the current rate of the second CCCV stage is strictly limited to 1C; in addition, the cycle aging discharge condition corresponding to all fast charging strategies is a uniform 1C constant current (CC) discharge.

[0042] Phase I charging current rate: As a core variable characterizing fast charging intensity, it determines the degree of battery polarization and local heat generation. This embodiment selects multiple discrete current rates within the 2C to 8C range to simulate battery operating conditions under different fast charging intensities, exploring the effect of current amplitude on the incremental capacity curve characteristics. Figure 2 In this context, xC represents the current multiplier for stage I, specifically 2C, 4C, 6C, and 8C.

[0043] CCCV switching SOC nodes, or State of Charge (SOC) switching points, are closely related to the internal phase transition process and lithium-ion migration kinetics of the battery. This embodiment sets four stepped SOC switching nodes at 25%, 45%, 65%, and 85% to analyze the effect of stage switching at different charging depths on the evolution of the incremental capacity curve.

[0044] The ambient temperature includes multiple temperature gradients within the range of 0-50 degrees Celsius; the constant current-constant voltage charging stages include one-stage constant current-constant voltage fast charging and two-stage constant current-constant voltage fast charging; the first-stage charging current rate includes multiple discrete current rates within the range of 2C-8C, where C is the rated capacity of the battery; the battery state of charge switching points corresponding to the multi-stage constant current-constant voltage charging include multiple battery state of charge switching points within the range of 20%-90%.

[0045] S2, using charging and discharging equipment, conducts systematic charging and discharging tests on the battery, fully covering all experimental conditions in the experimental test matrix, and simultaneously collects current, voltage, and capacity timing data under each condition. Each experimental condition is tested using a separate battery.

[0046] S3, every 50 cycles, performs a capacity calibration experiment to obtain the current maximum usable capacity of the battery and the corresponding incremental capacity curve (IC curve) label.

[0047] Further, in step S3, the capacity calibration experiment specifically involves: discharging the battery from the upper cutoff voltage to the lower cutoff voltage using a low-rate constant current discharge, and defining the cumulative capacity released during this quasi-static discharge process as the battery's current maximum usable capacity.

[0048] In step S3, the capacity calibration experiment uses a very small current (0.2A) for quasi-static constant current discharge, allowing the battery to discharge smoothly from the upper cutoff voltage to the lower cutoff voltage. This quasi-static discharge process can eliminate the influence of polarization internal resistance to the greatest extent. The cumulative capacity released is defined as the battery's current maximum usable capacity. The IC curve obtained by mathematically transforming the voltage sequence is then used as the label for the subsequent deep learning model.

[0049] S4. Establish a semi-mechanistic, semi-empirical model that can accurately describe the mapping relationship between external charging and discharging behavior and internal incremental capacity characteristics. Use the least squares algorithm to fit the model parameters and extract characteristic parameters such as peak height, peak position, and half-peak width of the incremental capacity curve.

[0050] Furthermore, in step S4, the semi-mechanistic semi-empirical model is specifically a semi-mechanistic semi-empirical model that uses the equivalent circuit model (ECM) as a framework and couples the SOC-OCV relationship.

[0051] In step S4, the semi-mechanistic semi-empirical model specifically involves establishing the voltage response relationship during the constant current charging stage based on the ECM, and combining it with the SOC-OCV mapping relationship to construct a semi-mechanistic semi-empirical model. Its main equations are as follows: .

[0052] in, This represents the cumulative capacity of a lithium-ion battery at voltage V. This represents the current maximum available capacity of lithium-ion batteries. This represents the normalized peak area of ​​the IC curve; This represents the number of peaks in the IC curve. Voltage; For the first The peak position of each peak, which is the voltage value corresponding to the peak value; W i Let be the half-width of the i-th peak of the IC curve; This is a constant term.

[0053] Specific fitting results are as follows: Figure 3 As shown, the overlapping characteristic peaks are decomposed into four peaks: peak 1, peak 2, peak 3, and peak 4.

[0054] The SOC-OCV (State of Charge-Open Circuit Voltage) curve is a characteristic curve describing the relationship between the battery's state of charge (SOC) and open circuit voltage (OCV). As an important input for battery modeling in a battery management system, it plays a core role in battery state estimation.

[0055] S5 uses the IQ and VQ sequences from the multi-stage CCCV charging process as model inputs and the incremental capacity curves and characteristic parameters obtained in steps S3 and S4 as labels to construct a dataset for training the deep learning model.

[0056] In step S5, during dataset construction, the voltage-time (Vt) and current-time (It) sequences in the time domain, together with the capacity-time (Qt) sequence, are mapped and transformed into voltage-capacity (VQ) and current-capacity (IQ) sequences. This transformation effectively eliminates time-scale distortions caused by different charging rates and unsteady dynamic operating conditions. Simultaneously, a multi-sequence alignment technique is employed to split the VQ and IQ sequences of the complete charging process into multiple segments with different SOC start points and SOC durations, achieving data augmentation of the input features and expanding the richness and diversity of the sample space.

[0057] like Figure 4 As shown, Figure 4 This is a graph showing the evolution of the peak voltage value and incremental capacity curves as the battery ages; where, Figure 4 (a) in the figure is the curve of the peak position shift of each peak as the battery ages; Figure 4 (b) in the figure shows the evolution curve of the incremental capacity curve as the battery ages. In this embodiment, three characteristic peaks were identified, denoted as peak 1, peak 2, and peak 4. The identified characteristic peaks were numbered sequentially from high to low voltage. Figure 4 In (b), the color gradient bar BoL-EoL represents the aging process of the battery from brand new to the end of its life.

[0058] S6 takes the incremental capacity dataset constructed in S5 as input and embeds the semi-mechanistic and semi-empirical model into the deep learning model in the form of joint loss function coupling, thereby completing the training of the incremental capacity curve reconstruction model.

[0059] In step S6, the deep learning model used is a dual-drive architecture of RCNN-Transformer, which deeply integrates a Residual Convolutional Neural Network (Residual-RCNN) and a Transformer model. Residual-RCNN is abbreviated as RCNN in this application. To overcome the "black box" limitation of purely data-driven models, this application proposes a mechanistic constraint embedding mechanism guided by physical information: specifically, electrochemical characteristic parameters representing the internal state of the battery (including the peak height, peak position, and half-maximum width of each incremental capacity characteristic peak) are used as prior physical knowledge, and their prediction error is coupled as a penalty term to the joint loss function of the model. Through the loss feedback of these multi-dimensional characteristic parameters, strict boundary constraints are imposed on the overall geometric morphology of the incremental capacity curve predicted by the network, thereby effectively suppressing the generation of local extrema or morphological distortions in the model under complex operating conditions, ensuring that the finally reconstructed incremental capacity curve conforms to the electrochemical physical laws of lithium-ion battery evolution.

[0060] Input and output: The input to the deep learning reconstruction network is the current-capacity sequence and the voltage-capacity sequence, and the output is the reconstructed incremental capacity curve (IC curve) and the corresponding characteristic parameters.

[0061] Training process: Based on the current-capacity sequence and voltage-capacity sequence, the RCNN-Transformer network forward inferences to obtain the predicted incremental capacity curve and predicted characteristic parameters; the curve loss and parameter loss are calculated according to the labels of the predicted incremental capacity curve and the actual incremental capacity curve, and the labels of the predicted characteristic parameters and the actual characteristic parameters, respectively, and the network weights are updated by backpropagation driven by the total joint loss function; the above process is iterated until the preset training rounds are reached or the early stopping condition is met to terminate the training.

[0062] The coupling method of the characteristic parameters is as follows: construct the prediction error loss based on the characteristic parameters such as peak height, peak position and half peak width of each incremental capacity peak, and embed it as a mechanism constraint term into the total joint loss function of the deep learning model to form a Physical Information Neural Network (PINN), that is, a reconstructed model.

[0063] S7: Input the time-series data of the battery under test, and output the corresponding incremental capacity curve through the trained reconstruction model, providing support for subsequent battery health status assessment and aging mode analysis.

[0064] like Figure 5As shown, the solid green line represents the predicted incremental capacity curve output by the model, while the dashed red line represents the actual incremental capacity curve obtained from actual measurements. It can be seen that the predicted incremental capacity curve maintains the same overall trend as the actual incremental capacity curve, and the peak position, peak size, and peak shape of each peak can be effectively reproduced with only minor errors. This verifies the reliability of the predicted incremental capacity curve in this application.

[0065] This application relates to a method for obtaining incremental capacity curves of lithium-ion batteries, comprising: constructing an experimental test matrix covering temperature, CCCV segment number, charging rate, and SOC switching point; conducting systematic charge-discharge tests and simultaneously collecting current, voltage, and capacity time-series data; periodically performing capacity calibration experiments to obtain maximum usable capacity and incremental capacity curve labels; establishing a semi-mechanistic, semi-empirical model and fitting and extracting characteristic parameters such as peak height, peak position, and half-peak width; constructing a dataset using IQ and VQ sequences as inputs and incremental capacity curves and their characteristic parameters as labels; coupling the characteristic parameters as mechanistic constraints to a joint loss function to train a deep learning model to obtain a reconstructed model; and finally, inputting the charge-discharge sequence of the battery under test to reconstruct the incremental capacity curve online. This application achieves a deep integration of physical mechanisms and deep learning, breaking through the strict limitations of traditional constant current and low-rate operating conditions, and can directly obtain incremental capacity curves under dynamic fast charging conditions, significantly improving engineering practicality.

[0066] Compared with the prior art, the method for obtaining the incremental capacity curve of lithium-ion batteries proposed in this application has the following beneficial effects.

[0067] (1) This application effectively overcomes the serious dependence of traditional incremental capacity curve analysis on extremely small constant current (such as 1 / 20C) and extremely long test time. It can directly adapt to the multi-stage CCCV fast charging and dynamic discharge conditions commonly found in practical complex applications, effectively eliminates the masking of characteristic peaks by polarization effect, greatly reduces test costs, and improves the engineering practicality of this technology in vehicle or energy storage scenarios.

[0068] (2) This application innovatively couples the key electrochemical characteristic parameters (peak height, peak position, and half-peak width) obtained from the semi-mechanistic and semi-empirical model into the joint loss function of deep learning in the form of pre-constraints, thus constructing a physical information neural network. This mechanism uses the underlying electrochemical physical laws to guide the optimization direction of the network gradient, effectively suppressing the local extrema or non-physical morphological distortions (such as peak shape distortion) that are prone to occur in pure data-driven models under complex conditions. This not only accelerates the convergence speed of the model, but also improves the physical consistency and interpretability of the deep learning prediction results.

[0069] (3) This application constructs an RCNN-Transformer dual-drive architecture, which uses RCNN to accurately capture the abrupt changes in local voltage / current response in VQ and IQ sequences, and combines the self-attention mechanism of Transformer to deeply analyze the global contextual dependencies of long sequence data. The two complement each other, fully exploring the spatiotemporal coupling features hidden under dynamic operating conditions, enabling the model to have high feature transfer capability and reconstruction accuracy when facing unknown temperatures, different fast charging strategies and different aging stages, providing a reliable technical foundation for battery state estimation and aging mode diagnosis across operating conditions and platforms.

[0070] In summary, one embodiment of this application is feasible and can be used to obtain incremental capacity curves of lithium-ion batteries, thereby providing support for subsequent battery health status assessment and aging mode analysis.

[0071] Obviously, the above embodiments of this application are merely examples for clearly illustrating this application, and are not intended to limit the implementation of this application. Those skilled in the art can make other changes within the main concept of this application, and these obvious variations should be included within the scope of protection claimed in this application.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for obtaining the incremental capacity curve of a lithium-ion battery, characterized in that, include: Construct an experimental test matrix for multi-stage constant current-constant voltage fast charging conditions; Based on the charging and discharging equipment, the lithium-ion battery is subjected to charge-discharge cycle tests using the experimental test matrix to obtain time-series data of the lithium-ion battery and the incremental capacity curve corresponding to the time-series data; the time-series data includes current-capacity sequence and voltage-capacity sequence. Based on a semi-mechanistic, semi-empirical model, characteristic parameters of the incremental capacity curve are extracted; wherein, the characteristic parameters include the peak height, peak position, and half-peak width of each peak in the incremental capacity curve. The incremental capacity dataset is determined based on the time-series data, the incremental capacity curve, and the characteristic parameters. Based on the incremental capacity dataset and joint loss function, and combined with a deep learning model, a reconstruction model is determined; wherein, the deep learning model is determined by fusing a residual convolutional neural network and a Transformer model. The time-series data of the lithium-ion battery under test is input into the reconstruction model to determine the incremental capacity curve of the lithium-ion battery under test.

2. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 1, characterized in that, Construct an experimental test matrix for multi-stage constant current-constant voltage fast charging conditions, specifically including: Based on ambient temperature, the number of constant current-constant voltage charging stages, the charging current rate of stage I, and the switching points of the battery state of charge corresponding to the multi-stage constant current-constant voltage charging, an experimental test matrix for multi-stage constant current-constant voltage fast charging is constructed.

3. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 1, characterized in that, Based on a semi-mechanistic, semi-empirical model, the characteristic parameters of the incremental capacity curve are extracted, specifically including: The characteristic parameters of the incremental capacity curve are determined by fitting the semi-mechanistic, semi-empirical model using the least squares algorithm.

4. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 1, characterized in that, Based on the charging and discharging equipment, and using the aforementioned experimental test matrix, charge-discharge cycle tests are performed on the lithium-ion battery to obtain time-series data of the lithium-ion battery and the incremental capacity curve corresponding to the time-series data. Specifically, this includes: Based on the charging and discharging equipment, all experimental conditions in the experimental test matrix are traversed to collect time-series data corresponding to different experimental conditions of lithium-ion batteries. Based on the preset interval test number and capacity calibration experiment, the lithium-ion battery is subjected to low-rate charge and discharge test to determine the current maximum usable capacity of the battery and the incremental capacity curve corresponding to the time series data.

5. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 1, characterized in that, Based on the incremental capacity dataset and joint loss function, and combined with the deep learning model, the reconstruction model is determined, specifically including: The incremental capacity dataset is input into the residual convolutional neural network, which outputs a deep feature sequence. The deep feature sequence is input into the Transformer model, which outputs the predicted incremental capacity curve and predicted characteristic parameters. Based on the predicted incremental capacity curve and the actual incremental capacity curve, the incremental capacity loss is determined; Based on the predicted characteristic parameters and the actual characteristic parameters, the characteristic parameter loss is determined; By combining the incremental capacity loss and the characteristic parameter loss, a joint loss function is determined. With the goal of minimizing the joint loss function, the residual convolutional neural network and the Transformer model are trained, and the trained deep learning model is determined as the reconstruction model.

6. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 1, characterized in that, The joint loss function is: ； in, ; Loss weights are assigned to characteristic parameters; Total loss; This represents the loss due to the incremental capacity curve. To predict the incremental capacity curve; This represents the actual incremental capacity curve; Mean squared error; α is the mean absolute error; The weights; ; For characteristic parameter loss; K This represents the total number of characteristic parameters. For predicting characteristic parameters, These are the actual characteristic parameters; This is the index of the characteristic parameter.

7. The method for obtaining the incremental capacity curve of a lithium-ion battery according to claim 2, characterized in that, The ambient temperature includes multiple temperature gradients within the range of 0°C to 50°C; the constant current-constant voltage charging stages include 1-stage constant current-constant voltage fast charging and 2-stage constant current-constant voltage fast charging; the first-stage charging current rate includes multiple discrete current rates within the range of 2C to 8C, where C is the rated capacity of the battery; the battery state of charge switching points corresponding to the multi-stage constant current-constant voltage charging include multiple state of charge switching points within the range of 20% to 90%.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for obtaining the incremental capacity curve of a lithium-ion battery according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for obtaining the incremental capacity curve of a lithium-ion battery as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for obtaining the incremental capacity curve of a lithium-ion battery as described in any one of claims 1-7.

Images