A battery cycle performance prediction method based on multi-modal structure representation and electrochemical dynamic coupling

Abstract

The application provides a battery cycle performance prediction method based on multi-modal structure representation and electrochemical dynamic coupling, and mainly relates to the technical field of batteries. The prediction method comprises the following steps: obtaining an electrochemical dynamic feature set of a to-be-tested battery, disassembling and taking out a positive plate to obtain a static multi-modal structure feature set of the positive plate of the to-be-tested battery, and inputting the electrochemical dynamic feature set, the static multi-modal structure feature set and a structure-electrochemical coupling term into a pre-trained structure-electrochemical coupling mapping model, wherein the structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multi-modal structure feature set, and finally outputting the grade, capacity retention rate, cycle life and aging risk point of the to-be-tested battery. Compared with the prior art, the prediction method can simultaneously output the cycle life, the capacity retention rate and the aging risk point, and identify the hidden micro defects of the to-be-tested battery.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and in particular to a method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling. Background Technology

[0002] Lithium iron phosphate (LFP) batteries have been widely used in new energy vehicles, energy storage power stations, and portable energy storage devices due to their advantages such as long cycle life, excellent safety performance, and low cost. Battery cycle life is a core indicator for evaluating the quality of LFP batteries. Traditional life verification methods involve long-term full-rate cycle aging tests, with testing cycles lasting several months. This results in extremely low testing efficiency and cannot meet the needs of rapid screening and incoming material quality grading for mass production of batteries.

[0003] Currently, the mainstream prediction technologies in the industry are mainly divided into three categories. The first category is the static structure characterization method, which mainly uses XRD and Raman spectroscopy to detect the single-point structure of the positive electrode sheet and roughly evaluates the battery performance through single parameters such as carbon coating crystallinity and carbon coating disorder. The second category is the electrochemical characteristic analysis method, which uses electrochemical signals such as charge-discharge curves, dQ / dV differential curves, and AC impedance spectroscopy to predict battery life. The third category is the simple data fusion scheme, which linearly superimposes structural parameters and electrochemical parameters for determination.

[0004] Existing technologies lack standardized procedures for various testing conditions and sample pretreatment processes, resulting in poor experimental repeatability. Judgment thresholds are mostly subjective and empirical values, and the interpretability and engineering feasibility of models are insufficient. Common problems include single characterization dimensions, disconnect between structural and electrochemical characteristics, lack of mechanistic support for models, non-standard procedures, and low prediction accuracy. Summary of the Invention

[0005] This invention proposes a method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling, which solves the technical problems of traditional detection methods that are difficult to identify hidden micro-defects and difficult to simultaneously output cycle life, capacity retention rate and aging risk points.

[0006] To achieve the above objectives, this invention proposes a method for predicting battery cycle performance based on the coupling of multimodal structural characterization and electrochemical dynamics. This method includes the following steps: The battery under test is activated to obtain the activated battery under test. During the activation process, open circuit voltage data and AC impedance spectrum data under different states of charge are collected. Based on the open circuit voltage data and the AC impedance spectrum data under different states of charge, an initial interface-impedance-state three-dimensional baseline of the battery under test is established. Charge-discharge tests were performed on the activated test battery to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, the AC impedance spectroscopy data, and the three-dimensional baseline of the initial interface-impedance-state, an electrochemical dynamic feature set was extracted. The activated battery under test is disassembled, the positive electrode sheet is removed and pre-treated to obtain a pre-treated positive electrode sheet, and the pre-treated positive electrode sheet is subjected to multi-modal structure characterization test to extract the static multi-modal structure feature set. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set; The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term are input into a pre-trained structure-electrochemical coupling mapping model, and the predicted equivalent cycle life of the battery under test is output. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set. The predicted equivalent cycle life of the battery under test is comprehensively checked and corrected with the preset grading judgment rules, and the grade, capacity retention rate, cycle life and aging risk points are output.

[0007] In one embodiment, the pre-trained structure-electrochemical coupling mapping model is obtained through the following steps: Multiple training battery samples were activated separately to obtain activated training battery samples. Open circuit voltage data and AC impedance spectrum data under different charging states were collected during the activation process. Charge-discharge tests were performed on the activated training battery samples to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data, features were extracted to obtain electrochemical dynamic feature sets of multiple training battery samples. The activated training battery sample was disassembled, the positive electrode sheet was removed and pre-treated to obtain the pre-treated positive electrode sheet, and the pre-treated positive electrode sheet was tested for multimodal structure characterization to obtain the static multimodal structure feature set of multiple training battery samples. The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term of the training battery samples are input into a multiple linear regression equation for calculation, resulting in the weight coefficients of the static multimodal structural features, the weight coefficients of the electrochemical dynamic features, and the weight coefficients of the structure-electrochemical coupling term for multiple training battery samples. The multiple linear regression equation is Y = w1·S + w2·E + w3·(S×E). The initial weight coefficients of the model are obtained by solving the static multimodal structural feature weight coefficients, electrochemical dynamic feature weight coefficients, and structure-electrochemical coupling term weight coefficients of the multiple training battery samples using the least squares iterative method. After calibration, the final weight coefficients w1, w2, and w3 of the model are obtained, thus obtaining the pre-trained structure-electrochemical coupling mapping model.

[0008] In one embodiment, the activation method for the battery under test is as follows: The battery under test was placed in a closed, constant temperature and humidity environment, and activated using a step-by-step activation method. The stepwise activation method includes the following steps: The charging and discharging steps are as follows: constant current charging is performed using constant current A1 until the cutoff voltage is V1, then the circuit is left to stand; then constant current discharging is performed using constant current A2 until the cutoff voltage is V2, then the circuit is left to stand. Wherein, 0.05A≤A1≤0.15A, 3.0V≤V1≤4.0V, 0.15A≤A2≤0.25A, and 2.0V≤V2≤3.0V; Repeat the charging and discharging steps to complete the stepwise activation treatment.

[0009] In one embodiment, in the step of collecting open-circuit voltage data and AC impedance spectrum data under different charging states during the activation process, the AC impedance spectrum data under different charging states includes AC impedance spectrum data under charging states of 15-25%, 45-55%, and 75-85%.

[0010] In one embodiment, the method for extracting the electrochemical dynamic feature set based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data includes the following steps: Capacity differential analysis is performed on the raw voltage-capacity data to obtain the dQ / dV curve. The charge-discharge peak position difference ΔVp, peak symmetry Sp, and main-side reaction peak height ratio Rp are extracted from the dQ / dV curve. Voltage differential analysis is performed on the raw voltage-capacity data to obtain the dV / dQ curve. The curve split degree Ds is extracted from the dV / dQ curve. The dV / dQ curve is compared with the fluctuation value of the three-dimensional baseline of the initial interface-impedance-state to obtain the baseline fluctuation value σb. The corresponding charge transfer impedance Rct is obtained from the AC impedance spectrum data under different charging states. The charge transfer impedance Rct is fitted with the open circuit voltage to construct the Rct-V impedance-voltage coupling curve. The impedance change rate kR and impedance change characteristic voltage VR are extracted from the Rct-V impedance-voltage coupling curve. Obtain the electrochemical dynamic feature set, which is expressed as: E=[ΔVp, Sp, Rp, Ds, σb, kR, VR].

[0011] In one embodiment, in the step of disassembling the activated training battery sample, removing the positive electrode sheet, and performing pretreatment to obtain a pretreated positive electrode sheet, the pretreatment includes the following steps: The positive electrode sheet was soaked in dimethyl carbonate and diethyl carbonate in sequence to obtain a cleaned positive electrode sheet. The cleaned positive electrode sheet is placed in a vacuum oven to dry, thereby obtaining a dried positive electrode sheet. The dried positive electrode sheet is placed in a glove box filled with inert gas and allowed to cool naturally to room temperature to obtain a cooled positive electrode sheet. The cooled positive electrode sheet is cut to obtain a pretreated positive electrode sheet.

[0012] In one embodiment, the tests for characterizing the multimodal structure include XRD, Raman spectroscopy, and XPS; and / or, The static multimodal structural feature set includes the crystallinity XC of the carbon coating, the proportion of harmful impurities, the shift of XRD diffraction characteristic peaks Δ2θ, and PO4. 3 -Characteristic peak half-width at half-maximum, Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, and F e2+ / Fe 3+ Price ratio; and / or, Obtain a static multimodal structural feature set, which is represented as: S = [Δ2θ, XC, percentage of harmful impurities, PO4]. 3- Peak half-width at half-maximum (HWHM), Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, Fe 2+ / Fe 3+ ].

[0013] In one embodiment, the process of comprehensively verifying and correcting the predicted equivalent cycle life value of the battery under test against a preset grading rule includes: When the output of the predicted equivalent cycle life value of the battery under test is displayed as the first set number of cycles, the judgment rule of the first-class high-quality battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the second set number of cycles, the judgment rule of the second-level qualified battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the third set number of cycles, the judgment rule of the third-level defective battery is adopted.

[0014] The criteria for determining the first-class high-quality battery are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously satisfy the following requirements: crystallinity of crystalline carbon coating XC ≥ 92%, total proportion of harmful impurity phases ≤ 1.2%, and XRD diffraction characteristic peak shift Δ2θ ≤ 0.08. ° PO4 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.32cm. -1 The Raman carbon peak intensity ratio was 0.85-1.1; the phosphate test cell / test cell carbon peak intensity ratio (IPO4 / ID) was 0.85-1.10; Fe... 2+ / Fe 3+ ≤0.1; The electrochemical parameters in the electrochemical dynamic characteristic set must simultaneously satisfy ΔVp≤0.02V, Sp is 0.95-1.05, Rp≤0.08, Ds≤0.1, σb≤0.002, kR≤4% / V, and the impedance change characteristic voltage VR does not show a sudden change inflection point; The structure-electrochemical coupling term must satisfy S×E≤0.35; The criteria for determining whether a battery is qualified at level two are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously satisfy the following conditions: 92% > XC ≥ 80%, 1.2% < total proportion of harmful impurity phases ≤ 3.0%, XRD diffraction characteristic peak shift 0.08° < Δ2θ ≤ 0.15°, and 0.32cm. -1 <PO4 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.55cm. -1 The Raman carbon peak intensity ratio I1350 / I1600 is 0.70-0.85 or 1.10-1.30; the phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID is 0.25-0.4 or 0.6-0.8; 0.1 < Fe 2+ / Fe 3+ ≤0.25; The electrochemical parameters in the electrochemical dynamic characteristic set must meet the following conditions: 0.02V < ΔVp ≤ 0.05V, Sp = 0.90-0.95 or 1.05-1.10, 0.08 < Rp ≤ 0.18, 0.1 < Ds ≤ 0.25, 0.002 < σb ≤ 0.006, 4% / V < kR ≤ 9% / V, and the impedance abrupt change characteristic voltage VR should show a single small abrupt change. The structure-electrochemical coupling term must satisfy: 0.35 < S × E ≤ 0.7; The criteria for determining the Level 3 defective battery are as follows: Among the structural parameters in the static multimodal structural feature set of the battery under test, 80% > XC, the total proportion of harmful impurity phases > 3.0%, the XRD diffraction characteristic peak shift Δ2θ > 0.15°, and PO4 3- Characteristic peak half-width at half-maximum > 0.55 cm -1 Raman carbon peak intensity ratio I1350 / I1600 < 0.70 or > 1.30; phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID < 0.25 or > 0.80; Fe 2+ / Fe 3+ >0.25; Among the electrochemical parameters in the electrochemical dynamic characteristic set, ΔVp > 0.05V, Sp is < 0.90 or > 1.10, Rp > 0.18, Ds > 0.25, σb > 0.006, kR > 9% / V, and the impedance mutation characteristic voltage VR shows multiple obvious abrupt changes; The structure-electrochemical coupling term S×E > 0.70.

[0015] In one embodiment, the output level, capacity retention, cycle life, and aging risk point include: Outputs a Class 1 high-quality battery, with a capacity retention of ≥85% and a cycle life of ≥2000 cycles at 1C rate, showing no obvious aging defects after 2000 cycles; or, The output is a Class II qualified battery, with a capacity retention of 80-85% after 2000 cycles at 1C rate, and a cycle life of 1500-2000 cycles. A slight interface instability issue exists. Alternatively, The output is a level 3 defective battery, and the output shows that the capacity retention rate is <80% and the cycle life is <1500 cycles after 2000 cycles at 1C rate. Simultaneously, it identifies and outputs aging risk points, which are one or more of the following: lattice distortion, abnormal proportion of harmful impurities, carbon coating failure, abnormal interface impedance, and deterioration of reaction reversibility.

[0016] This invention proposes a method for predicting battery cycle performance based on the coupling of multimodal structural characterization and electrochemical dynamics. The method includes the following steps: activating the battery under test to obtain an activated battery; during the activation process, collecting open-circuit voltage data and AC impedance spectroscopy data at different states of charge; establishing a three-dimensional baseline of the initial interface-impedance-state of the battery under test based on the open-circuit voltage data and the AC impedance spectroscopy data at different states of charge; performing charge-discharge tests on the activated battery to collect raw voltage-capacity data; and extracting an electrochemical dynamic feature set based on the raw voltage-capacity data, the open-circuit voltage data, the AC impedance spectroscopy data, and the three-dimensional baseline of the initial interface-impedance-state; and disassembling the activated battery. For the battery under test, the positive electrode sheet is removed and pre-processed to obtain a pre-processed positive electrode sheet. Multimodal structural characterization tests are performed on the pre-processed positive electrode sheet to extract a static multimodal structural feature set. A structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set. The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term are input into a pre-trained structure-electrochemical coupling mapping model, which outputs a predicted equivalent cycle life value for the battery under test. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set. The predicted equivalent cycle life value of the battery under test is comprehensively checked and corrected against preset grading rules, and the grade, capacity retention rate, cycle life, and aging risk points are output. The battery cycle performance prediction method based on multimodal structural characterization and electrochemical dynamic coupling described in this invention makes a comprehensive judgment based on structural parameters in the static multimodal structural feature set, electrochemical parameters in the electrochemical dynamic feature set, structure-electrochemical coupling terms, and three-level battery judgment rules. It can simultaneously output cycle life, capacity retention rate, and aging risk points, and identify hidden micro-defects in the battery under test. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Where specific conditions are not specified in the embodiments, conventional conditions or conditions recommended by the manufacturer shall apply. Where the manufacturers of reagents or instruments are not specified, they are all conventional products that can be purchased commercially. Furthermore, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, or solution B, or a solution where both A and B are satisfied simultaneously. In addition, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Currently, the mainstream prediction technologies in the industry are mainly divided into three categories. The first category is the static structure characterization method, which mainly uses XRD and Raman spectroscopy to detect the single-point structure of the positive electrode sheet and roughly evaluates the battery performance through single parameters such as carbon coating crystallinity and carbon coating disorder. This type of method only reflects the static intrinsic structure of the material and cannot be correlated with the actual electrochemical reaction process of the battery. The correlation between structural parameters and cycle life is weak, resulting in poor prediction accuracy. At the same time, existing schemes mostly use a combination of dual characterizations without combining elemental valence state and surface functional group analysis, which is insufficient for identifying electrode micro-defects and interface degradation problems. The second category is the electrochemical characteristic analysis method, which uses electrochemical signals such as charge-discharge curves, dQ / dV differential curves, and AC impedance spectroscopy to predict cycle life. This type of method focuses on macroscopic electrical performance but ignores intrinsic issues such as the crystal structure and carbon coating state of the positive electrode sheet material. When the material has hidden lattice defects or trace impurities, the electrochemical signal does not show obvious abnormalities in the early stage, which can easily lead to misjudgment. The third category is a simple data fusion scheme that linearly superimposes structural and electrochemical parameters for judgment. However, it fails to construct a coupled correlation model based on the aging mechanism of lithium iron phosphate batteries. This approach is a simple data combination rather than mechanism-driven fusion, resulting in poor model interpretability and weak anti-interference capabilities. Furthermore, existing technologies lack standardized procedures for various testing conditions and sample pretreatment processes, leading to poor experimental repeatability. Judgment thresholds are often based on subjective experience, resulting in insufficient model interpretability and engineering feasibility. Due to the shortcomings of existing technologies, such as limited characterization dimensions, disconnect between structural and electrochemical features, lack of mechanistic support in the model, non-standardized procedures, and low prediction accuracy, a multi-dimensional, standardized, and mechanism-driven method for rapid prediction of the cycle performance of lithium iron phosphate batteries is urgently needed.

[0019] In view of this, the present invention proposes a method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling, comprising the following steps: The battery under test is activated to obtain the activated battery under test. During the activation process, open circuit voltage data and AC impedance spectrum data under different states of charge are collected. Based on the open circuit voltage data and the AC impedance spectrum data under different states of charge, an initial interface-impedance-state three-dimensional baseline of the battery under test is established. Charge-discharge tests were performed on the activated test battery to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, the AC impedance spectroscopy data, and the three-dimensional baseline of the initial interface-impedance-state, an electrochemical dynamic feature set was extracted. The activated battery under test is disassembled, the positive electrode sheet is removed and pre-treated to obtain a pre-treated positive electrode sheet, and the pre-treated positive electrode sheet is subjected to multi-modal structure characterization test to extract the static multi-modal structure feature set. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set; The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term are input into a pre-trained structure-electrochemical coupling mapping model, and the predicted equivalent cycle life of the battery under test is output. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set. The predicted equivalent cycle life of the battery under test is comprehensively checked and corrected with the preset grading judgment rules, and the grade, capacity retention rate, cycle life and aging risk points are output.

[0020] This invention proposes a multi-dimensional, standardized, and mechanism-driven method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling. This method standardizes the testing process, ensures parameter reproducibility, and allows for cross-sectional comparison of test results from different batches of batteries. The overall prediction accuracy for validation samples reaches 96.70%, higher than the 78.30% of traditional single-structure characterization methods and the 81.70% of traditional single-electrochemical analysis methods. Furthermore, the entire testing cycle for a single battery does not exceed one working day, balancing prediction accuracy, testing efficiency, model interpretability, and engineering feasibility. By comprehensively judging the structural parameters from the static multimodal structural feature set, the electrochemical parameters from the electrochemical dynamic feature set, and the structure-electrochemical coupling terms with the three-level battery judgment rules, the method can simultaneously output cycle life, capacity retention, and aging risk points, identifying latent micro-defects in the battery under test.

[0021] Brand-new batteries exhibit individual variations in charge, interfacial impedance, and internal polarization. By acquiring an initial three-dimensional baseline of interface-impedance-state, the interference of initial state differences on subsequent electrochemical feature extraction can be avoided, eliminating these initial state variations and establishing a unified reference standard. Subsequent charge / discharge cycles and the extraction of electrochemical dynamic feature sets are all referenced to this initial three-dimensional baseline. This allows for precise calculation of the interface impedance shift during charge / discharge, quantification of the degree of electrode interface degradation, and quantitative comparison of the dynamic evolution of interface impedance.

[0022] Meanwhile, the initial interface-impedance-state three-dimensional baseline can distinguish between the inherent initial impedance deviation caused by fluctuations in production assembly and liquid injection processes and the irreversible impedance degradation caused by material lattice defects and carbon coating defects, thereby reducing the probability of misjudging initial process fluctuations as intrinsic material defects. By adopting a unified baseline calibration rule for all batteries under test, it is also possible to correct data deviations generated during testing in different laboratories, with different testing equipment, and in different batches, enabling horizontal comparison of test results and improving the standardization of the testing process and the repeatability of tests in industrial applications.

[0023] In one embodiment, the pre-trained structure-electrochemical coupling mapping model is obtained through the following steps: Multiple training battery samples were activated separately to obtain activated training battery samples. Open circuit voltage data and AC impedance spectrum data under different charging states were collected during the activation process. Charge-discharge tests were performed on the activated training battery samples to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data, features were extracted to obtain electrochemical dynamic feature sets of multiple training battery samples. The activated training battery sample was disassembled, the positive electrode sheet was removed and pre-treated to obtain the pre-treated positive electrode sheet, and the pre-treated positive electrode sheet was tested for multimodal structure characterization to obtain the static multimodal structure feature set of multiple training battery samples. The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term of the training battery samples are input into a multiple linear regression equation for calculation, resulting in the weight coefficients of the static multimodal structural features, the weight coefficients of the electrochemical dynamic features, and the weight coefficients of the structure-electrochemical coupling term for multiple training battery samples. The multiple linear regression equation is Y = w1·S + w2·E + w3·(S×E). The initial weight coefficients of the model are obtained by solving the static multimodal structural feature weight coefficients, electrochemical dynamic feature weight coefficients, and structure-electrochemical coupling term weight coefficients of the multiple training battery samples using the least squares iterative method. After calibration, the final weight coefficients w1, w2, and w3 of the model are obtained, thus obtaining the pre-trained structure-electrochemical coupling mapping model.

[0024] By employing the aforementioned training method, a correspondence is established between static multimodal structural features, electrochemical dynamic features, and structure-electrochemical coupling terms and battery cycle life. This quantifies the contribution of these features to cycle life, avoiding the mere linear superposition of structural and electrochemical parameters. Finally, the final weight coefficients w1, w2, and w3 of the model are calculated, resulting in a pre-trained structure-electrochemical coupling mapping model. This model can correlate microstructural defects in the cathode electrode with macroscopic electrochemical performance anomalies in the battery, improving the model's adaptability to different batches of batteries, prediction accuracy, and ability to explain aging risks.

[0025] In one embodiment, the activation method for the battery under test is as follows: The battery under test was placed in a closed, constant temperature and humidity environment, and activated using a step-by-step activation method. The stepwise activation method includes the following steps: The charging and discharging steps are as follows: constant current charging is performed using constant current A1 until the cutoff voltage is V1, then the circuit is left to stand; then constant current discharging is performed using constant current A2 until the cutoff voltage is V2, then the circuit is left to stand. Wherein, 0.05A≤A1≤0.15A, 3.0V≤V1≤4.0V, 0.15≤A2≤0.25, 2.0≤V2≤3.0; Repeat the charging and discharging steps to complete the stepwise activation treatment.

[0026] By using a step-by-step standardized activation process, the electrolyte is fully wetted into the electrode pores, and the electrode interface is gradually stabilized. This reduces the individual differences in battery charge, interface impedance, and internal polarization at the factory. It allows different batteries under test to be tested under relatively consistent initial conditions for open-circuit voltage acquisition, multi-state-of-charge AC impedance testing, and subsequent charge-discharge testing. This reduces the interference of initial state differences on the electrochemical dynamic feature extraction results, and provides a unified and reproducible data foundation for establishing a three-dimensional baseline of battery initial interface-impedance-state and for subsequent cycle performance prediction.

[0027] In one embodiment, in the step of collecting open-circuit voltage data and AC impedance spectrum data under different charging states during the activation process, the AC impedance spectrum data under different charging states includes AC impedance spectrum data under charging states of 15-25%, 45-55%, and 75-85%.

[0028] Multiple state-of-charge (SOC) testing can capture the dynamic changes in interfacial impedance. AC impedance spectra collected across low, medium, and high SOC ranges cover the interfacial charge transfer and diffusion states at different lithiation levels. Low SOC helps observe the interfacial impedance at higher delithiation degrees; medium SOC reflects phase transitions and reaction reversibility within the operating range; and high SOC helps identify polarization and impedance anomalies near the charging end. By extracting the charge transfer impedance Rct at different SOCs, Rct-V curves can be fitted, and the impedance change rate kR = ΔRct / ΔV can be calculated. Simultaneously, the characteristic voltage VR of impedance abrupt change can be extracted, rather than using AC impedance spectrum data from a single SOC to replace the dynamic response.

[0029] In one embodiment, the method for extracting the electrochemical dynamic feature set based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data includes the following steps: Capacity differential analysis is performed on the raw voltage-capacity data to obtain the dQ / dV curve. The charge-discharge peak position difference ΔVp, peak symmetry Sp, and main-side reaction peak height ratio Rp are extracted from the dQ / dV curve. Voltage differential analysis is performed on the raw voltage-capacity data to obtain the dV / dQ curve. The curve split degree Ds is extracted from the dV / dQ curve. The dV / dQ curve is compared with the fluctuation value of the three-dimensional baseline of the initial interface-impedance-state to obtain the baseline fluctuation value σb. The corresponding charge transfer impedance Rct is obtained from the AC impedance spectrum data under different charging states. The charge transfer impedance Rct is fitted with the open circuit voltage to construct the Rct-V impedance-voltage coupling curve. The impedance change rate kR and impedance change characteristic voltage VR are extracted from the Rct-V impedance-voltage coupling curve. Obtain the electrochemical dynamic feature set, which is expressed as: E=[ΔVp, Sp, Rp, Ds, σb, kR, VR].

[0030] This invention utilizes seven characteristics—charge-discharge peak position difference ΔVp, peak symmetry Sp, main-side reaction peak height ratio Rp, curve splitting degree Ds, baseline fluctuation value σb, impedance change rate kR, and impedance jump characteristic voltage VR—to construct an electrochemical dynamic feature set for determining the battery under test, rather than relying solely on a single capacity or impedance parameter for lifetime assessment. Specifically, the dQ / dV curve obtained from capacity differential analysis is used to evaluate the reversibility of electrochemical reactions. By identifying the charge-discharge peak position difference, peak symmetry, and main-side reaction peak height ratio, it reflects the reversibility of redox reactions and the degree of side reactions. The dV / dQ curve obtained from voltage differential analysis is used to evaluate electrode phase transition uniformity and interface stability, revealing curve splitting and baseline fluctuation, reflecting electrode phase transition uniformity and interface stability. At different SOCs, the charge transfer impedance Rct data is fitted to obtain the Rct-V curve, and the impedance change rate kR = ΔRct / ΔV is calculated, allowing for further extraction of the impedance change rate and impedance jump characteristic voltage.

[0031] In one embodiment, in the step of disassembling the activated training battery sample, removing the positive electrode sheet, and performing pretreatment to obtain a pretreated positive electrode sheet, the pretreatment includes the following steps: The positive electrode sheet was soaked in dimethyl carbonate and diethyl carbonate in sequence to obtain a cleaned positive electrode sheet. The cleaned positive electrode sheet is placed in a vacuum oven to dry, thereby obtaining a dried positive electrode sheet. The dried positive electrode sheet is placed in a glove box filled with inert gas and allowed to cool naturally to room temperature to obtain a cooled positive electrode sheet. The cooled positive electrode sheet is cut to obtain a pretreated positive electrode sheet.

[0032] This invention reduces characterization bias caused by battery differences through pretreatment, ensuring that the subsequently obtained static structural features correspond to the true positive electrode state of the same battery after activation. Firstly, sequential soaking in dimethyl carbonate and diethyl carbonate removes residual electrolyte and lithium salt from the surface of the positive electrode, preventing residues from interfering with XRD, Raman spectroscopy, and XPS signals. Vacuum drying removes solvent and moisture, while natural cooling in an inert gas environment reduces the risk of electrode oxidation and moisture absorption. Trimming ensures consistency in the measured area and size across different samples.

[0033] In one embodiment, the tests for characterizing the multimodal structure include XRD, Raman spectroscopy, and XPS; and / or, The static multimodal structural feature set includes the crystallinity XC of the carbon coating, the proportion of harmful impurities, the shift of XRD diffraction characteristic peaks Δ2θ, and PO4. 3-Characteristic peak half-width at half-maximum, Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, and F e2+ / Fe 3+ Price ratio; and / or, Obtain a static multimodal structural feature set, which is represented as: S = [Δ2θ, XC, percentage of harmful impurities, PO4]. 3- Peak half-width at half-maximum (HWHM), Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, Fe 2+ / Fe 3+ ].

[0034] This invention employs a characterization method combining XRD, Raman spectroscopy, and XPS to determine the static structural characteristics of a battery from multiple dimensions, including lattice distortion, carbon coating crystallinity, impurity phases, carbon coating, and elemental valence states. Specifically, XRD allows for the quantitative calculation of structural parameters, extracting the crystallinity of the carbon coating, the proportion of harmful impurity phases, and characteristic peak shifts to reflect lattice integrity and phase purity. Raman spectroscopy avoids damage to the sample from excessively high laser power and can extract the full width at half maximum (FWHM) of the phosphate characteristic peak, the Raman carbon peak intensity ratio I1350 / I1600, and the phosphate-based battery / carbon peak intensity ratio IPO4 / ID to reflect the local structure of the phosphate group and the orderliness of the carbon coating. XPS is used for quantitative analysis of elemental valence states and surface functional groups, obtaining data on Fe... 2+ / Fe 3+ Surface chemical information such as valence ratios is used to identify valence anomalies and interface aging trends.

[0035] In one embodiment, the predicted equivalent cycle life value of the battery under test is comprehensively checked and corrected against a preset grading rule, including: When the output of the predicted equivalent cycle life value of the battery under test is displayed as the first set number of cycles, the judgment rule of the first-class high-quality battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the second set number of cycles, the judgment rule of the second-level qualified battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the third set number of cycles, the judgment rule of the third-level defective battery is adopted.

[0036] To avoid misjudgments caused by judging based on a single prediction count, the following rules are triggered: when the output shows the first set number of predictions, it is determined as a Level 1 high-quality battery; when the output shows the second set number of predictions, it is determined as a Level 2 qualified battery; and when the output shows the third set number of predictions, it is determined as a Level 3 defective battery.

[0037] The criteria for determining the first-class high-quality battery are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously meet the following requirements: crystallinity of crystalline carbon coating XC ≥ 92%, total proportion of harmful impurity phases ≤ 1.2%, XRD diffraction characteristic peak shift Δ2θ ≤ 0.08°, and PO4. 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.32cm. -1 The Raman carbon peak intensity ratio was 0.85-1.1; the phosphate test cell / test cell carbon peak intensity ratio (IPO4 / ID) was 0.85-1.10; Fe... 2+ / Fe 3+ ≤0.1; The electrochemical parameters in the electrochemical dynamic characteristic set must simultaneously satisfy ΔVp≤0.02V, Sp is 0.95-1.05, Rp≤0.08, Ds≤0.1, σb≤0.002, kR≤4% / V, and the impedance change characteristic voltage VR does not show a sudden change inflection point; The structure-electrochemical coupling term must satisfy S×E≤0.35; The criteria for determining whether a battery is qualified at level two are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously satisfy the following conditions: 92% > XC ≥ 80%, 1.2% < total proportion of harmful impurity phases ≤ 3.0%, XRD diffraction characteristic peak shift 0.08° < Δ2θ ≤ 0.15°, and 0.32cm. -1 <PO4 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.55cm. -1 The Raman carbon peak intensity ratio I1350 / I1600 is 0.70-0.85 or 1.10-1.30; the phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID is 0.25-0.4 or 0.6-0.8; 0.1 < Fe 2+ / Fe 3+ ≤0.25; The electrochemical parameters in the electrochemical dynamic characteristic set must meet the following conditions: 0.02V < ΔVp ≤ 0.05V, Sp = 0.90-0.95 or 1.05-1.10, 0.08 < Rp ≤ 0.18, 0.1 < Ds ≤ 0.25, 0.002 < σb ≤ 0.006, 4% / V < kR ≤ 9% / V, and the impedance abrupt change characteristic voltage VR should show a single small abrupt change. The structure-electrochemical coupling term must satisfy: 0.35 < S × E ≤ 0.7; The criteria for determining the Level 3 defective battery are as follows: Among the structural parameters in the static multimodal structural feature set of the battery under test, 80% > XC, the total proportion of harmful impurity phases > 3.0%, the XRD diffraction characteristic peak shift Δ2θ > 0.15°, and PO4 3-Characteristic peak half-width at half-maximum > 0.55 cm -1 Raman carbon peak intensity ratio I1350 / I1600 < 0.70 or > 1.30; phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID < 0.25 or > 0.80; Fe 2+ / Fe 3+ >0.25; Among the electrochemical parameters in the electrochemical dynamic characteristic set, ΔVp > 0.05V, Sp is < 0.90 or > 1.10, Rp > 0.18, Ds > 0.25, σb > 0.006, kR > 9% / V, and the impedance mutation characteristic voltage VR shows multiple obvious abrupt changes; The structure-electrochemical coupling term S×E > 0.70.

[0038] Once the output items have been processed a set number of times, the structural parameters in the static multimodal structural feature set, the electrochemical parameters in the electrochemical dynamic feature set, and the structure-electrochemical coupling terms of the battery under test are comprehensively checked and corrected. This further reduces the probability of misclassification and improves the ability to distinguish between Grade 1 high-quality, Grade 2 qualified, and Grade 3 defective batteries. If the predicted cycle life meets the requirements of the corresponding grade, but the structural parameters, electrochemical parameters, or structure-electrochemical coupling terms exceed the threshold of the corresponding grade, the battery under test can be downgraded according to the degree of parameter anomaly. This allows for the screening of batteries whose predicted cycle life is temporarily up to standard but already have microstructural defects or abnormal interface impedance.

[0039] In one embodiment, the output level, capacity retention, cycle life, and aging risk point include: Outputs a Class 1 high-quality battery, with a capacity retention of ≥85% and a cycle life of ≥2000 cycles at 1C rate, showing no obvious aging defects after 2000 cycles; or, The output is a Class II qualified battery, with a capacity retention of 80-85% after 2000 cycles at 1C rate, and a cycle life of 1500-2000 cycles. A slight interface instability issue exists. Alternatively, The output is a level 3 defective battery, and the output shows that the capacity retention rate is <80% and the cycle life is <1500 cycles after 2000 cycles at 1C rate. Simultaneously, it identifies and outputs aging risk points, which are one or more of the following: lattice distortion, abnormal proportion of harmful impurities, carbon coating failure, abnormal interface impedance, and deterioration of reaction reversibility. The output items for the rating, capacity retention, cycle life, and aging risk points are summarized in Table 2.

[0040] By providing outputs such as battery quality grade, capacity retention rate, cycle life, and aging risk points, this invention not only provides the quality grade and estimated cycle life of the battery under test, but also simultaneously provides the corresponding capacity retention rate and aging risk points. It can determine the possible causes of battery performance degradation based on specific risks such as lattice distortion, abnormal proportion of harmful impurities, carbon coating failure, abnormal interface impedance, and deterioration of reaction reversibility. In addition to classifying battery quality grades, estimating cycle life and capacity retention rate, it can also accurately locate specific aging risk points, reducing the time and equipment costs for enterprises' production line sampling, incoming material inspection, and product grading. It provides data support for battery process optimization, raw material selection, and defect traceability, and is suitable for all scenarios of lithium battery production, quality inspection, and R&D.

[0041] Table 1

[0042] Table 2

[0043] Example 1 A method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling includes the following steps: S10. Activate the battery under test to obtain the activated battery under test. During the activation process, collect open circuit voltage data and AC impedance spectrum data under different charging states. Based on the open circuit voltage data and the AC impedance spectrum data under different charging states, establish the initial interface-impedance-state three-dimensional baseline of the battery under test. The stepwise activation method includes the following steps: The charging and discharging steps are as follows: constant current charging is performed at 0.1C until the cutoff voltage is 3.65V, then constant voltage charging is switched until the charging current is ≤0.01C, and then left to stand; then constant current discharging is performed at 0.2C until the cutoff voltage is 2.5V, and then left to stand. Repeat the charging and discharging steps to complete the stepwise activation treatment; S20. Perform charge-discharge tests on the activated battery to be tested, collect raw voltage-capacity data, and extract the electrochemical dynamic feature set E based on the raw voltage-capacity data, the open circuit voltage data, the AC impedance spectrum data, and the three-dimensional baseline of the initial interface-impedance-state. S30. Disassemble the activated battery to be tested, take out the positive electrode sheet and pre-process it to obtain the pre-processed positive electrode sheet, perform multi-modal structure characterization test on the pre-processed positive electrode sheet, and extract the static multi-modal structure feature set S. S40. Input the electrochemical dynamic feature set E, the static multimodal structural feature set S, and the structure-electrochemical coupling term S×E into the pre-trained structure-electrochemical coupling mapping model Y=w1·S+w2·E+w3·(S×E), and output the predicted equivalent cycle life value Y of the battery under test, wherein the structure-electrochemical coupling term S×E is calculated based on the electrochemical dynamic feature set E and the static multimodal structural feature set S; S50. The predicted equivalent cycle life value Y of the battery under test is comprehensively checked and corrected with the preset grading judgment rules, and the grade, capacity retention rate, cycle life and aging risk points of the battery under test are output.

[0044] The judgment rules for first-level high-quality batteries, second-level qualified batteries, and third-level defective batteries are summarized in Table 1. The output items for grade, capacity retention rate, cycle life, and aging risk points are summarized in Table 2.

[0045] In step S40 of Example 1, the pre-trained structure-electrochemical coupling mapping model Y=w1·S+w2·E+w3·(S×E) is obtained through the following steps: S401. Activate 90 training battery samples to obtain activated training battery samples. During the activation process, collect open circuit voltage data and AC impedance spectrum data under different charging states. S402. Charge and discharge tests are performed on the activated training battery samples to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectrum data, features are extracted to obtain an electrochemical dynamic feature set E of multiple training battery samples. S403. Disassemble the activated training battery sample, take out the positive electrode sheet and pre-process it to obtain the pre-processed positive electrode sheet, and test the multi-modal structure characterization of the pre-processed positive electrode sheet to obtain the static multi-modal structure feature set S of multiple training battery samples. S404. Using the true lifetime label of the training battery sample as the dependent variable, and the electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term of the training battery sample as independent variables, the results are input into a multiple linear regression equation to calculate the weight coefficients of the static multimodal structural features, the electrochemical dynamic features, and the structure-electrochemical coupling term for multiple training battery samples. The multiple linear regression equation is Y = w1·S + w2·E + w3·(S×E). S405. The initial weight coefficients of the model are obtained by solving the static multimodal structural feature weight coefficients, electrochemical dynamic feature weight coefficients, and structure-electrochemical coupling term weight coefficients of the multiple training battery samples using the least squares iterative method. After calibration, the final weight coefficients w1, w2, and w3 of the model are obtained, thus obtaining the pre-trained structure-electrochemical coupling mapping model. The calibration method for the manual weight calibration is as follows: referring to the weight proportions of key influencing factors of battery degradation in the national standard GB / T 30084-2013 "Lithium-ion Batteries for Energy Storage" and the industry standard HB / T 4069-2019 "Test Method for Cycle Performance of Lithium-ion Batteries", and combining the contribution of LFP material structure and electrochemical performance to cycle life in relevant literature on lithium iron phosphate material aging, the initial weight coefficients of the model are manually corrected and iterated.

[0046] Comparative Example 1 A method for predicting battery cycle performance based on traditional single-structure characterization method. The difference between Comparative Example 1 and Example 1 is that the prediction method in Comparative Example 1 is based solely on XRD structural characterization.

[0047] Comparative Example 2 A method for predicting battery cycle performance based on a traditional single electrochemical analysis method is presented. The difference between Comparative Example 2 and Example 1 is that Comparative Example 2's prediction method only analyzes linear fitting through dQ / dV and dV / dQ.

[0048] The differences between Example 1 and Comparative Examples 1-2 are shown in Table 3.

[0049] Table 3

[0050] Test Example 1 120 commercially available square lithium iron phosphate batteries from different production batches with known actual cycle life labels were selected. 90 batteries were randomly selected as training battery samples and trained according to the training method described in steps S401-S405 of Example 1 to obtain Y=w1·S+w2·E+w3·(S×E). The remaining batteries were randomly divided into three groups of 10 each, serving as the verification groups for Example 1, Comparative Example, and Comparative Example 2.

[0051] The methods are respectively used to verify the battery cycle performance prediction method based on multimodal structural characterization and electrochemical dynamic coupling in Example 1, to verify the battery cycle performance prediction method based on traditional single structural characterization in Comparative Example 1, and to verify the battery cycle performance prediction method based on traditional single electrochemical analysis in Comparative Example 2.

[0052] The predicted cycle life obtained by the battery cycle performance prediction method of each verification group was compared with its known actual cycle life to obtain the overall prediction accuracy of the verification samples, the prediction accuracy of Level 1 batteries (capacity retention rate ≥ 85%), the prediction accuracy of Level 2 batteries (capacity retention rate 80-85%), and the prediction accuracy of Level 3 batteries (capacity retention rate < 80%). The full-process testing cycle, aging risk points (ability to identify latent micro-defects), model interpretability, and engineering feasibility were also statistically analyzed. The test results are summarized in Table 4.

[0053] Table 4

[0054] As shown in Table 3, the battery cycle performance prediction method based on multimodal structural characterization and electrochemical dynamic coupling in Example 1 tested the lithium iron phosphate battery. Compared with the battery cycle performance prediction method based on the traditional single structural characterization method in Comparative Example 1 and the battery cycle performance prediction method based on the traditional single electrochemical analysis method in Comparative Example 2, it has higher prediction accuracy, model interpretability, and engineering feasibility. Moreover, the entire process detection cycle of the activated battery sample of a single battery under test in Example 1 is shorter, and it also has the ability to identify latent micro-defects. It can identify defects of all dimensions such as lattice distortion, trace impurities, carbon coating defects, and interface degradation. It solves the technical problem that existing detection methods cannot detect latent micro-defects and cannot simultaneously identify and output the estimated cycle life, capacity retention rate, and aging risk points.

[0055] In Comparative Example 1, the traditional single-structure characterization method of XRD was used to judge the lithium iron phosphate battery under test. It can only identify the integrity of the macroscopic crystal structure, but cannot identify latent defects such as interface and valence state. It cannot evaluate important factors affecting cycle life such as carbon coating state, surface element valence state and interface reaction. Therefore, it is difficult to identify latent defects such as slight lattice distortion, trace impurities, carbon coating, and interface degradation in a timely manner, resulting in a certain degree of misjudgment of both high-quality and poor batteries. Its overall prediction accuracy rate is only 78.30%.

[0056] Comparative Example 2 uses only the traditional single electrochemical linear fitting analysis method (dQ / dV) to predict the performance of the lithium iron phosphate battery under test. Although it can reflect the abnormal macroscopic electrochemical performance during the charging and discharging process, it cannot identify the fundamental material defects that lead to performance degradation. When the cathode material has early latent defects such as lattice distortion, trace impurities, and carbon coating defects, its electrochemical response often does not yet show obvious abnormalities. Therefore, the overall prediction accuracy is only improved to a limited extent, reaching only 81.70%, and it cannot locate the root cause of the defects or predict the aging risk points.

[0057] The battery cycle performance prediction method based on multimodal structural characterization and electrochemical dynamic coupling in Example 1 constructs a static multimodal structural feature set using XRD, Raman spectroscopy, and XPS to identify intrinsic defects in the cathode material from multiple dimensions, including crystal structure, carbon coating state, and elemental valence state. Simultaneously, it establishes an electrochemical dynamic feature set by combining multi-SOC AC impedance data, dQ / dV, dV / dQ, and Rct-V impedance-voltage coupling analysis. This yields dynamic characterization information such as interfacial impedance evolution, electrochemical reaction reversibility, electrode phase transition uniformity, interface stability, and electrode interface dynamic evolution rate under different charge states. Furthermore, through a mechanism-driven coupling model of the static multimodal structural feature set and the electrochemical dynamic feature set, it establishes a correlation between microscopic defects and macroscopic electrical performance anomalies, achieving a comprehensive analysis from the intrinsic state of the material to the cycle life evolution process.

[0058] In summary, the battery cycle performance prediction method based on multimodal structural characterization and electrochemical dynamic coupling of the present invention has the advantages of comprehensive characterization dimensions, good experimental repeatability, clear model mechanism, high prediction accuracy, high detection efficiency, and strong engineering applicability. It can accurately identify defects of all dimensions such as lattice distortion, trace impurities, carbon coating defects, and interface degradation, and simultaneously output the estimated cycle life, capacity retention rate, and aging risk points, so that the overall prediction accuracy of the verification sample reaches 96.70%. At the same time, the entire detection process can be completed within one working day, while taking into account the requirements of detection efficiency, latent micro-defect identification capability, model interpretability, and engineering applicability.

[0059] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the patent protection scope of the present invention.

Claims

1. A method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling, characterized in that, Includes the following steps: The battery under test is activated to obtain the activated battery under test. During the activation process, open circuit voltage data and AC impedance spectrum data under different states of charge are collected. Based on the open circuit voltage data and the AC impedance spectrum data under different states of charge, an initial interface-impedance-state three-dimensional baseline of the battery under test is established. Charge-discharge tests were performed on the activated test battery to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, the AC impedance spectroscopy data, and the three-dimensional baseline of the initial interface-impedance-state, an electrochemical dynamic feature set was extracted. The activated battery under test is disassembled, the positive electrode sheet is removed and pre-treated to obtain a pre-treated positive electrode sheet, and the pre-treated positive electrode sheet is subjected to multi-modal structure characterization test to extract the static multi-modal structure feature set. The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term are input into a pre-trained structure-electrochemical coupling mapping model, and the predicted equivalent cycle life of the battery under test is output. The structure-electrochemical coupling term is calculated based on the electrochemical dynamic feature set and the static multimodal structural feature set. The predicted equivalent cycle life of the battery under test is comprehensively checked and corrected with the preset grading judgment rules, and the grade, capacity retention rate, cycle life and aging risk points are output.

2. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, The pre-trained structure-electrochemical coupling mapping model is obtained through the following steps: Multiple training battery samples were activated separately to obtain activated training battery samples. Open circuit voltage data and AC impedance spectrum data under different charging states were collected during the activation process. Charge-discharge tests were performed on the activated training battery samples to collect raw voltage-capacity data. Based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data, features were extracted to obtain electrochemical dynamic feature sets of multiple training battery samples. The activated training battery sample was disassembled, the positive electrode sheet was removed and pre-treated to obtain the pre-treated positive electrode sheet, and the pre-treated positive electrode sheet was tested for multimodal structure characterization to obtain the static multimodal structure feature set of multiple training battery samples. The electrochemical dynamic feature set, the static multimodal structural feature set, and the structure-electrochemical coupling term of the training battery samples are input into a multiple linear regression equation for calculation to obtain the weight coefficients of the static multimodal structural features, the weight coefficients of the electrochemical dynamic features, and the weight coefficients of the structure-electrochemical coupling term of multiple training battery samples. The multiple linear regression equation is Y=w1·S+w2·E+w3·(S×E). The initial weight coefficients of the model are obtained by solving the static multimodal structural feature weight coefficients, electrochemical dynamic feature weight coefficients, and structure-electrochemical coupling term weight coefficients of multiple training battery samples using the least squares iterative method. After calibration, the final weight coefficients w1, w2, and w3 of the model are obtained, thus obtaining the pre-trained structure-electrochemical coupling mapping model.

3. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, In the step of activating the battery to be tested, the activation process includes: The battery under test is placed in a sealed constant temperature and humidity environment and charged with a constant current of A1 until the cutoff voltage is V1. Then it is left to stand. Then it is discharged with a constant current of A2 until the cutoff voltage is V2. Then it is left to stand. Wherein, 0.05A≤A1≤0.15A, 3.0V≤V1≤4.0V, 0.15A≤A2≤0.25A, 2.0V≤V2≤3.0V; Repeat the charging and discharging steps to complete the stepwise activation treatment.

4. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, In the step of collecting open-circuit voltage data and AC impedance spectrum data under different charging states during the activation process, the AC impedance spectrum data under different charging states include AC impedance spectrum data under charging states of 15-25%, 45-55%, and 75-85%.

5. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, In the step of extracting the electrochemical dynamic feature set based on the raw voltage-capacity data, the open-circuit voltage data, and the AC impedance spectroscopy data, the method for extracting the electrochemical dynamic feature set includes the following steps: Capacity differential analysis is performed on the raw voltage-capacity data to obtain the dQ / dV curve. The charge-discharge peak position difference ΔVp, peak symmetry Sp, and main-side reaction peak height ratio Rp are extracted from the dQ / dV curve. Voltage differential analysis is performed on the raw voltage-capacity data to obtain the dV / dQ curve. The curve split degree Ds is extracted from the dV / dQ curve. The dV / dQ curve is compared with the fluctuation value of the three-dimensional baseline of the initial interface-impedance-state to obtain the baseline fluctuation value σb. The corresponding charge transfer impedance Rct is obtained from the AC impedance spectrum data under different charging states. The charge transfer impedance Rct is fitted with the open circuit voltage to construct the Rct-V impedance-voltage coupling curve. The impedance change rate kR and impedance change characteristic voltage VR are extracted from the Rct-V impedance-voltage coupling curve. Obtain the electrochemical dynamic feature set, which is expressed as: E=[ΔVp, Sp, Rp, Ds, σb, kR, VR].

6. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, In the step of disassembling the activated training battery sample, removing the positive electrode sheet, and pre-treating it to obtain the pre-treated positive electrode sheet, the pre-treatment includes the following steps: The positive electrode sheet was soaked in dimethyl carbonate and diethyl carbonate in sequence to obtain a cleaned positive electrode sheet. The cleaned positive electrode sheet is placed in a vacuum oven to dry, thereby obtaining a dried positive electrode sheet. The dried positive electrode sheet is placed in a glove box filled with inert gas and allowed to cool naturally to room temperature to obtain a cooled positive electrode sheet. The cooled positive electrode sheet is cut to obtain a pretreated positive electrode sheet.

7. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, The tests used to characterize the multimodal structure include XRD, Raman spectroscopy, and XPS. The static multimodal structural feature set includes the crystallinity XC of the carbon coating, the proportion of harmful impurities, the shift of XRD diffraction characteristic peaks Δ2θ, and PO4. 3 -Characteristic peak half-width at half-maximum, Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, and F e2+ / Fe 3+ Price-state ratio; Obtain a static multimodal structural feature set, which is represented as: S = [Δ2θ, XC, percentage of harmful impurities, PO4]. 3- Peak half-width at half-maximum (HWHM), Raman carbon peak intensity ratio I1350 / I1600, phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID, Fe 2+ / Fe 3+ ].

8. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 1, characterized in that, The predicted equivalent cycle life value of the battery under test is comprehensively checked and corrected against the preset grading rules, including: When the output of the predicted equivalent cycle life value of the battery under test is displayed as the first set number of cycles, the judgment rule of the first-class high-quality battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the second set number of cycles, the judgment rule of the second-level qualified battery is adopted for judgment. When the output of the predicted equivalent cycle life value of the battery under test is displayed as the third set number of cycles, the judgment rule of the third-level defective battery is adopted.

9. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 8, characterized in that, The criteria for determining the first-class high-quality battery are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously meet the following requirements: crystallinity of crystalline carbon coating XC ≥ 92%, total proportion of harmful impurity phases ≤ 1.2%, XRD diffraction characteristic peak shift Δ2θ ≤ 0.08°, and PO4. 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.32cm. -1 The Raman carbon peak intensity ratio was 0.85-1.1; the phosphate test cell / test cell carbon peak intensity ratio (IPO4 / ID) was 0.85-1.10; Fe... 2+ / Fe 3+ ≤0.1; The electrochemical parameters in the electrochemical dynamic characteristic set must simultaneously satisfy ΔVp≤0.02V, Sp is 0.95-1.05, Rp≤0.08, Ds≤0.1, σb≤0.002, kR≤4% / V, and the impedance change characteristic voltage VR does not show a sudden change inflection point; The structure-electrochemical coupling term must satisfy S×E≤0.35; The criteria for determining whether a battery is qualified at level two are as follows: The structural parameters in the static multimodal structural feature set of the battery under test must simultaneously satisfy the following conditions: 92% > XC ≥ 80%, 1.2% < total proportion of harmful impurity phases ≤ 3.0%, XRD diffraction characteristic peak shift 0.08° < Δ2θ ≤ 0.15°, and 0.32cm. -1 <PO4 3- The characteristic peak half-width at half-maximum (FWHM) is ≤0.55cm. -1 The Raman carbon peak intensity ratio I1350 / I1600 is 0.70-0.85 or 1.10-1.30; the phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID is 0.25-0.4 or 0.6-0.8; 0.1 < Fe 2+ / Fe 3+ ≤0.25; The electrochemical parameters in the electrochemical dynamic characteristic set must meet the following conditions: 0.02V < ΔVp ≤ 0.05V, Sp = 0.90-0.95 or 1.05-1.10, 0.08 < Rp ≤ 0.18, 0.1 < Ds ≤ 0.25, 0.002 < σb ≤ 0.006, 4% / V < kR ≤ 9% / V, and the impedance abrupt change characteristic voltage VR should show a single small abrupt change. The structure-electrochemical coupling term must satisfy: 0.35 < S × E ≤ 0.7; The criteria for determining the Level 3 defective battery are as follows: Among the structural parameters in the static multimodal structural feature set of the battery under test, 80% > XC, the total proportion of harmful impurity phases > 3.0%, the XRD diffraction characteristic peak shift Δ2θ > 0.15°, and PO4 3- Characteristic peak half-width at half-maximum > 0.55 cm -1 Raman carbon peak intensity ratio I1350 / I1600 < 0.70 or > 1.30; phosphate test cell / test cell carbon peak intensity ratio IPO4 / ID < 0.25 or > 0.80; Fe 2+ / Fe 3+ >0.25; The electrochemical parameters in the electrochemical dynamic characteristic set are ΔVp > 0.05V, Sp < 0.90 or > 1.10, Rp > 0.18, Ds > 0.25, σb > 0.006, kR > 9% / V, and the impedance mutation characteristic voltage VR shows multiple obvious abrupt changes. The structure-electrochemical coupling term S×E > 0.

70.

10. The method for predicting battery cycle performance based on multimodal structural characterization and electrochemical dynamic coupling as described in claim 8, characterized in that, Output rating, capacity retention, cycle life, and aging risk factors, including: Outputs a Class 1 high-quality battery, with a capacity retention of ≥85% and a cycle life of ≥2000 cycles at 1C rate, showing no obvious aging defects after 2000 cycles; or, The output is a Class II qualified battery, with a capacity retention of 80-85% after 2000 cycles at 1C rate, and a cycle life of 1500-2000 cycles. A slight interface instability issue exists. Alternatively, The output is a level 3 defective battery, and the output shows that the capacity retention rate is <80% and the cycle life is <1500 cycles after 2000 cycles at 1C rate. Simultaneously, it identifies and outputs aging risk points, which are one or more of the following: lattice distortion, abnormal proportion of harmful impurities, carbon coating failure, abnormal interface impedance, and deterioration of reaction reversibility.

Images