Battery capacity jump-down early warning method, device, equipment and storage medium

Abstract

This invention discloses a method, apparatus, device, and storage medium for early warning of battery capacity drops. The method includes: detecting particulate crack features in a target battery based on acoustic emission signals to generate a crack event rate time series; measuring the electrochemical impedance spectroscopy of the target battery based on a multi-frequency sinusoidal current excitation signal, and determining a crack sensitivity index time series based on the electrochemical impedance spectroscopy; time-series fusion of the crack event rate time series and the crack sensitivity index time series to construct a dual-feature time series matrix; and quantitative identification of the dual-feature time series matrix based on preset crack event rate thresholds and crack sensitivity index thresholds to determine whether the target battery has entered the early stage of capacity drops. This invention can accurately determine whether a target battery has entered the early stage of capacity drops, thereby achieving quantitative characterization and reliable early warning of battery capacity abrupt changes.

Description

Technical Field

[0001] This invention relates to the field of battery energy storage technology, and in particular to a method, device, equipment, and storage medium for early warning of battery capacity drops. Background Technology

[0002] Battery technology involves methods or devices for directly converting chemical energy into electrical energy, such as battery packs. Sudden capacity fade refers to a nonlinear failure mode in which the capacity retention rate of a battery abruptly changes from a normal decay trajectory to a rapid decline during cycling. Existing methods for predicting sudden capacity fade include prediction methods based on electrochemical mechanism models, data-driven feature extraction methods, and sensing and monitoring methods based on single physical fields. However, these methods still suffer from technical problems such as poor real-time performance, delayed warnings, and unclear physical mechanisms, making it difficult to achieve early and reliable warnings of sudden capacity fade in batteries (especially power batteries). Summary of the Invention

[0003] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a method, apparatus, device, and storage medium for early warning of battery capacity drops, which can quantitatively identify the characteristics of battery capacity changes, so as to achieve early and reliable early warning of battery capacity drops.

[0004] In a first aspect, embodiments of the present invention provide a method for early warning of battery capacity drop, comprising: Particle crack feature detection of target battery based on acoustic emission signal to generate crack event rate time series; The electrochemical impedance spectroscopy of the target battery is measured based on a multi-frequency sinusoidal current excitation signal, and the crack sensitivity index time series is determined based on the electrochemical impedance spectroscopy. The crack event rate time series sequence and the crack sensitivity index time series sequence are fused in time series to construct a dual-feature time series matrix; Based on preset crack event rate thresholds and crack sensitivity index thresholds, the dual-feature time series matrix is ​​quantitatively identified to determine whether the target battery has entered the early stage of capacity drop.

[0005] According to some embodiments of the present invention, the step of detecting particle crack features in the target battery based on acoustic emission signals to generate a time-series sequence of crack event rates includes: Acoustic emission signals targeting the target battery are acquired at a preset sampling frequency; The acoustic emission signal is bandpass filtered to extract the signal components located within the characteristic frequency band of the particle crack; Based on the rise time, ring count, and peak frequency of the signal components, particle crack initiation events are identified and their initiation times are recorded to obtain an event sequence. Based on a sliding time window, the number of crack events in the event sequence within a unit time window is counted to generate a crack event rate time series.

[0006] According to some embodiments of the present invention, the step of identifying particle crack initiation events and recording the initiation times of these events based on the rise time, ring count, and peak frequency of the signal components to obtain an event sequence includes: If the rise time of the signal component is less than 10 microseconds, the ring count is greater than 5 times, and the peak frequency is within the frequency range [150, 350] kHz, it is identified as a particle crack initiation event, and the initiation time of the event is recorded to obtain an event sequence.

[0007] According to some embodiments of the present invention, the step of measuring the electrochemical impedance spectroscopy of the target battery based on a multi-frequency sinusoidal current excitation signal and determining the crack sensitivity index time series based on the electrochemical impedance spectroscopy includes: During the current intervals of the charging and discharging process, a multi-frequency sinusoidal current excitation signal is injected into the target battery and the voltage response signal is acquired simultaneously. The electrochemical impedance spectroscopy is determined based on the current excitation signal and the voltage response signal; The high-frequency real impedance and mid-frequency imaginary impedance of the electrochemical impedance spectrum are extracted, and the first rate of change of the real impedance and the second rate of change of the imaginary impedance are determined based on a healthy baseline. The crack sensitivity index time series is obtained by weighted summation of the first rate of change and the second rate of change.

[0008] According to some embodiments of the present invention, the step of quantifying and identifying the dual-feature time series matrix based on a preset crack event rate threshold and a crack sensitivity index threshold to determine whether the target battery has entered the early stage of capacity drop includes: Within multiple consecutive time windows, when the crack event rate of the dual-feature time series matrix exceeds the crack event threshold and the crack sensitivity index exceeds the crack sensitivity index threshold, the target battery is determined to have entered the early stage of capacity drop.

[0009] According to some embodiments of the present invention, the battery capacity drop warning method further includes: The warning confidence level is determined based on the Pearson correlation coefficient between the crack event rate and the crack sensitivity index in the dual-feature time series matrix. Based on the warning confidence level and the preset confidence threshold, graded warnings are issued and graded battery management strategies are implemented.

[0010] According to some embodiments of the present invention, determining the warning confidence level based on the Pearson correlation coefficient between the crack event rate and the crack sensitivity index in the dual-feature time series matrix includes: The Pearson correlation coefficient is determined based on the crack event rate and crack sensitivity index of the dual-feature time series matrix within the sliding time window. The first ratio is determined based on the maximum crack event rate within the sliding time window and the crack event rate threshold. The second ratio is determined based on the maximum value of the crack sensitivity index within the sliding time window and the crack sensitivity index threshold. The smaller value is determined based on the first ratio and the second ratio; The warning confidence level is determined by multiplying the Pearson correlation coefficient by the smaller value.

[0011] Thirdly, embodiments of the present invention provide a battery capacity drop warning device, comprising: The generation module is used to detect particle crack features of the target battery based on acoustic emission signals in order to generate a time series sequence of crack event rates. The determination module is used to measure the electrochemical impedance spectrum of the target battery based on a multi-frequency sinusoidal current excitation signal, and to determine the crack sensitivity index time sequence based on the electrochemical impedance spectrum. The construction module is used to perform time-series fusion of the crack event rate time series sequence and the crack sensitivity index time series sequence to construct a dual-feature time series matrix; The quantization identification module is used to quantize and identify the dual-feature time series matrix based on a preset crack event rate threshold and crack sensitivity index threshold, so as to determine whether the target battery has entered the early stage of capacity drop.

[0012] Thirdly, embodiments of the present invention provide a battery capacity drop warning device, including a processor and a memory, wherein the memory stores a computer program, and the processor runs the computer program to implement the above-mentioned battery capacity drop warning method.

[0013] Fourthly, embodiments of the present invention provide a storage medium storing a computer program, which, when run, implements the aforementioned battery capacity drop warning method.

[0014] The embodiments of the present invention have at least the following beneficial effects: By fusing acoustic emission signals with multi-spectral sinusoidal current excitation signals, the joint sensing of acoustic and electrochemical dual physical fields is achieved. This not only enables the active capture of high-frequency acoustic emission signals to accurately detect particle crack events, but also allows the construction of a dual-feature time series matrix based on the dynamic fusion of time series characteristics. By quantifying and identifying the dual-feature time series matrix, it is possible to accurately determine whether the target battery has entered the early stage of capacity drop, thereby achieving quantitative characterization and reliable early warning of battery capacity mutation characteristics.

[0015] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0016] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is one of the flowcharts of the battery capacity drop warning method according to an embodiment of the present invention; Figure 2 This is a power spectral density variation diagram of a healthy battery according to an embodiment of the present invention; Figure 3 This is a power spectral density change diagram of the early stage of battery capacity drop in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the evolution of the Nyquist plot of electrochemical impedance spectroscopy in an embodiment of the present invention. Figure 5 This is a diagram showing the evolution of the crack event rate according to an embodiment of the present invention. Figure 6 This is a diagram illustrating the evolution of the crack sensitivity index according to an embodiment of the present invention. Figure 7 This is a change diagram of the dual-feature fusion early warning decision in an embodiment of the present invention; Figure 8 This is the second flowchart of the battery capacity drop warning method according to an embodiment of the present invention; Figure 9 This is a scatter plot showing the distribution of early warning confidence and Pearson correlation coefficient in an embodiment of the present invention. Figure 10 This is a schematic diagram of the battery capacity drop warning device according to an embodiment of the present invention; Figure 11 This is a schematic diagram of the battery capacity drop warning device according to an embodiment of the present invention. Detailed Implementation

[0017] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0018] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," "exceeding," etc. are understood to exclude the stated number, and "above," "below," "within," etc. are understood to include the stated number. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the order of the indicated technical features.

[0019] To address the issue of rapidly decreasing battery capacity, various detection methods have been proposed, such as: 1) Prediction methods based on electrochemical mechanism models The capacity decay trajectory is solved by establishing a set of coupled differential equations between particle cracking and electrolyte consumption. This type of method has the following drawbacks: a. Relying on finite element or multiphysics simulation, a single calculation takes several minutes to several hours, making it impossible to embed into the vehicle battery management system for real-time execution; b. Model parameters (such as diffusion coefficient, reaction rate constant, etc.) drift with battery aging, resulting in decreased long-term prediction accuracy; c. Unable to distinguish the temporal contributions of different physical failure mechanisms.

[0020] 2) Data-driven feature extraction methods Early warning is provided by monitoring the characteristic peak shift of the charging voltage curve (i.e., the dQ / dV curve) or the capacity-internal resistance mapping relationship. This type of method has the following drawbacks: a. It requires accumulating historical data from hundreds of cycles to establish a statistical baseline, resulting in a lag in early warning of sudden failures; b. The voltage curve characteristics are not sensitive enough to capacity drops, and usually only show significant changes when the capacity loss is >5%; c. It belongs to passive monitoring and cannot locate the physical location where the failure started.

[0021] 3) Sensing and monitoring methods based on a single physical field Existing acoustic emission technology is used for early warning of battery thermal runaway by monitoring high-amplitude acoustic emission signals (frequency <100kHz) generated by electrolyte vaporization to identify precursors to thermal runaway. This method has the following drawbacks: a. The monitoring target is thermal runaway (violent reaction at temperatures >80°C), while particle cracks are found in very mild cycles; b. No correlation was established with electrochemical parameters, making it impossible to quantify the extent of the impact of cracks on electrochemical performance; c. Lack of exploration into the temporal evolution of acoustic emission signals.

[0022] Therefore, existing technologies suffer from technical defects such as poor real-time performance, delayed early warning, and unclear physical mechanisms.

[0023] Please refer to Figure 1 This embodiment discloses a method for early warning of battery capacity drops, including steps S100 to S400. It should be noted that the numbering of the steps in this embodiment is only for ease of review and understanding, and not to limit the execution order of the steps. The content of each step is described in detail below: S100. Based on acoustic emission signals, particle crack features are detected in the target battery to generate a time series sequence of crack event rates. For example, a sudden drop in battery capacity is typically caused by rapid electrolyte consumption and solid-liquid interface deterioration due to particle cracking in the negative electrode active material. During this process, the instantaneous generation of elastic waves due to stress release in the material is known as acoustic emission (AE), or stress wave emission. These transient elastic waves carry crucial information about the evolution of internal defects in the material and are ideal physical quantities for capturing early events of particle cracking.

[0024] To non-invasively acquire the internal state of a battery, acoustic emission sensors are placed on the surface of the target battery's casing. The working principle of these sensors is as follows: when the active material particles inside the battery deform, experience lattice slip, or crack propagate, the released strain energy propagates to the battery casing surface in the form of elastic waves, causing minute vibrations. The acoustic emission sensors capture these weak mechanical vibration signals and linearly convert them into processable electrical signals (i.e., acoustic emission signals). By continuously acquiring and processing the acoustic emission signals in real time at a preset sampling frequency, crack initiation events (referred to as crack events) can be identified. Since the acquisition and processing of acoustic emission signals are strictly synchronous and real-time, each identified crack event has a precise timestamp. The number of effective crack events detected within a unit time window is defined as the "crack event rate," which characterizes the number of effective crack events within a unit time window. As the target battery continues to operate, the crack event rates within consecutive time windows are arranged sequentially according to time, generating a crack event rate time series.

[0025] S200, based on multi-frequency sinusoidal current excitation signal, measures the electrochemical impedance spectrum of the target battery, and determines the crack sensitivity index time series based on the electrochemical impedance spectrum; For example, electrochemical impedance spectroscopy (EIS), also known as AC impedance spectroscopy, is a frequency domain analysis technique that allows for in-depth, non-destructive exploration of the microscopic processes within an electrochemical system. The basic principle involves applying a variable-frequency, small-amplitude (typically a few millivolts to tens of millivolts) sinusoidal AC current signal as a perturbation to an electrochemical system (such as a target battery) in equilibrium or steady state, while simultaneously measuring the voltage response signal output by the system. By calculating the impedance of the electrochemical system as a function of frequency (i.e., the impedance spectrum), the structure of the electrode / electrolyte interface, electrochemical reaction kinetics, and various physical and electrochemical processes within the material can be studied. In this embodiment, a crack sensitivity index is used to characterize the potential probability and activity level of particle cracks within the battery. During the current gap in the charging and discharging process of the target battery, the electrochemical impedance spectrum of the target battery is measured based on a multi-frequency sinusoidal current excitation signal. The crack sensitivity index time series is then determined based on the electrochemical impedance spectrum, enabling real-time sensing and probability assessment of crack initiation and propagation events within the battery from the perspective of electrochemical response.

[0026] S300. The crack event rate time series and the crack sensitivity index time series are fused together to construct a dual-feature time series matrix. For example, both the crack event rate time series and the crack sensitivity index time series are time series detected during the use of the target battery. The crack event rate time series is obtained through continuous sampling, while the crack sensitivity index time series is obtained through intermittent sampling. The original data of the two are often not synchronized on the time axis. To enable subsequent quantitative analysis of the two time series, they are aligned along the time axis to achieve time series fusion, constructing a dual-feature time series matrix to achieve joint sensing of the acoustic and electrochemical fields.

[0027] S400 quantifies and identifies the dual-feature time series matrix based on preset crack event rate threshold and crack sensitivity index threshold to determine whether the target battery has entered the early stage of capacity drop.

[0028] For example, the crack event rate threshold and the crack sensitivity index threshold are thresholds determined based on historical experience. Based on the preset crack event rate threshold and crack sensitivity index threshold, the dual-feature time series matrix is ​​quantitatively identified, which can determine whether the target battery has entered the early stage of capacity drop.

[0029] Thus, by fusing acoustic emission signals with multi-spectral sinusoidal current excitation signals, the joint sensing of acoustic and electrochemical dual physical fields can be achieved. This not only enables the active capture of high-frequency acoustic emission signals to accurately detect particle crack events, but also allows the construction of a dual-feature time-series matrix based on the dynamic fusion of time-series characteristics. By quantifying and identifying the dual-feature time-series matrix, it is possible to accurately determine whether the target battery has entered the early stage of capacity drop, thereby achieving quantitative characterization and reliable early warning of battery capacity mutation characteristics.

[0030] In some application examples, step S100, which involves detecting particle crack features in the target battery based on acoustic emission signals to generate a time-series sequence of crack event rates, includes: Acoustic emission signals targeting the battery are acquired at a preset sampling frequency; Bandpass filtering is applied to the acoustic emission signal to extract signal components located within the characteristic frequency range of particle cracks; Based on the rise time, ring count, and peak frequency of the signal components, particle crack initiation events are identified and their initiation times are recorded to obtain an event sequence. Based on a sliding time window, the number of crack events in the event sequence within a unit time window is counted to generate a crack event rate time series.

[0031] For example, an acoustic emission sensor is arranged on the surface of the target battery casing to collect acoustic emission signals in real time at a preset sampling frequency. The acoustic emission signals are then bandpass filtered to extract signal components within the characteristic frequency range of particle cracks. By analyzing these signal components, particle crack initiation events are identified, and the initiation times of these events are recorded to obtain an event sequence. A unit time window is set, denoted as Δt, and the number of crack events within each unit time window is counted, denoted as N. crack (t, t+Δt), and slide the unit time window along the time axis to count the number of crack events in the event sequence along the time axis. That is, based on the sliding time window, count the number of crack events in the event sequence within the unit time window. During the statistical process, a moving average filter can be applied to the number of crack events in multiple consecutive time windows to suppress random noise and perform normalization processing to generate a crack event rate time series. For example, a moving average filter can be applied to the number of crack events in M=3 consecutive time windows to obtain the filtered average number of crack events, denoted as: Normalizing the moving average sequence to the interval between 0 and 1 yields the crack event rate time series, denoted as: Where, N maxThe normalized benchmark is determined based on the statistical distribution of batteries in the same batch, preferably twice the 95th percentile of the healthy baseline period, or a preset absolute threshold of 100.

[0032] The preset sampling frequency is 1-10MHz, preferably 2MHz; the bandpass filter frequency band is 100-500kHz, preferably 150-350kHz; the identification criteria for particle crack initiation events are: rise time <10μs, ring count >5, and peak frequency in the range of 150~350kHz; the unit time window can be set to 1~60 seconds, preferably 10 seconds.

[0033] In a specific application example, the above steps—based on the rise time, ring count, and peak frequency of the signal components—identify particle crack initiation events and record the initiation times of these events to obtain an event sequence, including: If the rise time of the signal component is less than 10 microseconds, the ring count is greater than 5 times, and the peak frequency is within the frequency range [150, 350] kHz, it is identified as a particle crack initiation event, and the initiation time of the event is recorded to obtain the event sequence.

[0034] For example, although acoustic emission technology has been applied in the field of battery thermal runaway early warning, existing technologies generally suffer from limitations in frequency band selection and mechanism understanding. Specifically, existing solutions mostly focus on monitoring low-frequency acoustic emission signals (usually signals with frequencies below 100 kHz), which mainly correspond to late-stage thermal runaway characteristics such as gas evolution inside the battery, electrolyte boiling, or macroscopic structural deformation. However, existing technologies have not yet revealed the application value of high-frequency acoustic emission signals in battery micro-damage detection, nor have they established a clear mapping relationship between high-frequency signal characteristics and the microscopic physical event of particle crack initiation in the negative electrode active material. Due to the lack of in-depth exploration of the correlation between the spectral characteristics of acoustic emission signals and the evolution law of battery capacity drop, existing technologies have significant blind spots in their ability to perceive the early stages of capacity drop, making it difficult to achieve effective early warning during the micro-damage accumulation period before thermal runaway occurs.

[0035] This embodiment, based on the microscopic initiation mechanism of battery capacity drop—namely, rapid electrolyte consumption and solid-liquid interface deterioration caused by particle cracking in the negative electrode active material—discovered a significant and strong correlation between acoustic emission signals in specific high-frequency bands (e.g., [150, 350] kHz) and particle crack initiation events through extensive experimental research and spectral analysis. By combining the rise time of signal components, ring count, and the frequency band of the peak frequency for multidimensional analysis, particle crack initiation events can be effectively identified, thereby obtaining the event sequence.

[0036] like Figure 2 As shown, healthy batteries show no significant signal in the 150~350kHz characteristic frequency band, while... Figure 3 As shown, before the drop, the battery exhibited a dense cluster of crack events in this frequency band, manifested as multiple high-amplitude peaks. Figure 2 and Figure 3 The study clearly demonstrated the characteristics of elastic waves released during the initiation of granular cracks, proving the sensitivity of high-frequency acoustic emission signals to microcracks.

[0037] In some application examples, step S200, measuring the electrochemical impedance spectroscopy of the target cell based on a multi-frequency sinusoidal current excitation signal, and determining the crack sensitivity index time series based on the electrochemical impedance spectroscopy, includes: During the current intervals of the charging and discharging process, a multi-frequency sinusoidal current excitation signal is injected into the target battery and the voltage response signal is acquired simultaneously. Electrochemical impedance spectroscopy was determined based on current excitation signal and voltage response signal; The real part impedance in the high-frequency band and the imaginary part impedance in the mid-frequency band of the electrochemical impedance spectrum were extracted, and the first rate of change of the real part impedance and the second rate of change of the imaginary part impedance were determined based on the healthy baseline. The crack sensitivity index time series is obtained by weighted summation of the first and second rates of change.

[0038] For example, the current interval is the resting period during the charging or discharging process when the absolute value of the current is <0.05C (coulomb); the frequency range of the multi-frequency sinusoidal current excitation signal is 1mHz-10kHz, containing at least 10 logarithmically distributed frequency points; the amplitude of the current excitation signal is <5mA to avoid polarization interference; the calculation of the electrochemical impedance spectrum adopts the fast Fourier transform algorithm, and the single measurement time is <2 seconds.

[0039] Please refer to Figure 4 , Figure 4 The evolution of the Nyquist plot of the electrochemical impedance spectroscopy (EIS) shown illustrates the evolution of the EIS with cycling, demonstrating the sensitivity of the high-frequency real part and the mid-frequency imaginary part to cracks. As cycling progresses, the Nyquist plot exhibits a regular evolution: the high-frequency intercept (horizontal axis) shifts to the right, reflecting the increase in contact resistance caused by particle cracks; the mid-frequency semicircle diameter expands, reflecting the increase in charge transfer impedance. The impedance spectra of healthy cells (green curve, 100 cycles), cells in the warning period (orange dashed line, 300 cycles), and cells in the critical period (red dotted line, 500 cycles) show significant differences, demonstrating the sensitivity of the EIS to capacity drops.

[0040] The real part of the impedance in the high-frequency band is the average real part R of the impedance at frequencies >1kHz. HF The imaginary impedance in the mid-frequency band is the average value of the imaginary impedance from 10 to 100 Hz. MF The difference in the real part of the impedance is denoted as ΔR. HF The difference in the imaginary impedance is denoted as ΔX. MFThe rate of change is calculated as the relative change with respect to the battery's factory condition or its health status within the first 100 cycles, where R... HF,0 and X MF,0 The healthy baseline values ​​for the real and imaginary impedances are ΔR, respectively. HF / R HF,0 Let ΔX be the first rate of change. MF / X MF,0 The second rate of change; crack sensitivity index (denoted as Z). crack The formula for calculating ) is: Z crack = w1 (ΔR HF / R HF,0 ) + w2 (ΔX MF / X MF,0 ) Where w1 and w2 are weighting coefficients, satisfying w1 + w2 = 1, and for the graphite anode system, w1 = 0.6 and w2 = 0.4 are preferred.

[0041] In some application examples, step S400, based on preset crack event rate thresholds and crack sensitivity index thresholds, quantifies and identifies the dual-feature time series matrix to determine whether the target battery has entered the early stage of capacity degradation, including: Within multiple consecutive time windows, when the crack event rate of the dual-feature time series matrix exceeds the crack event threshold and the crack sensitivity index exceeds the crack sensitivity index threshold, the target battery is determined to have entered the early stage of capacity drop.

[0042] For example, using the acoustic emission data acquisition timestamp as a reference, the crack sensitivity index is interpolated to the same time resolution to construct a dual-feature time series matrix. The dual-feature time series matrix is ​​a two-dimensional time series, with the first dimension being the crack event rate R(t) and the second dimension being the crack sensitivity index Z(t).

[0043] Please refer to Figures 5 to 7 , Figure 5 The evolution trajectory of the crack event rate R(t) is shown. Figure 6 The evolution trajectory of the crack sensitivity index Z(t) is shown. Figure 7 The results of the joint determination based on two features are shown. Figure 5 The evolution trajectory of the crack event rate R(t) begins to exceed the threshold R around cycle 280. th =0.35, but Figure 6 The evolution trajectory of the crack sensitivity index Z(t) did not exceed the threshold Z. th =0.18, at which point no alarm was triggered; around loop 300, Figure 6 The evolution trajectory of the crack sensitivity index Z(t) begins to exceed the threshold Z. th=0.18, at which point the evolution trajectory of both the crack event rate R(t) and the crack sensitivity index Z(t) exceeds the corresponding threshold; when the double threshold condition is met simultaneously for 5 consecutive windows (L=5), a yellow warning is triggered (cycle 320); around cycle 350, the evolution trajectory of the crack sensitivity index Z(t) exceeds twice the crack sensitivity index threshold, and after another 5 consecutive windows of detection, a red warning is triggered (cycle 380).

[0044] The early warning judgment logic is as follows: Set a sliding time window of length L (preferably L=5 time points); when all time points within the window satisfy R(t)>R th (Crack event threshold) and Z(t)>Z th When the crack sensitivity index threshold is reached, an early warning is triggered; where R... th and Z th To determine the threshold based on the statistical distribution of batteries in the same batch, R is preferred. th =0.3 (normalized value), Z th =0.15.

[0045] Please refer to Figure 8 In some application examples, battery capacity drop warning methods also include: The Pearson correlation coefficient between the crack event rate and the crack sensitivity index in the dual-feature time series matrix is ​​used to determine the early warning confidence level. Based on the warning confidence level and the preset confidence threshold, graded warnings are issued and graded battery management strategies are implemented.

[0046] For example, the Pearson correlation coefficient, also known as the Pearson product-moment correlation coefficient (PPMCC or PCCs), is used to measure the correlation (linear correlation) between two variables (such as crack event rate and crack sensitivity index). The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation of the two variables. Based on the Pearson correlation coefficient, the warning confidence level is determined, and then, according to the warning confidence level and confidence threshold, tiered warnings are issued and tiered battery management strategies are implemented, effectively reducing the false alarm rate.

[0047] In a specific application example, the above steps, based on the Pearson correlation coefficient between the crack event rate and the crack sensitivity index in the dual-feature time series matrix, determine the warning confidence level, including: The Pearson correlation coefficient is determined based on the crack event rate and crack sensitivity index of the dual-feature time series matrix within the sliding time window. The first ratio is determined based on the maximum crack event rate and the crack event rate threshold within the sliding time window; The second ratio is determined based on the maximum value of the crack sensitivity index and the crack sensitivity index threshold within the sliding time window; The smaller value is determined based on the first ratio and the second ratio; The confidence level of the warning is determined by multiplying the Pearson correlation coefficient by the smaller value.

[0048] For example, calculate the Pearson correlation coefficient ρ between the crack event rate R(t) and the crack sensitivity index Z(t) within the sliding window; based on the maximum crack event rate (R) within the sliding time window... max ) and crack event rate threshold (R) th Determine the first ratio, i.e., R. max / R th Based on the maximum crack sensitivity index (Z) within the sliding time window max ) and crack sensitivity index threshold (Z) th ), determine the second ratio, i.e., Z max / Z th Based on the first ratio and the second ratio, determine the smaller value, denoted as min(R). max / R th Z max / Z th If the warning confidence level is C = ρ·min(R), then the warning confidence level is C = ρ·min(R). max / R th Z max / Z th In one specific application example, the confidence threshold is configured as 0.7.

[0049] Please refer to Figure 9 , Figure 9 The distribution of the warning confidence-Pearson correlation coefficient for a large number of samples is shown in a scatter plot. The healthy status (green dots), noise (gray triangles), yellow warning (orange diamonds), and red warning (red asterisks) show clear partitions, which proves the effectiveness of dual feature fusion.

[0050] The tiered early warning system includes the following levels: When C < 0.7, no warning is triggered; When 0.7 ≤ C < 0.9, a yellow warning is triggered, and it is recommended to limit the charge / discharge rate to 0.5C. When C ≥ 0.9, a red alert is triggered, and it is recommended to prohibit fast charging and arrange for maintenance.

[0051] The technical effects of this embodiment are illustrated below based on comparative experimental data: For the power battery pack (100Ah, 1P96S), acoustic emission sensors were arranged on the large surface of each cell, with a total of 96 measurement points; the sampling rate was set to 2MHz for continuous acquisition; real-time filtering was used to extract the 150~350kHz frequency band; the crack event rate was calculated every 10 seconds, and the moving average window was 3 time points; the normalized baseline was determined to be the 95th percentile of the first 100 cycles.

[0052] Utilizing the existing charge / discharge circuitry of the BMS (Battery Management System), inject excitation during the resting period (current <5A) within the 50%-80% SOC (State of Charge) range; frequency selection: 10mHz, 50mHz, 100mHz, 500mHz, 1Hz, 10Hz, 100Hz, 1kHz, 5kHz, 10kHz; current amplitude 3mA, duration 1.5 seconds; calculate Z... crack = 0.6 (ΔR HF / R HF,0 ) + 0.4 (ΔX MF / X MF,0 ).

[0053] Time alignment: Based on the acoustic emission data acquisition timestamp, the crack sensitivity index is linearly interpolated to the same time resolution; sliding window length L=5 (corresponding to 50 seconds); threshold setting: crack event rate threshold R. th =0.35, crack sensitivity index threshold Z th =0.18 (based on statistics of 10 batteries in the same batch).

[0054] Calculate the Pearson correlation coefficient ρ and the warning confidence level C; during a yellow warning, the BMS will limit fast charging requests to 0.5C; during a red warning, the BMS will disable fast charging and illuminate the instrument panel malfunction indicator light.

[0055] Experimental verification: Control group: Batteries from the same batch that did not use this method experienced a significant drop in capacity after 450 cycles (capacity retention rate plummeted to 82%). Experimental group: Using this method, a yellow alert was triggered at 320 cycles (capacity retention rate 97%), and a red alert was triggered at 380 cycles (capacity retention rate 91%), with alerts issued 70-130 cycles in advance.

[0056] In addition, several comparison schemes are provided. The main characteristics of the comparison schemes are shown in Table 1 below. The test conditions are: NCM523 / graphite soft-pack battery, ambient temperature during the test: 45℃, charge / discharge rate: 2C / 2C charge / discharge, cutoff voltage range during the charge / discharge process: 3.0~4.2V. The main evaluation indicators of the experimental results are shown in Table 2. Table 1 Table 2 The AE-EIS time series fusion and dual-threshold determination are used to represent the time series fusion of the crack event rate time series obtained by acoustic emission (AE) processing and the crack sensitivity index time series obtained by electrochemical impedance spectroscopy (EIS) processing in this embodiment. The dual-feature time series matrix is ​​then quantified and identified based on the crack event rate threshold and the crack sensitivity index threshold. As shown in the comparison results in Table 2, this embodiment can cycle 120 times in advance when issuing an early warning, with a capacity retention rate of 97% at the time of warning, only 2 false alarms, and 0 missed alarms, achieving early and reliable early warning when capacity loss is <3%.

[0057] Please refer to Figure 10 Based on the above technical concept, this embodiment also provides a battery capacity drop warning device, including: The generation module 110 is used to perform particle crack feature detection on the target battery based on acoustic emission signals in order to generate a crack event rate time series. The determination module 120 is used to measure the electrochemical impedance spectrum of the target battery based on the multi-frequency sinusoidal current excitation signal, and to determine the crack sensitivity index time sequence based on the electrochemical impedance spectrum. Module 130 is used to perform time-series fusion of the crack event rate time series and the crack sensitivity index time series to construct a dual-feature time series matrix. The quantization identification module 140 is used to quantize and identify the dual-feature time series matrix based on the preset crack event rate threshold and crack sensitivity index threshold, so as to determine whether the target battery has entered the early stage of capacity drop.

[0058] The inventive concept of this battery capacity drop warning device embodiment is the same as that of the battery capacity drop warning method embodiment described above. Content not covered in this battery capacity drop warning device embodiment can be referred to in the battery capacity drop warning method embodiment described above, and will not be repeated here. By fusing acoustic emission signals and multi-spectral sinusoidal current excitation signals, joint sensing of acoustic and electrochemical dual-physical fields is achieved. This not only actively captures high-frequency acoustic emission signals to accurately detect particle crack events, but also constructs a dual-feature time-series matrix based on the dynamic fusion of time-series characteristics. By quantifying and identifying the dual-feature time-series matrix, it is possible to accurately determine whether the target battery has entered the early stage of capacity drop, thereby achieving quantitative characterization and reliable early warning of battery capacity mutation characteristics.

[0059] Please refer to Figure 11This embodiment also provides a battery capacity drop warning device, including a processor 210 and a memory 220. The memory 220 stores a computer program, and the processor 210 executes the computer program to implement the aforementioned battery capacity drop warning method. The specific details of the battery capacity drop warning method are described above and will not be repeated here. By fusing acoustic emission signals and multi-spectral sinusoidal current excitation signals, joint sensing of acoustic and electrochemical dual-physical fields is achieved. This not only actively captures high-frequency acoustic emission signals to accurately detect particle crack events, but also constructs a dual-feature time-series matrix based on the dynamic fusion of time-series characteristics. By quantifying and identifying the dual-feature time-series matrix, it is possible to accurately determine whether the target battery has entered the early stage of capacity drop, thereby achieving quantitative characterization and reliable early warning of battery capacity change characteristics.

[0060] This embodiment also provides a storage medium storing a computer program. When the computer program is run, it implements the aforementioned battery capacity drop warning method. The specific details of the battery capacity drop warning method are described above and will not be repeated here. By fusing acoustic emission signals and multi-spectral sinusoidal current excitation signals, joint sensing of acoustic and electrochemical dual-physical fields is achieved. This not only actively captures high-frequency acoustic emission signals to accurately detect particle crack events, but also constructs a dual-feature time-series matrix based on the dynamic fusion of time-series characteristics. By quantifying and identifying the dual-feature time-series matrix, it is possible to accurately determine whether the target battery has entered the early stage of capacity drop, thereby achieving quantitative characterization and reliable early warning of battery capacity abrupt changes.

[0061] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for early warning of battery capacity drop, characterized in that, include: Particle crack feature detection of target battery based on acoustic emission signal to generate crack event rate time series; The electrochemical impedance spectroscopy of the target battery is measured based on a multi-frequency sinusoidal current excitation signal, and the crack sensitivity index time series is determined based on the electrochemical impedance spectroscopy. The crack event rate time series sequence and the crack sensitivity index time series sequence are fused in time series to construct a dual-feature time series matrix; Based on preset crack event rate threshold and crack sensitivity index threshold, the dual-feature time series matrix is ​​quantitatively identified to determine whether the target battery has entered the early stage of capacity drop. The step of detecting particle crack features in the target battery based on acoustic emission signals to generate a time-series sequence of crack event rates includes: Acoustic emission signals targeting the target battery are acquired at a preset sampling frequency; The acoustic emission signal is bandpass filtered to extract the signal components located within the characteristic frequency band of the particle crack; Based on the rise time, ring count, and peak frequency of the signal components, particle crack initiation events are identified and their initiation times are recorded to obtain an event sequence. Based on a sliding time window, the number of crack events in the event sequence within a unit time window is counted to generate a crack event rate time series. The method of measuring the electrochemical impedance spectroscopy of the target battery based on a multi-frequency sinusoidal current excitation signal, and determining the crack sensitivity index time series based on the electrochemical impedance spectroscopy, includes: During the current intervals of the charging and discharging process, a multi-frequency sinusoidal current excitation signal is injected into the target battery and the voltage response signal is acquired simultaneously. The electrochemical impedance spectroscopy is determined based on the current excitation signal and the voltage response signal; The high-frequency real impedance and mid-frequency imaginary impedance of the electrochemical impedance spectrum are extracted, and the first rate of change of the real impedance and the second rate of change of the imaginary impedance are determined based on a healthy baseline. The crack sensitivity index time series is obtained by weighted summation of the first rate of change and the second rate of change.

2. The battery capacity drop warning method according to claim 1, characterized in that, The process of identifying particle crack initiation events and recording their initiation times based on the rise time, ring count, and peak frequency of the signal components yields an event sequence, including: If the rise time of the signal component is less than 10 microseconds, the ring count is greater than 5 times, and the peak frequency is within the frequency range [150, 350] kHz, it is identified as a particle crack initiation event, and the initiation time of the event is recorded to obtain an event sequence.

3. The battery capacity drop warning method according to claim 1, characterized in that, The method of quantifying and identifying the dual-feature time series matrix based on preset crack event rate thresholds and crack sensitivity index thresholds to determine whether the target battery has entered the early stage of capacity degradation includes: Within multiple consecutive time windows, when the crack event rate of the dual-feature time series matrix exceeds the crack event rate threshold and the crack sensitivity index exceeds the crack sensitivity index threshold, the target battery is determined to have entered the early stage of capacity drop.

4. The battery capacity drop warning method according to any one of claims 1 to 3, characterized in that, The battery capacity drop warning method also includes: The warning confidence level is determined based on the Pearson correlation coefficient between the crack event rate and the crack sensitivity index in the dual-feature time series matrix. Based on the warning confidence level and the preset confidence threshold, graded warnings are issued and graded battery management strategies are implemented.

5. The battery capacity drop warning method according to claim 4, characterized in that, The determination of the warning confidence level based on the Pearson correlation coefficient of the crack event rate and crack sensitivity index in the dual-feature time series matrix includes: The Pearson correlation coefficient is determined based on the crack event rate and crack sensitivity index of the dual-feature time series matrix within the sliding time window. The first ratio is determined based on the maximum crack event rate within the sliding time window and the crack event rate threshold. The second ratio is determined based on the maximum value of the crack sensitivity index within the sliding time window and the crack sensitivity index threshold. The smaller value is determined based on the first ratio and the second ratio; The warning confidence level is determined by multiplying the Pearson correlation coefficient by the smaller value.

6. A battery capacity drop warning device, characterized in that, include: The generation module is used to detect particle crack features of the target battery based on acoustic emission signals in order to generate a time series sequence of crack event rates. The determination module is used to measure the electrochemical impedance spectrum of the target battery based on a multi-frequency sinusoidal current excitation signal, and to determine the crack sensitivity index time sequence based on the electrochemical impedance spectrum. The construction module is used to perform time-series fusion of the crack event rate time series sequence and the crack sensitivity index time series sequence to construct a dual-feature time series matrix; The quantization identification module is used to quantize and identify the dual-feature time series matrix based on a preset crack event rate threshold and crack sensitivity index threshold, so as to determine whether the target battery has entered the early stage of capacity drop. The step of detecting particle crack features in the target battery based on acoustic emission signals to generate a time-series sequence of crack event rates includes: Acoustic emission signals targeting the target battery are acquired at a preset sampling frequency; The acoustic emission signal is bandpass filtered to extract the signal components located within the characteristic frequency band of the particle crack; Based on the rise time, ring count, and peak frequency of the signal components, particle crack initiation events are identified and their initiation times are recorded to obtain an event sequence. Based on a sliding time window, the number of crack events in the event sequence within a unit time window is counted to generate a crack event rate time series. The method of measuring the electrochemical impedance spectroscopy of the target battery based on a multi-frequency sinusoidal current excitation signal, and determining the crack sensitivity index time series based on the electrochemical impedance spectroscopy, includes: During the current intervals of the charging and discharging process, a multi-frequency sinusoidal current excitation signal is injected into the target battery and the voltage response signal is acquired simultaneously. The electrochemical impedance spectroscopy is determined based on the current excitation signal and the voltage response signal; The high-frequency real impedance and mid-frequency imaginary impedance of the electrochemical impedance spectrum are extracted, and the first rate of change of the real impedance and the second rate of change of the imaginary impedance are determined based on a healthy baseline. The crack sensitivity index time series is obtained by weighted summation of the first rate of change and the second rate of change.

7. A battery capacity drop warning device, comprising a processor and a memory, wherein the memory stores a computer program, characterized in that, When the processor runs the computer program, it is used to implement the battery capacity drop warning method as described in any one of claims 1 to 5.

8. A storage medium storing a computer program, characterized in that, When the computer program is run, it implements the battery capacity drop warning method as described in any one of claims 1 to 5.

Images