A dynamic EIS, pressure signal and electric signal combined liquid leakage battery diagnosis method

Abstract

The present application relates to a kind of dynamic EIS, pressure signal and electric signal combined liquid leakage battery diagnosis method, three kinds of signals are collected in cycle four stages, extract EIS impedance amplitude, phase, change and Nyquist area;Extract pressure rate, extreme value, amplitude and slope, extract electric signal voltage and power, offset rate and slope, with health benchmark standardization, weighted abnormal index, compare threshold, determine health, suspected, confirm liquid leakage or other failure.The dynamic EIS, pressure signal and electric signal combined liquid leakage battery diagnosis method, realize early warning and whole process monitoring of liquid leakage, dynamic EIS can capture early electrochemical anomaly of leakage, combined with pressure signal to carry out physical verification, effectively reduce the risk of misjudgment caused by single signal interference, introduce electric signal to distinguish liquid leakage and other non-liquid leakage failure accurately, establish health benchmark and adaptive weight system, significantly improve the reliability and robustness of diagnosis result.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery fault detection technology, specifically to a method for diagnosing leaking batteries that combines dynamic EIS, pressure signals, and electrical signals. Background Technology

[0002] Lithium-ion battery fault detection collects operating parameters such as cell voltage, circuit current, and casing temperature. Combined with offline internal resistance testing, capacity calibration, and online parameter identification methods, it monitors characteristic quantities such as single-cell voltage difference, internal resistance change, voltage decay rate, and temperature rise rate. It also incorporates casing leakage sensing to detect electrolyte leakage, thereby identifying typical faults such as cell inconsistency degradation, capacity decay, abnormal self-discharge, internal micro-short circuits, poor terminal contact, and electrolyte leakage. Based on threshold judgment and time-series data trend analysis, and combined with equivalent circuit models or intelligent algorithms, it estimates the battery's state of harm (SOH), distinguishes between battery-related faults and BMS sampling anomalies, and achieves fault early warning and failure identification to prevent the risk of thermal runaway.

[0003] Currently, there are three main methods for diagnosing battery leakage faults. The first is the traditional method based on threshold judgment, which judges battery faults based on abnormal alarms such as voltage, temperature, and current. The second is the method based on static electrochemical impedance spectroscopy (EIS) analysis, which judges battery faults by identifying SEI film thickening and charge transfer impedance increase through relaxation time distribution (DRT). The third is the chemical monitoring method based on gas signal changes, which uses gas sensors to identify the gas components and gaseous chemical substances leaked from the leaking battery to provide early warning and judge battery faults. In practical use, the three methods mentioned above have problems such as signal lag, confusing mechanism and weak anti-interference ability when dealing with the developing fault of electrolyte leakage. For example, traditional threshold-based methods struggle to capture early signs of electrolyte leakage, failing to provide early warnings. Abnormalities such as voltage or current only manifest in the later stages of leakage, leading to delayed alarms. Static electrochemical impedance spectroscopy (EIS) analysis suffers from poor model parameter identification due to operational conditions, insufficient specificity, and a high false alarm rate. EIS features are easily confused with normal aging and other faults, resulting in misdiagnosis of battery malfunctions. Chemical monitoring methods based on gas signal changes exhibit characteristic peak changes similar to various battery degradation modes, leading to confusion. Furthermore, static EIS information cannot reflect the entire battery leakage process and is highly susceptible to factors such as battery charging / discharging expansion and heat generation, resulting in poor reliability. Therefore, a method for diagnosing leaking batteries that combines dynamic EIS, pressure signals, and electrical signals is proposed to solve the above problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for diagnosing leaking batteries that combines dynamic EIS, pressure signals, and electrical signals. This method overcomes the limitations of relying solely on pressure or electrical signals, which are susceptible to interference from battery expansion and heat generation. It enables early, comprehensive, and reliable fault diagnosis of battery leaks. This solves the problems of traditional threshold-based methods, which struggle to capture early signs of electrolyte leakage, fail to provide early warnings, and whose voltage or current anomalies only manifest in the later stages of leakage, leading to delayed alarms. Furthermore, methods based on static electrochemical impedance spectroscopy (EIS) analysis suffer from model parameter identification issues due to susceptibility to operating conditions, insufficient specificity, and a high false positive rate. EIS features are also easily confused with normal aging and other faults, leading to misdiagnosis. Finally, chemical monitoring methods based on gas signal changes suffer from characteristic peak changes that resemble various battery degradation modes, causing confusion. Static EIS information cannot reflect the entire leakage process and is highly susceptible to factors such as battery charging / discharging expansion and heat generation, resulting in poor reliability.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals, comprising the following steps: S1. Divide the single cycle process of the soft-pack lithium-ion battery into four stages, and collect the pressure signal data, electrical signal data and dynamic EIS signal data of the soft-pack lithium-ion battery under test at each stage and perform preprocessing. S2. Based on the dynamic EIS signal data of the soft-pack lithium-ion battery under test, the full-band complex impedance spectrum is obtained, and the basic impedance feature set is constructed; The statistical characteristics, impedance change characteristics, and curve morphology characteristics were calculated using the basic impedance characteristic set, and a dynamic EIS leakage sensitivity characteristic set was constructed. S3. Based on the pressure signal data of the soft-pack lithium-ion battery under test, calculate the stage pressure change characteristics, cycle pressure characteristics, and pressure cycle evolution characteristics, and construct a pressure leakage sensitive feature set; S4. Based on the electrical signal data of the soft-pack lithium-ion battery to be tested, calculate the electrical signal characteristics, electrical signal offset characteristics, and electrical signal cyclic evolution characteristics, and construct a set of leakage-sensitive features of electrical signals; S5. Based on the pressure signal data, electrical signal data and dynamic EIS signal data of healthy soft-pack lithium-ion batteries in historical data, construct the dynamic EIS feature set, pressure feature set and electrical signal feature set using the methods of S2, S3 and S4; The health mean value corresponding to each feature data is calculated based on the dynamic EIS feature set, pressure feature set, and electrical signal feature set, and the health standard deviation of the corresponding feature data is calculated based on the health mean value. Based on the dynamic EIS leakage sensitive feature set, pressure leakage sensitive feature set, and electrical signal leakage sensitive feature set, the standardized outlier value of each feature data and the health standard deviation of the corresponding feature data is calculated, and the dynamic EIS abnormal feature set, pressure abnormal feature set, and electrical signal abnormal feature set are constructed. The dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set, respectively. Compare the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index with the corresponding abnormality thresholds, and output the battery type based on the comparison results.

[0006] Furthermore, a full-band complex impedance spectrum calculation method is introduced in S2: The AC voltage disturbance signal and AC current response signal in the dynamic EIS signal data are separated. Discrete Fourier transform is performed on the AC voltage disturbance signal and AC current response signal respectively to obtain the voltage spectrum and current spectrum corresponding to each frequency point in the full frequency band. The complex impedance spectrum of the full frequency band is calculated by using the complex ratio of the voltage spectrum and the current spectrum. The real part, imaginary part, impedance magnitude, and phase angle of the impedance are extracted from the full-band complex impedance spectrum, and a basic impedance feature set is constructed. Furthermore, S2 further introduces methods for constructing statistical features, impedance change features, and curve shape features; Based on the electrochemical response mechanism, the frequency points within the entire frequency band are divided into three characteristic frequency bands: high frequency, mid frequency, and low frequency. Based on the basic impedance characteristic set, the average impedance amplitude and the average absolute phase angle of the three characteristic frequency bands are calculated respectively. Statistical features are constructed based on the average impedance amplitude and the average absolute phase angle. The impedance change, impedance change slope, and phase angle change are calculated based on the average impedance amplitude and the average absolute value of the phase angle from the statistical characteristics. The impedance change characteristics are then constructed based on the impedance change, impedance change slope, and phase angle change. Dynamic impedance is constructed based on the real and imaginary parts of impedance in the full-band complex impedance spectrum. Curves, extract dynamics The curve's projected area is used to construct the curve's morphological features.

[0007] Furthermore, S3 introduces a method for constructing stage pressure change characteristics, cyclic pressure characteristics, and pressure cyclic evolution characteristics; S1 is divided into four stages: charging stage, post-charging resting stage, discharging stage, and post-discharging resting stage. Extract the pressure at the start and end of each stage from the pressure signal data. Based on the pressure at the start and end of the stage and the stage duration, calculate the pressure rise rate during the charging stage, the pressure fall rate during the discharging stage, the pressure recovery rate during the post-charging resting stage, and the pressure decay rate during the post-discharging resting stage. Construct stage pressure change characteristics based on the pressure rise rate, pressure fall rate, pressure recovery rate, and pressure decay rate. Extract the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure from the single cycle pressure signal data of each stage, and construct cycle pressure features based on the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure. Linear fitting is performed on the pressure change amplitude and the cyclic average pressure to obtain the pressure baseline change slope and the pressure amplitude change slope. Based on the pressure baseline change slope and the pressure amplitude change slope, the pressure cyclic evolution characteristics are constructed.

[0008] Furthermore, S4 introduces a method for constructing electrical signal features, electrical signal offset features, and electrical signal cyclic evolution features; Extract the voltage and power during the charging and discharging phases from the electrical signal data, and calculate the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The electrical signal characteristics are constructed based on the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The charging voltage deviation rate, discharging voltage deviation rate, charging power deviation rate, and discharging power deviation rate are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding health reference values. Electrical signal offset features are constructed based on charging voltage offset rate, discharging voltage offset rate, charging power offset rate, and discharging power offset rate. The slopes of the charging voltage change, discharging voltage change, charging power change, and discharging power change are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding fitting intercepts. The cyclic evolution characteristics of the electrical signal are constructed based on the slopes of the charging voltage change, the discharging voltage change, the charging power change, and the discharging power change.

[0009] Furthermore, in S5, the weights of the dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are dynamically adjusted based on the sensitivity of different characteristic data to leakage faults. Based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set of historical healthy batteries and leaking batteries, the historical feature mean of each feature data is calculated. The intra-class variance was calculated based on the historical characteristic mean of healthy batteries and leaking batteries, and the discrimination ability evaluation value was calculated through the intra-class variance of healthy samples and leaking samples. Then, the discrimination ability evaluation values ​​are normalized within groups to obtain new weights, and the dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the new weights.

[0010] Furthermore, an abnormal threshold determination method is introduced in S5; The historical data of the health battery is sorted from smallest to largest according to different health index values, and the midpoint between two adjacent health indices is used as the candidate threshold. The abnormality index is compared with the candidate threshold, and the true positive (TP), false positive (FP), true negative (TN), and false negative (FN) are counted based on the comparison results. Then, the sensitivity and specificity are calculated. Calculate the Youden index for each candidate threshold and select the candidate threshold with the largest Youden index as the final threshold; If multiple candidate thresholds correspond to the same Youden index, then the smaller threshold will be selected.

[0011] Furthermore, in S5, the dynamic EIS abnormal index, pressure abnormal index, and electrical signal abnormal index are compared with the corresponding abnormal thresholds, and the battery type is output based on the comparison results, thus further introducing a faulty battery classification mechanism. Among them, the corresponding abnormal thresholds include dynamic EIS early warning threshold, pressure abnormal threshold, and electrical signal abnormal threshold; If the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index are all not greater than the corresponding abnormal threshold, the battery is judged to be healthy. If the dynamic EIS abnormal index is greater than the corresponding abnormal threshold, and the pressure abnormal index is not greater than the corresponding abnormal threshold, it is judged as a suspected leaking battery. If both the dynamic EIS abnormality index and the pressure abnormality index are greater than the corresponding abnormal threshold, the battery is judged to be leaking. If the dynamic EIS abnormality index is greater than the corresponding abnormality threshold, and the pressure abnormality threshold and electrical signal abnormality threshold are not greater than the corresponding abnormality threshold, then it is judged as another faulty battery.

[0012] Compared with the prior art, the present invention provides a method for diagnosing leaking batteries by combining dynamic EIS, pressure signals and electrical signals, which has the following advantages: 1. This method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals can keenly detect subtle electrochemical changes such as ion transport obstruction and increased interfacial reaction impedance caused by the initial stage of electrolyte leakage by analyzing features such as full-band complex impedance spectrum, low-frequency impedance change, and Nyquist curve projection area. Compared with traditional methods that rely on voltage / current abnormality alarms, this invention can issue early warnings in the early stage of leakage fault development, solving the problems of signal lag and difficulty in capturing the initial characteristics of leakage in traditional methods. 2. This dynamic EIS, pressure signal and electrical signal combined leakage battery diagnosis method uses the pressure anomaly index to verify whether there are physical changes related to the failure of the sealing performance. By using the dual confirmation mode of electrochemical anomaly and mechanical response anomaly, it can effectively distinguish leakage fault from normal battery aging, polarization enhancement and other situations, and greatly reduce the misjudgment rate caused by the susceptibility of a single signal source to operating condition interference. 3. This method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals incorporates electrical signal characteristics as a verification of operating status. In the diagnostic logic, when only the electrical signal is abnormal while the dynamic EIS and pressure signals are normal, the system will classify it as an "other faulty battery" rather than a leaking battery. This three-level joint diagnostic mechanism ensures the refinement and accuracy of fault classification, avoids misjudging non-leaking faults as leaking faults, and improves the reliability of diagnostic results. 4. This dynamic EIS, pressure signal and electrical signal combined leakage battery diagnosis method adopts Fisher discriminant coefficient to adaptively determine the weight of each feature in the anomaly index, so that features with high sensitivity to leakage occupy a larger proportion in the diagnosis. Combined with the threshold dynamic optimization method based on maximizing Youden index, it ensures the scientificity and adaptability of the classification threshold selection, so that the whole diagnosis method has strong anti-interference ability and robustness to different working conditions. Attached Figure Description

[0013] Figure 1 This is the overall flowchart of the present invention; Figure 2 This is a flowchart of the multi-source signal acquisition and feature extraction process of the present invention; Figure 3 This is a flowchart of the multi-source anomaly index joint judgment and classification process of the present invention; Figure 4 This is a schematic diagram of the dynamic Nyquist curve of the historical leakage battery 1 of the present invention. Figure 5 This is a schematic diagram of the dynamic Nyquist curve of the historical leakage battery 2 of the present invention; Figure 6 This is a schematic diagram of the dynamic Nyquist curve of the historical healthy battery of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0015] Example 1: Please refer to Figure 1-6 The method for diagnosing leaking batteries by combining dynamic EIS, pressure signal and electrical signal in this embodiment includes the following steps: S1. Divide the single cycle process of the soft-pack lithium-ion battery into four stages, and collect the pressure signal data, electrical signal data and dynamic EIS signal data of the soft-pack lithium-ion battery to be tested at each stage and perform preprocessing. S2. Based on the dynamic EIS signal data of the soft-pack lithium-ion battery under test, the full-band complex impedance spectrum is obtained, and the basic impedance feature set is constructed; The statistical characteristics, impedance change characteristics, and curve morphology characteristics were calculated using the basic impedance characteristic set, and a dynamic EIS leakage sensitivity characteristic set was constructed. S3. Based on the pressure signal data of the soft-pack lithium-ion battery under test, calculate the stage pressure change characteristics, cycle pressure characteristics, and pressure cycle evolution characteristics, and construct a pressure leakage sensitive feature set; S4. Based on the electrical signal data of the soft-pack lithium-ion battery to be tested, calculate the electrical signal characteristics, electrical signal offset characteristics, and electrical signal cyclic evolution characteristics, and construct a set of leakage-sensitive features of electrical signals; S5. Based on the pressure signal data, electrical signal data and dynamic EIS signal data of healthy soft-pack lithium-ion batteries in historical data, construct the dynamic EIS feature set, pressure feature set and electrical signal feature set using the methods of S2, S3 and S4; The health mean value corresponding to each feature data is calculated based on the dynamic EIS feature set, pressure feature set, and electrical signal feature set, and the health standard deviation of the corresponding feature data is calculated based on the health mean value. Based on the dynamic EIS leakage sensitive feature set, pressure leakage sensitive feature set, and electrical signal leakage sensitive feature set, the standardized outlier value of each feature data and the health standard deviation of the corresponding feature data is calculated, and the dynamic EIS abnormal feature set, pressure abnormal feature set, and electrical signal abnormal feature set are constructed. The dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set, respectively. Compare the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index with the corresponding abnormality thresholds, and output the battery type based on the comparison results.

[0016] In S2, a method for calculating the complex impedance spectrum across the entire frequency band is further introduced: The AC voltage disturbance signal and AC current response signal in the dynamic EIS signal data are separated. Discrete Fourier transform is performed on the AC voltage disturbance signal and AC current response signal respectively to obtain the voltage spectrum and current spectrum corresponding to each frequency point in the full frequency band. The complex impedance spectrum of the full frequency band is calculated by using the complex ratio of the voltage spectrum and the current spectrum. The real part, imaginary part, impedance amplitude, and phase angle of the impedance are extracted from the full-band complex impedance spectrum, and a basic impedance feature set is constructed.

[0017] It should be noted that the impedance excitation employs a multi-frequency sinusoidal superposition technique; Preferably, 15 frequency points are set at approximately logarithmic intervals within the range of 0.2384 to 5004.896 Hz. An example of the selected frequency sequence is as follows: ; Calculate the impedance and the time-domain voltage response. and current Perform Discrete Fourier Transform (DFT) to calculate complex impedance : ; in, and These are the real and imaginary parts of the impedance, respectively. Then, it outputs a high-quality EIS data point set at each frequency point within that time period for each measurement. ; In obtaining complex impedance Next, the impedance magnitude and phase angle are calculated. The impedance magnitude is expressed as: ; The phase angle is expressed as: ; For each cycle, each stage, and each frequency point, the following fundamental impedance characteristics can be obtained: Real part of impedance Imaginary part of impedance Impedance amplitude Phase angle ; The aforementioned basic characteristics can be used to reflect changes in battery ohmic impedance, interfacial reaction impedance, diffusion impedance, and polarization behavior, respectively.

[0018] In addition, S2 further introduces methods for constructing statistical features, impedance change features, and curve shape features; Based on the electrochemical response mechanism, the frequency points within the entire frequency band are divided into three characteristic frequency bands: high frequency, mid frequency, and low frequency. Based on the basic impedance characteristic set, the average impedance amplitude and the average absolute phase angle of the three characteristic frequency bands are calculated respectively. Statistical features are constructed based on the average impedance amplitude and the average absolute phase angle. The impedance change, impedance change slope, and phase angle change are calculated based on the average impedance amplitude and the average absolute value of the phase angle from the statistical characteristics. The impedance change characteristics are then constructed based on the impedance change, impedance change slope, and phase angle change. Dynamic impedance is constructed based on the real and imaginary parts of impedance in the full-band complex impedance spectrum. Curves, extract dynamics The curve's projected area is used to construct the curve's morphological features.

[0019] It should be noted that, in order to reduce the impact of occasional fluctuations at a single frequency point, the 15 EIS frequency points are divided into three frequency bands: the frequency band is above 100Hz, the frequency band is 10-100Hz, and the frequency band is below 10Hz. The system can extract the mean of the real part of impedance, the mean of the absolute value of the imaginary part of impedance, the mean of the impedance amplitude, and the mean of the absolute value of the phase angle for each frequency band. It can also extract one or more of the following: standard deviation, maximum value, minimum value, median, and interquartile range. Among them, the high-frequency band mainly reflects the changes in the battery's internal ohmic resistance and connection state; the mid-frequency band mainly reflects the changes in charge transfer process and interface reaction; and the low-frequency band mainly reflects the diffusion process, concentration polarization, and the degree of restriction on ion transport. Electrolyte leakage typically leads to a more sensitive impedance response in the low-frequency and mid-low-frequency ranges. Therefore, low-frequency impedance amplitude, low-frequency impedance change, and low-frequency impedance slope can be used as preferred leakage-sensitive characteristics. For a certain frequency band Its average impedance amplitude can be expressed as: ; in, This indicates the selected frequency band, which can be any one of the high-frequency, mid-frequency, or low-frequency bands. This indicates the number of frequency points within the frequency band. Representing frequency point The impedance amplitude at the following values; Accordingly, the mean absolute value of the phase angle within the frequency band can be calculated: ; In this process, the absolute values ​​of the imaginary part of the impedance and the phase angle are taken first, and then the average value is calculated to avoid the cancellation of positive and negative values.

[0020] Suppose that the impedance amplitude at a certain frequency point in a certain stage is initially... The final impedance amplitude is Then the impedance change during this stage is: ; For a certain frequency band The change in the average amplitude of the frequency band impedance within a given period, i.e., the change in impedance, can be expressed as: ; in, This represents the average band impedance amplitude of the initial data for all frequency points within this stage. This represents the average band impedance amplitude of the last data at all frequency points within the band during this stage. To further reflect the rate of impedance change with the development of each stage, the slope of impedance change within each stage can be calculated: ; in, Indicates frequency band The slope of the impedance change below, This indicates the time interval between the last EIS measurement and the first EIS measurement in this phase; The change in the phase angle of the frequency band within the stage is expressed as: ; in, This represents the difference between the mean absolute values ​​of the frequency band phase angles of the last and first EIS measurements in this phase, i.e., the change in phase angle. Start and end represent the first and last EIS measurements in this phase, respectively. When leakage leads to a reduction in electrolyte and obstruction of ion transport, the impedance in the low-frequency region may show a more obvious upward trend as the cycle or stage develops. Therefore, the amount of change in low-frequency impedance and the slope of the change in low-frequency impedance can be used as preferred leakage-sensitive characteristics of dynamic EIS.

[0021] Using the real part of the impedance of historical healthy batteries and leaking batteries as the x-axis and the inverse of the imaginary part of the impedance as the y-axis, and connecting the discrete frequency points in descending order of frequency, three sets of dynamic Nyquist curves are constructed, as follows: Figure 4-6 ; The Nyquist curves of the first two groups of leaking batteries showed a significant overall shift, expansion at the low-frequency end, and increased curve spacing as the measurement sequence progressed, indicating that their impedance spectrum morphology continued to evolve with cycling. The curves of the second group of normal batteries had a high degree of overlap, and the changes in the semicircular region and the low-frequency tail were relatively small, showing good impedance stability. Therefore, the above-mentioned differences between leaking batteries and normal batteries can serve as the basis for curve morphology feature extraction and leakage identification. Transforming the above curve differences into quantifiable features, for the first... The next EIS measurement will be the same group The Nyquist coordinates of each frequency point are as follows: Extract the projected area of ​​the Nyquist curve. Its expression is: ; in, Indicates the dynamic EIS measurement sequence number. Indicates the frequency point number. This indicates the number of frequency points included in each measurement.

[0022] The dynamic EIS leakage sensitive feature set is as follows: .

[0023] Furthermore, S3 introduces a method for constructing stage pressure change characteristics, cyclic pressure characteristics, and pressure cyclic evolution characteristics; S1 is divided into four stages: charging stage, post-charging resting stage, discharging stage, and post-discharging resting stage. Extract the pressure at the start and end of each stage from the pressure signal data. Based on the pressure at the start and end of the stage and the stage duration, calculate the pressure rise rate during the charging stage, the pressure fall rate during the discharging stage, the pressure recovery rate during the post-charging resting stage, and the pressure decay rate during the post-discharging resting stage. Construct stage pressure change characteristics based on the pressure rise rate, pressure fall rate, pressure recovery rate, and pressure decay rate. Extract the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure from the single cycle pressure signal data of each stage, and construct cycle pressure features based on the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure. Linear fitting is performed on the pressure change amplitude and the cyclic average pressure to obtain the pressure baseline change slope and the pressure amplitude change slope. Based on the pressure baseline change slope and the pressure amplitude change slope, the pressure cyclic evolution characteristics are constructed.

[0024] It should be noted that the number is... The pressure data in each cycle is Then we can obtain them separately. , , , ; in, Indicates the first Pressure data for each charging cycle phase Indicates the first Pressure data during the resting phase after each charging cycle. Indicates the first Pressure data for each cycle of discharge. Indicates the first Pressure data during the resting phase after each discharge cycle; Let the pressure at the beginning of a certain stage be... The pressure at the end is The duration of the phase is Then the rate of pressure change during this stage is: ; During the charging phase, the rate of pressure rise can be expressed as: ; For the discharge phase, the pressure drop rate can be expressed as: ; For the resting phase after charging, the pressure recovery rate can be expressed as follows: ; For the post-discharge resting phase, the pressure decay rate can be expressed as follows: ; Due to the reduction of internal electrolyte, decreased wetting state, or changes in the sealing of the package, the pressure change rate of a leaking battery may differ significantly from that of a normal battery. Therefore, the aforementioned stage rate characteristics can be used to determine whether the pressure signal supports the determination of leakage.

[0025] In addition to the rate of change of pressure in each stage, it also starts from the first stage. Complete stress data for each cycle Extracting cyclic pressure features; Cyclic pressure characteristics include maximum pressure Minimum pressure Pressure change amplitude and cyclic mean pressure ; The magnitude of the pressure change is calculated from the difference between the maximum and minimum pressures in the same cycle, i.e.: , , ; in, Indicates the first The maximum pressure in each cycle, Indicates the first Minimum pressure in each cycle, Indicates the first The magnitude of pressure change in each cycle; Calculate the cyclic mean pressure: ; in, Indicates the first Number of pressure sampling points within a cycle; If leakage causes changes in the internal support state and stress response of the battery, the leaking battery may exhibit phenomena such as a shift in the cycle pressure baseline, a decrease in the magnitude of pressure changes, or an increase in pressure fluctuations.

[0026] For cyclic mean pressure Perform linear fitting: ; in, This represents the slope of the change in mean cyclic pressure as the cycle progresses, i.e., the slope of the pressure baseline change. Indicates the fit intercept; For cyclic pressure amplitude Perform linear fitting: ; in, This represents the slope of the pressure change amplitude as the cycle progresses, i.e., the slope of the pressure amplitude change. As a leakage fault gradually develops, the pressure signal may show a continuous shift or a decrease in phased response capability. Therefore, the above-mentioned cyclic evolution slope can be used to characterize the development trend of leakage faults.

[0027] Then construct a pressure leakage sensitive feature set: .

[0028] In addition, S4 further introduces a method for constructing electrical signal features, electrical signal offset features, and electrical signal cyclic evolution features; Extract the voltage and power during the charging and discharging phases from the electrical signal data, and calculate the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The electrical signal characteristics are constructed based on the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The charging voltage deviation rate, discharging voltage deviation rate, charging power deviation rate, and discharging power deviation rate are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding health reference values. Electrical signal offset features are constructed based on charging voltage offset rate, discharging voltage offset rate, charging power offset rate, and discharging power offset rate. The slopes of the charging voltage change, discharging voltage change, charging power change, and discharging power change are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding fitting intercepts. The cyclic evolution characteristics of the electrical signal are constructed based on the slopes of the charging voltage change, the discharging voltage change, the charging power change, and the discharging power change.

[0029] It should be noted that, let the first... The electrical signals collected in each cycle are represented as voltages. Current ,power ; Based on the cycle phase division, the charging phase voltage is extracted separately. Discharge stage voltage Power during charging phase Discharge stage power ; For each stage, an average value can be further calculated: Average voltage during charging phase ; Average voltage during discharge phase ; Average power during charging phase ; Average power during discharge phase ; in, Indicates the current loop number. This represents the number of sampling points during the charging phase. The above average value represents the number of sampling points during the discharge phase, and can reflect the operating status of the entire phase, thus reducing the impact of fluctuations at individual sampling points.

[0030] For each battery under test, select the first few stable and abnormal cycles as reference cycles; Take the first 3 stable cycles as the base cycle: ; Similarly, we can obtain ; ; ; in, and They represent the first Average voltage during the charging phase and average voltage during the discharging phase of each cycle. and They represent the first Average power during the charging phase and average power during the discharging phase of each cycle.

[0031] For any feature's offset ; in, The current loop feature value, For health reference values; For the offset rate of any feature: ; Charging voltage offset rate: ; Discharge voltage offset rate: ; Charging power offset rate: ; Discharge power offset rate: .

[0032] The average charging voltage can be expressed as: ; The average discharge voltage can be expressed as: ; The average charging power can be expressed as: ; The average discharge power can be expressed as: ; Let N be the number of iterations involved in the fitting process, and let the k-th iteration be numbered as follows: The corresponding feature value of the i-th class is (i=1, 2, 3, 4), determined by minimizing the sum of squared residuals. and :

[0033] The mean, slope of change, and intercept of each variable are calculated using the following formulas: ; ; ; in, These correspond to the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase, respectively. This indicates the slope of the feature as the cycle number changes. Indicates the slope of the charging voltage change, Indicates the slope of the discharge voltage change. Indicates the slope of the change in charging power. This represents the slope of the change in discharge power. Indicates that the fitted line is in The intercept at the point; Used to characterize the changing trend of the corresponding electrical signal as the cycle progresses, and as a feature for subsequent diagnosis; The intercept parameter is only required to complete the linear fitting and is not included in the set of sensitive features for leakage of electrical signals. If leakage continues to develop, the voltage and power curves will usually show a continuous shift. Therefore, the slope of change can also be used as an auxiliary diagnostic feature.

[0034] Then construct a leakage-sensitive feature set of electrical signals: .

[0035] Example 2: Please refer to Figure 1-6Based on Example 1, a method for diagnosing leaking batteries by combining dynamic EIS, pressure signal and electrical signal is proposed. In S5, the weights of dynamic EIS abnormality index, pressure abnormality index and electrical signal abnormality index are dynamically adjusted based on the sensitivity of different feature data to leaking faults. Based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set of historical healthy batteries and leaking batteries, the historical feature mean of each feature data is calculated. The intra-class variance was calculated based on the historical characteristic mean of healthy batteries and leaking batteries, and the discrimination ability evaluation value was calculated through the intra-class variance of healthy samples and leaking samples. Then, the discrimination ability evaluation values ​​are normalized within groups to obtain new weights, and the dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the new weights.

[0036] It should be noted that the dynamic EIS anomaly index is constructed as follows: ; in, For the first A dynamic EIS anomaly characteristic, For the corresponding weights, Number of features; When the dynamic EIS anomaly index And the number of cycles that continuously satisfy this condition reaches When the test battery is deemed to have a high risk of leakage, a dynamic EIS early warning signal is output. in, For dynamic EIS early warning thresholds, The pressure anomaly index is the number of consecutive abnormal cycles and is mainly used to verify whether early anomalies in dynamic EIS have physical leakage characteristics.

[0037] Construct an abnormal stress index: ; in, For the first One characteristic of abnormal pressure, For the corresponding weights, The number of features; When the abnormal stress index At that time, the pressure signal is used to determine leakage. in, This is the pressure anomaly threshold; When the dynamic EIS abnormality index and the pressure abnormality index both exceed their respective thresholds, it can be considered that the battery is simultaneously exhibiting abnormal electrochemical impedance and abnormal mechanical response, thereby improving the reliability of leak diagnosis.

[0038] Constructing an electrical signal anomaly index: ; in, For the first Abnormal characteristics of electrical signals For the corresponding weights, The number of features; When the electrical signal abnormality index At that time, it was determined that the operating state of the battery under test had significantly deviated from a healthy state; in, The threshold for electrical signal abnormality; When the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index all exceed their respective thresholds, it indicates that the battery under test not only has leakage-related electrochemical impedance abnormalities and mechanical response abnormalities, but these abnormalities have also affected the battery's operating performance. In this case, a higher level of leakage fault alarm can be output.

[0039] It should be noted that, let the first... The anomaly feature set is as follows: ; in, These represent the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set of historical batteries, respectively. For the first The number of anomaly features; For abnormal features First, calculate its ability to distinguish between healthy batteries and leaking batteries. Then determine the corresponding weights. ; The corresponding weights are represented as follows: ; in, Abnormal characteristics The weight, This is the evaluation value for the discriminative power of this abnormal feature; For the The first in the set of abnormal features One abnormal feature Let the number of historical healthy battery samples be... The number of historical leakage battery samples is ; No. The feature values ​​corresponding to the healthy samples and the leakage samples are respectively and ; The feature means of the two types of samples are as follows: ; ; The within-class variances of healthy samples and leaked samples are as follows: ; ; Abnormal characteristics Fisher's discrimination coefficient, i.e., the rating of his ability to distinguish. Calculate using the following formula: ; in, To prevent extremely small positive numbers with a denominator of zero; The larger the value, the more significant the inter-class difference between healthy samples and leaking samples is relative to the intra-class fluctuation, and the stronger the discriminative power of this feature. Calculated Then, the weights are obtained according to the aforementioned within-group normalization formula. .

[0040] In addition, an abnormal threshold determination method is further introduced in S5; The historical data of the health battery is sorted from smallest to largest according to different health index values, and the midpoint between two adjacent health indices is used as the candidate threshold. The abnormality index is compared with the candidate threshold, and the true positive (TP), false positive (FP), true negative (TN), and false negative (FN) are counted based on the comparison results. Then, the sensitivity and specificity are calculated. Calculate the Youden index for each candidate threshold and select the candidate threshold with the largest Youden index as the final threshold; If multiple candidate thresholds correspond to the same Youden index, then the smaller threshold will be selected.

[0041] It should be noted that the abnormal index category is... ; in, This indicates the category of the anomaly index currently undergoing threshold optimization; when hour, Indicates the first Dynamic EIS anomaly index of a modeling sample; when hour, Indicates the first The stress anomaly index of the modeled sample; when hour, Indicates the first The electrical signal anomaly index of a modeling sample; For modeling sample numbers, The total number of modeling samples, each with an actual class label. ,in Indicates a healthy sample. This indicates a leaking sample.

[0042] For category The anomaly indices of the modeling samples are arranged in ascending order as follows: ; in, Indicate category The One abnormal index value; in parentheses Indicates the sorting position. Indicate category Number of abnormal index values; The midpoint between two adjacent distinct index values ​​is defined as the candidate threshold: ; in, Indicate category The One candidate threshold, The candidate threshold number; All candidate thresholds constitute the candidate set. ; in, Indicate category The candidate threshold set uses the midpoint of adjacent different index values ​​as the candidate threshold, which can cover all boundaries where the classification results change when the sample ranking occurs. For any candidate threshold An anomaly is defined as an anomaly index exceeding a threshold. ; in, Indicates the category used The When the candidate threshold is , for the , The predicted labels obtained from each sample This is an indicator function; it takes the value 1 if the condition inside the parentheses is true, and 0 otherwise. therefore, This indicates that the sample has been identified as abnormal. This indicates that the sample was determined to be normal; In candidate threshold Below, the number of true positives, false positives, true negatives, and false negatives are respectively: ; ; ; ; in, The true positive count is the number of samples that were actually leaking and were correctly identified as abnormal. This refers to the number of false positives, which are samples that are actually healthy but were mistakenly identified as abnormal. This represents the number of true negatives, i.e., the number of samples that are actually healthy and correctly classified as normal. This represents the number of false negatives, i.e., the number of samples that were actually leaking but were mistakenly identified as normal. For each candidate threshold, the sensitivity and specificity are calculated as follows: ; ; in, or This indicates sensitivity or true positive rate, which is the proportion of all actual leak samples that are correctly identified as abnormal. or This indicates specificity or true negative rate, which is the proportion of all actual healthy samples that are correctly identified as normal. The false positive rate is defined as: ; in, This represents the proportion of all actual healthy samples that were misclassified as abnormal. by x-axis Using the vertical axis, we can obtain the category. A corresponding operating point on the ROC curve; The Youden index for each candidate threshold is calculated as follows:

[0043] in, Indicate category In the The Youden index under each candidate threshold; The larger the threshold, the better the healthy samples can be excluded while improving the leakage detection rate, and the stronger its comprehensive discrimination ability. Let the optimal candidate threshold number be: ; Then category The final abnormality index threshold is: ; in, This represents the candidate threshold number that causes the Youden index to reach its maximum value. Indicate category The final anomaly index threshold adopted; If multiple candidate thresholds correspond to the same maximum Youden index, the threshold with the smaller value is selected using the following formula: ; in, This is a temporary sequence number used when iterating through all candidate thresholds; This indicates that the minimum value is selected from the candidate thresholds that satisfy the conditions within the parentheses; To each , and Perform the above calculations and record the corresponding optimal candidate threshold indices as follows: , and Then the three final thresholds are ; ; ; in, , and These represent the optimal candidate threshold numbers corresponding to the dynamic EIS, pressure, and electrical signal anomaly indices, respectively. , and These are the final judgment thresholds for the three types of abnormal indices.

[0044] Example 3: Please refer to Figure 1-6 Based on Example 2, a method for diagnosing leaking batteries by combining dynamic EIS, pressure signal and electrical signal is proposed. In S5, the dynamic EIS abnormal index, pressure abnormal index, electrical signal abnormal index and corresponding abnormal threshold are compared, and the battery type is output according to the comparison result. A faulty battery classification mechanism is further introduced. Among them, the corresponding abnormal thresholds include dynamic EIS early warning threshold, pressure abnormal threshold, and electrical signal abnormal threshold; If the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index are all not greater than the corresponding abnormal threshold, the battery is judged to be healthy. If the dynamic EIS abnormal index is greater than the corresponding abnormal threshold, and the pressure abnormal index is not greater than the corresponding abnormal threshold, it is judged as a suspected leaking battery. If both the dynamic EIS abnormality index and the pressure abnormality index are greater than the corresponding abnormal threshold, the battery is judged to be leaking. If the dynamic EIS abnormal index is greater than the corresponding abnormal threshold, and the pressure abnormal threshold and electrical signal abnormal threshold are not greater than the corresponding abnormal threshold, then it is judged as another faulty battery.

[0045] It should be noted that when the battery under test simultaneously meets the following conditions... , , ; This indicates that the battery's dynamic impedance response, pressure response, and electrical signal operating status are all within the healthy baseline range, and no fault alarm is output. When the battery under test meets ,and ; This indicates that the battery has exhibited abnormal dynamic impedance, but the pressure signal has not yet shown any obvious abnormalities. This state may correspond to the early stage of leakage, or it may be caused by normal aging, enhanced polarization, or test disturbance. Therefore, the system does not immediately confirm leakage, but outputs a suspected leakage warning signal and increases the monitoring frequency in subsequent cycles. When the battery under test simultaneously meets , ; This indicates that the battery under test exhibits both dynamic impedance anomalies and pressure anomalies. Dynamic EIS anomalies reflect abnormalities in the battery's internal ion transport, interface reaction, or diffusion processes, while pressure anomalies reflect abnormalities in the battery's external mechanical response or sealing condition. The two types of signals can verify each other, reducing the risk of misjudgment caused by a single impedance signal due to normal aging or polarization enhancement. When the battery under test meets ,and , ; This indicates that the battery's voltage, power, and other operating conditions have deviated from a healthy state, but neither the dynamic EIS nor the pressure signal shows typical leakage characteristics. Therefore, this anomaly is more likely caused by capacity decay, poor contact, abnormal temperature, enhanced polarization, or other non-leakage faults, rather than being directly identified as a leaking battery.

[0046] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals, characterized in that, Includes the following steps: S1. Divide the single cycle process of the soft-pack lithium-ion battery into four stages, and collect the pressure signal data, electrical signal data and dynamic EIS signal data of the soft-pack lithium-ion battery under test at each stage and perform preprocessing. S2. Based on the dynamic EIS signal data of the soft-pack lithium-ion battery under test, the full-band complex impedance spectrum is obtained, and the basic impedance feature set is constructed; Statistical characteristics, impedance change characteristics, and curve morphology characteristics were calculated using the basic impedance characteristic set, and a dynamic EIS leakage sensitivity characteristic set was constructed. S3. Based on the pressure signal data of the soft-pack lithium-ion battery under test, calculate the stage pressure change characteristics, cycle pressure characteristics, and pressure cycle evolution characteristics, and construct a pressure leakage sensitive feature set; S4. Based on the electrical signal data of the soft-pack lithium-ion battery to be tested, calculate the electrical signal characteristics, electrical signal offset characteristics, and electrical signal cyclic evolution characteristics, and construct a set of leakage-sensitive features of the electrical signal. S5. Based on the pressure signal data, electrical signal data and dynamic EIS signal data of healthy soft-pack lithium-ion batteries in historical data, construct the dynamic EIS feature set, pressure feature set and electrical signal feature set using the methods of S2, S3 and S4; The health mean value corresponding to each feature data is calculated based on the dynamic EIS feature set, pressure feature set, and electrical signal feature set, and the health standard deviation of the corresponding feature data is calculated based on the health mean value. Based on the dynamic EIS leakage sensitive feature set, pressure leakage sensitive feature set, and electrical signal leakage sensitive feature set, the standardized outlier value of each feature data and the health standard deviation of the corresponding feature data is calculated, and the dynamic EIS abnormal feature set, pressure abnormal feature set, and electrical signal abnormal feature set are constructed. The dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set, respectively. Compare the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index with the corresponding abnormality thresholds, and output the battery type based on the comparison results.

2. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 1, characterized in that: S2 further introduces a method for calculating the full-band complex impedance spectrum: The AC voltage disturbance signal and AC current response signal in the dynamic EIS signal data are separated. Discrete Fourier transform is performed on the AC voltage disturbance signal and AC current response signal respectively to obtain the voltage spectrum and current spectrum corresponding to each frequency point in the full frequency band. The complex impedance spectrum of the full frequency band is calculated by using the complex ratio of the voltage spectrum and the current spectrum. The real part, imaginary part, impedance amplitude, and phase angle of the impedance are extracted from the full-band complex impedance spectrum, and a basic impedance feature set is constructed.

3. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 2, characterized in that: The S2 further introduces methods for constructing statistical features, impedance change features, and curve shape features; Based on the electrochemical response mechanism, the frequency points within the entire frequency band are divided into three characteristic frequency bands: high frequency, mid frequency, and low frequency. Based on the basic impedance characteristic set, the average impedance amplitude and the average absolute phase angle of the three characteristic frequency bands are calculated respectively. Statistical features are constructed based on the average impedance amplitude and the average absolute phase angle. The impedance change, impedance change slope, and phase angle change are calculated based on the average impedance amplitude and the average absolute value of the phase angle from the statistical characteristics. The impedance change characteristics are then constructed based on the impedance change, impedance change slope, and phase angle change. Dynamic impedance is constructed based on the real and imaginary parts of impedance in the full-band complex impedance spectrum. Curves, extract dynamics The curve's projected area is used to construct the curve's morphological features.

4. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 1, characterized in that: The S3 further introduces a method for constructing stage pressure change characteristics, cyclic pressure characteristics, and pressure cyclic evolution characteristics; S1 is divided into four stages: charging stage, post-charging resting stage, discharging stage, and post-discharging resting stage. Extract the pressure at the start and end of each stage from the pressure signal data. Based on the pressure at the start and end of the stage and the stage duration, calculate the pressure rise rate during the charging stage, the pressure fall rate during the discharging stage, the pressure recovery rate during the post-charging resting stage, and the pressure decay rate during the post-discharging resting stage. Construct stage pressure change characteristics based on the pressure rise rate, pressure fall rate, pressure recovery rate, and pressure decay rate. Extract the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure from the single cycle pressure signal data of each stage, and construct cycle pressure features based on the maximum pressure, minimum pressure, pressure change amplitude, and cycle average pressure. Linear fitting is performed on the pressure change amplitude and the cyclic average pressure to obtain the pressure baseline change slope and the pressure amplitude change slope. Based on the pressure baseline change slope and the pressure amplitude change slope, the pressure cyclic evolution characteristics are constructed.

5. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 1, characterized in that: The method for constructing electrical signal features, electrical signal offset features, and electrical signal cyclic evolution features is further introduced in S4; Extract the voltage and power during the charging and discharging phases from the electrical signal data, and calculate the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The electrical signal characteristics are constructed based on the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, and the average power during the discharging phase. The charging voltage deviation rate, discharging voltage deviation rate, charging power deviation rate, and discharging power deviation rate are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding health reference values. Electrical signal offset features are constructed based on charging voltage offset rate, discharging voltage offset rate, charging power offset rate, and discharging power offset rate. The slopes of the charging voltage change, discharging voltage change, charging power change, and discharging power change are calculated using the average voltage during the charging phase, the average voltage during the discharging phase, the average power during the charging phase, the average power during the discharging phase, and the corresponding fitting intercepts. The cyclic evolution characteristics of the electrical signal are constructed based on the slopes of the charging voltage change, the discharging voltage change, the charging power change, and the discharging power change.

6. A method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to any one of claims 2-5, characterized in that: S5 further introduces a dynamic adjustment of the weights of the dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index based on the sensitivity of different characteristic data to leakage faults. Based on the dynamic EIS anomaly feature set, pressure anomaly feature set, and electrical signal anomaly feature set of historical healthy batteries and leaking batteries, the historical feature mean of each feature data is calculated. The intra-class variance was calculated based on the historical characteristic mean of healthy batteries and leaking batteries, and the discrimination ability evaluation value was calculated through the intra-class variance of healthy samples and leaking samples. Then, the discrimination ability evaluation values ​​are normalized within groups to obtain new weights, and the dynamic EIS anomaly index, pressure anomaly index, and electrical signal anomaly index are calculated based on the new weights.

7. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 6, characterized in that: S5 further introduces an abnormal threshold determination method; The historical data of the health battery is sorted from smallest to largest according to different health index values, and the midpoint between two adjacent health indices is used as the candidate threshold. The abnormality index is compared with the candidate threshold, and the true positive (TP), false positive (FP), true negative (TN), and false negative (FN) are counted based on the comparison results. Then, the sensitivity and specificity are calculated. Calculate the Youden index for each candidate threshold and select the candidate threshold with the largest Youden index as the final threshold; If multiple candidate thresholds correspond to the same Youden index, then the smaller threshold will be selected.

8. The method for diagnosing leaking batteries by combining dynamic EIS, pressure signals, and electrical signals according to claim 7, characterized in that: In S5, the dynamic EIS abnormal index, pressure abnormal index, and electrical signal abnormal index are compared with the corresponding abnormal thresholds, and the battery type is output based on the comparison results, further introducing a faulty battery classification mechanism. Among them, the corresponding abnormal thresholds include dynamic EIS early warning threshold, pressure abnormal threshold, and electrical signal abnormal threshold; If the dynamic EIS abnormality index, pressure abnormality index, and electrical signal abnormality index are all not greater than the corresponding abnormal threshold, the battery is judged to be healthy. If the dynamic EIS abnormal index is greater than the corresponding abnormal threshold, and the pressure abnormal index is not greater than the corresponding abnormal threshold, it is judged as a suspected leaking battery. If both the dynamic EIS abnormality index and the pressure abnormality index are greater than the corresponding abnormal threshold, the battery is judged to be leaking. If the dynamic EIS abnormality index is greater than the corresponding abnormality threshold, and the pressure abnormality threshold and electrical signal abnormality threshold are not greater than the corresponding abnormality threshold, then it is judged as another faulty battery.

Images