Lithium battery health state estimation method based on electrochemical impedance spectroscopy characteristic peak tracking

Abstract

The invention relates to the technical field of lithium battery health management, in particular to a lithium battery health state estimation method based on electrochemical impedance spectroscopy characteristic peak tracking, which comprises the following steps: acquiring battery voltage, current, temperature and SOC (State of Charge) data, obtaining local analysis data through working condition screening, constructing a local electrochemical impedance spectroscopy through short-time frequency domain conversion, and calculating the state of health of a lithium battery. Characteristic peak distribution is obtained through self-adaptive peak enhancement, a limited search interval is constructed by combining temperature, SOC and historical peak positions, characteristic peaks are identified and tracked in the interval, parameter sequences under abnormal working conditions such as drifting, overlapping and disappearance are corrected, health characteristics such as peak positions, amplitudes and peak widths are extracted, and the SOH of the lithium battery is calculated according to a preset mapping relation. According to the method, the local electrochemical impedance spectrum is constructed through working condition screening, the health features are extracted and the SOH is estimated through adaptive peak enhancement and feature peak identification tracking, and the problems that a traditional method depends on a single parameter, the working condition adaptability is poor and the estimation stability is insufficient are solved.

Description

Technical Field

[0001] This invention relates to the field of lithium battery health management technology, and in particular to a method for estimating the health status of lithium batteries based on the tracking of characteristic peaks in electrochemical impedance spectroscopy. Background Technology

[0002] With the rapid development of new energy vehicles, energy storage systems, and portable electronic devices, lithium-ion batteries are widely used as the main electrochemical energy storage unit. During long-term charge and discharge processes, batteries undergo aging phenomena such as capacity decay and internal resistance increase. Therefore, accurate assessment of battery health status has become one of the key technologies in battery management systems.

[0003] Traditional lithium-ion battery health state estimation methods mostly rely on single operating parameters or static impedance characteristics for assessment. These methods typically assume the battery operates under stable conditions, inferring health status through offline calibrated impedance spectra or empirical models. However, in actual dynamic operating conditions, battery current, temperature, and state of charge continuously change, causing characteristic peaks in the electrochemical impedance spectroscopy to drift, overlap, or even transiently disappear. Traditional full-frequency peak detection methods, searching for peaks across the entire frequency range, struggle to distinguish between noise spurious peaks and true characteristic peaks. Furthermore, the fixed search interval cannot adapt to dynamic changes in characteristic peaks, resulting in poor temporal consistency in feature extraction and abrupt changes in SOH estimation results, failing to meet the battery management system's requirements for estimation stability.

[0004] Therefore, how to achieve stable tracking of impedance spectrum characteristic peaks under complex dynamic operating conditions has become an urgent technical problem to be solved in the estimation of the health status of lithium batteries. Summary of the Invention

[0005] To overcome the above shortcomings, this invention provides a lithium battery health status estimation method based on electrochemical impedance spectroscopy characteristic peak tracking. By using operating condition screening, adaptive peak enhancement, restricted search interval construction, and time series tracking correction, it solves the problems of unstable characteristic peak identification and low health status estimation accuracy of traditional methods under dynamic operating conditions.

[0006] This invention provides the following technical solution: a method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy, comprising the following steps:

[0007] S1. Collect operating data of lithium battery during operation, including voltage data, current data, temperature data and state of charge (SOC) data. Filter the collected operating data according to preset operating condition filtering rules to obtain local analysis data that meets the filtering conditions.

[0008] S2. Perform short-time frequency domain transformation on the voltage and current data in the local analysis data, calculate the frequency domain impedance and construct the local electrochemical impedance spectrum;

[0009] S3. Perform adaptive peak enhancement transformation on the local electrochemical impedance spectrum to obtain characteristic peak distribution data;

[0010] S4. Combining temperature data, state of charge (SOC) data, and historical characteristic peak position data, determine the target characteristic peak position prediction data and construct the restricted search interval of the target characteristic peak.

[0011] S5. Identify the target feature peak within the restricted search range, extract the target feature peak parameters, and construct the target feature peak parameter time series; for parameter jumps or missing parameters in the target feature peak parameter time series under abnormal working conditions such as drift, overlap, and disappearance, perform continuous tracking and correction processing based on time-series correlation on the target feature peak parameter time series.

[0012] S6. Extract health features from the corrected target characteristic peak parameter time series, and calculate the lithium battery health state SOH estimation result according to the preset mapping relationship between health features and SOH (State of Health).

[0013] By employing the above technical solution, operating conditions are screened from lithium battery operating data, and a local electrochemical impedance spectroscopy is constructed. An adaptive peak enhancement transformation is performed on the local electrochemical impedance spectroscopy, and target characteristic peaks are identified within a limited search range. Then, health features are extracted based on the time series of target characteristic peak parameters, and the state of health (SOH) estimation result of the lithium battery is calculated. This improves upon the problem that traditional lithium battery health state estimation methods mostly rely on a single operating parameter or static impedance feature for state assessment. Due to the difficulty in stably extracting characteristic information reflecting changes in the internal state of the battery under complex operating conditions, the stability of health state estimation is insufficient.

[0014] Furthermore, in S1, the step of filtering the collected operational data according to preset operating condition filtering rules includes:

[0015] The running data is divided into multiple time windows according to a preset time length;

[0016] Calculate the rate of change of current data, the rate of change of voltage data, the amount of change of temperature data, and the amount of change of state of charge (SOC) data within each time window.

[0017] The rate of change of current data is compared with a preset current rate of change threshold, the rate of change of voltage data is compared with a preset voltage rate of change threshold, the amount of change of temperature data is compared with a preset temperature amount of change threshold, and the amount of change of state of charge (SOC) data is compared with a preset SOC change threshold.

[0018] When the rate of change of current data is less than the preset threshold for the rate of change of current, the rate of change of voltage data is less than the preset threshold for the rate of change of voltage, the amount of change of temperature data is less than the preset threshold for the amount of change of temperature, and the amount of change of state of charge (SOC) data is less than the preset threshold for the amount of change of SOC, the running data within the corresponding time window will be determined as local analysis data.

[0019] Further, in S2, the step of performing short-time frequency domain transformation on the voltage and current data in the local analysis data includes:

[0020] The voltage and current data in the local analysis data are divided into multiple time windows according to a preset time length;

[0021] Within each time window, the voltage and current data are weighted using a window function.

[0022] Frequency domain transformations are performed on the weighted voltage data and the weighted current data respectively to obtain the frequency domain representations of the voltage data and the current data for the corresponding time windows;

[0023] The frequency domain representations of voltage and current data obtained from each time window are arranged in chronological order to form short-time frequency domain conversion results for voltage and current data.

[0024] Further, in S2, the step of calculating the frequency domain impedance and constructing the local electrochemical impedance spectrum includes:

[0025] Calculate the impedance value at the corresponding frequency point based on the frequency domain representation of voltage data and current data;

[0026] The impedance value is decomposed to obtain the real part and the imaginary part of the impedance;

[0027] The real part and imaginary part of the impedance are combined with the corresponding frequency values ​​to form an impedance frequency sequence;

[0028] Local electrochemical impedance spectroscopy is constructed based on impedance frequency sequences.

[0029] Further, in S3, the step of performing adaptive peak enhancement transformation on the local electrochemical impedance spectrum includes:

[0030] The impedance amplitude sequence is extracted from the impedance frequency sequence of the local electrochemical impedance spectrum, and the impedance amplitude sequence is smoothed.

[0031] The peak response sequence is obtained by calculating the rate of change and the difference between adjacent frequency points based on the smoothed impedance amplitude sequence.

[0032] Calculate the weighting coefficients corresponding to each frequency point based on the impedance amplitude sequence;

[0033] The enhanced peak response sequence is obtained by combining the peak response sequence with the weighting coefficients, and characteristic peak distribution data is generated based on the enhanced peak response sequence.

[0034] Furthermore, in S3, the adaptive peak enhancement transformation enhances the intensity of characteristic peak responses related to charge transfer and diffusion processes in the electrochemical impedance spectroscopy by using a second-order difference response based on local curvature weighting.

[0035] Further, in S4, the step of determining the target feature peak prediction data and constructing the restricted search interval of the target feature peak includes:

[0036] The historical characteristic peak data are arranged in chronological order to obtain the historical characteristic peak sequence;

[0037] Based on temperature data and state of charge (SOC) data, determine the subset of historical characteristic peak data for the corresponding operating condition range from the historical characteristic peak sequence;

[0038] The peak position change sequence between historical characteristic peaks is calculated based on a subset of historical characteristic peak data, and the target characteristic peak prediction data is calculated by combining temperature data and state of charge (SOC) data.

[0039] The peak position offset range is determined based on historical characteristic peak position data and target characteristic peak position prediction data, and the restricted search interval of the target characteristic peak is determined based on the target characteristic peak position prediction data and the peak position offset range.

[0040] Furthermore, in S5, the step of identifying the target feature peak within the restricted search interval includes:

[0041] Based on the restricted search range of the target feature peak, extract the feature peak distribution data of the corresponding frequency range from the feature peak distribution data and form a feature peak distribution data sequence;

[0042] Traverse each data point in the characteristic peak distribution data sequence in frequency order, and determine the current data point as a candidate characteristic peak when the value of the current data point is greater than that of the adjacent data points.

[0043] Based on the frequency value and peak amplitude of the candidate feature peak, extract and filter the candidate feature peak parameters, and determine the target feature peak parameters;

[0044] The target feature peak parameters obtained at each time point are arranged in chronological order to construct a time series of target feature peak parameters.

[0045] Furthermore, in S5, the step of performing continuous tracking processing and correction on the time series of target characteristic peak parameters includes:

[0046] Traverse adjacent time points in the time series of target feature peak parameters in chronological order, and calculate the rate of change of peak position in the target feature peak parameters corresponding to adjacent time points;

[0047] When the peak position change rate exceeds the preset change range, the corresponding time point is marked as a peak position drift state;

[0048] When no target feature peak parameter is detected within the restricted search range of the target feature peak, the corresponding time point is marked as the peak disappearance state;

[0049] When multiple candidate feature peak parameters are detected at the same time point, the target feature peak parameter is selected and determined based on the temporal correlation between each candidate feature peak parameter and the historical feature peak position, and the corresponding time point is marked as peak overlap state;

[0050] Based on the target characteristic peak parameters at adjacent time points, interpolation correction or neighborhood smoothing is performed on time points marked as peak position drifting or peak disappearance states. Further, in S6, the health characteristics include at least one of the following: the target characteristic peak position change rate, the target characteristic peak value amplitude change rate, and the target characteristic peak width change rate.

[0051] The present invention has the following beneficial effects:

[0052] 1. In this invention, by screening the operating conditions of lithium battery operating data and constructing a local electrochemical impedance spectroscopy, an adaptive peak enhancement transformation is performed on the local electrochemical impedance spectroscopy, and target characteristic peaks are identified within a limited search range. Then, based on the time series of the target characteristic peak parameters, health features are extracted and the state of health (SOH) estimation result of the lithium battery is calculated. This improves upon the problem that traditional lithium battery health state estimation methods mostly rely on a single operating parameter or static impedance feature for state assessment. Due to the difficulty in stably extracting characteristic information reflecting changes in the internal state of the battery under complex operating conditions, the stability of health state estimation is insufficient.

[0053] 2. In this invention, by performing adaptive peak enhancement transformation on the local electrochemical impedance spectrum and generating characteristic peak distribution data, the characteristic peak information reflecting the changes in electrochemical characteristics is highlighted in the impedance spectrum data. This improves the problem that traditional electrochemical impedance spectrum feature extraction methods mostly directly detect the peaks of the original impedance spectrum. Due to the weak peak shape changes or noise interference in the original impedance spectrum, the identification of characteristic peaks is unstable.

[0054] 3. In this invention, by combining temperature data and state of charge (SOC) data, historical characteristic peak position data are matched under operating conditions and a restricted search interval for the target characteristic peak is constructed. At the same time, continuous tracking processing is performed on the time series of the identified target characteristic peak parameters, and peak position drift, peak overlap, and peak disappearance are corrected. This maintains the continuity of the target characteristic peak in the time series, thereby improving the problem that traditional impedance characteristic analysis methods mostly identify peaks independently in the full frequency domain. Since the temporal changes of characteristic peaks are not constrained, the stability of characteristic peak tracking is insufficient.

[0055] 4. This invention constructs local electrochemical impedance spectroscopy (EIS) through operating condition screening, providing a steady-state data foundation for subsequent analysis; it highlights electrochemical characteristic peaks and suppresses noise interference through adaptive peak enhancement transformation; it constructs a restricted search interval by combining temperature, SOC, and historical data, avoiding spurious peaks introduced by full-frequency domain search; and it maintains the continuous evolution of characteristic peaks in the time dimension by correcting the time-series parameters under abnormal operating conditions. These four steps work together to form a complete technical closed loop of "data screening - feature enhancement - interval constraint - time-series tracking." Specifically, operating condition screening provides a high-quality steady-state data foundation for peak enhancement, making the enhanced characteristic peaks more representative; peak enhancement provides a clear distribution of characteristic peaks for interval construction, making the search interval location more accurate; interval constraint provides an effective search range for time-series tracking, avoiding false detections caused by full-frequency domain search; and time-series correction further verifies and optimizes interval prediction, ensuring continuous and stable characteristic peak tracking in the time dimension. Each step supports the others and progresses step by step, jointly solving the technical problem of unstable characteristic peak tracking under dynamic operating conditions and achieving a robust improvement in SOH estimation. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating the lithium battery health status estimation method based on electrochemical impedance spectroscopy characteristic peak tracking proposed in this invention.

[0057] Figure 2 This is a schematic diagram of the architecture of the lithium battery health status estimation system based on electrochemical impedance spectroscopy characteristic peak tracking proposed in an embodiment of the present invention. Detailed Implementation

[0058] The technical solutions in 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.

[0059] Example 1: In the first embodiment of the present invention, a method for estimating the health status of lithium batteries based on electrochemical impedance spectroscopy characteristic peak tracking is provided, such as... Figure 1 As shown, the process includes the following steps: S1, collecting operating data of the lithium battery during operation, including voltage data, current data, temperature data and state of charge (SOC) data, and filtering the collected operating data according to preset operating condition filtering rules to obtain local analysis data that meet the filtering conditions.

[0060] Furthermore, in S1, the step of filtering the collected operational data according to preset operating condition filtering rules includes:

[0061] The running data is divided into multiple time windows according to a preset time length;

[0062] Calculate the rate of change of current data, the rate of change of voltage data, the amount of change of temperature data, and the amount of change of state of charge (SOC) data within each time window.

[0063] The rate of change of current data is compared with a preset current rate of change threshold, the rate of change of voltage data is compared with a preset voltage rate of change threshold, the amount of change of temperature data is compared with a preset temperature amount of change threshold, and the amount of change of state of charge (SOC) data is compared with a preset SOC change threshold.

[0064] When the rate of change of current data is less than the preset threshold for the rate of change of current, the rate of change of voltage data is less than the preset threshold for the rate of change of voltage, the amount of change of temperature data is less than the preset threshold for the amount of change of temperature, and the amount of change of state of charge (SOC) data is less than the preset threshold for the amount of change of SOC, the running data within the corresponding time window will be determined as local analysis data.

[0065] Specifically, the collected voltage, current, temperature, and state of charge (SOC) data from the lithium battery operation process are processed according to a preset time period. Divided into multiple consecutive time windows, with a preset time length. The time constant of the electrochemical process in a lithium battery was predetermined through calibration experiments, and the rate of change of current data was calculated within each time window. Rate of change of voltage data Changes in temperature data and the change in state of charge (SOC) data rate of change of current data The first derivative of the current data with respect to time within the time window is calculated as the ratio of the current difference between adjacent sampling points within the time window to the sampling time interval. The rate of change of the voltage data is also mentioned. The first derivative of the voltage data with respect to time within the time window is calculated as the ratio of the voltage difference between adjacent sampling points within the time window to the sampling time interval. The change in temperature data is also considered. The difference between the maximum and minimum values ​​of temperature data within the time window represents the change in state of charge (SOC) data. The difference between the maximum and minimum values ​​of SOC data within the time window is calculated. Compared with the preset current change rate threshold Compare and calculate the results Compared with the preset voltage change rate threshold Compare and calculate the results Compared with the preset temperature change threshold Compare and calculate the results Compared with the preset SOC change threshold Comparison, preset current change rate threshold Preset voltage change rate threshold Preset temperature change threshold Preset SOC change threshold All were predetermined through lithium battery calibration tests. , , and When both conditions are met, the running data within the corresponding time window is determined as local analysis data, and the obtained local analysis data is used for subsequent short-time frequency domain conversion processing and construction of local electrochemical impedance spectroscopy.

[0066] S2. Perform short-time frequency domain transformation on the voltage and current data in the local analysis data, calculate the frequency domain impedance and construct the local electrochemical impedance spectrum;

[0067] Furthermore, in S2, the steps of performing short-time frequency domain transformation on the voltage and current data in the local analysis data include:

[0068] The voltage and current data in the local analysis data are divided into multiple time windows according to a preset time length;

[0069] Within each time window, the voltage and current data are weighted using a window function.

[0070] Frequency domain transformations are performed on the weighted voltage data and the weighted current data respectively to obtain the frequency domain representations of the voltage data and the current data for the corresponding time windows;

[0071] The frequency domain representations of voltage and current data obtained from each time window are arranged in chronological order to form short-time frequency domain conversion results for voltage and current data.

[0072] Specifically, voltage data from local analysis data and current data For input, the sampling point number is The sampling frequency is , voltage data and current data According to the preset time length Divided into multiple consecutive time windows, with a preset time length. The corresponding window length is Window length The overlap length of adjacent time windows is Window length With overlap length Based on sampling frequency The frequency coverage of the lithium battery electrochemical process is pre-calibrated, and voltage data is analyzed within each time window. and current data The window function performs weighted processing, using a Hanning window, and the expression is: ;in ... The weighted voltage data is The weighted current data is The weighted voltage data were analyzed separately. and weighted current data Performing a discrete Fourier transform yields the frequency domain representation of the voltage data for the corresponding time window. Frequency domain representation of current data The expression for the discrete Fourier transform is: , ;in Using the frequency point number, the voltage data obtained in each time window is represented in the frequency domain. Frequency domain representation of current data Arranged in chronological order, the voltage and current data are converted into short-time frequency domain results. These results are then used for subsequent calculations of frequency domain impedance and construction of local electrochemical impedance spectra.

[0073] Furthermore, in S2, the steps of calculating the frequency domain impedance and constructing the local electrochemical impedance spectrum include:

[0074] Calculate the impedance value at the corresponding frequency point based on the frequency domain representation of voltage data and current data;

[0075] The impedance value is decomposed to obtain the real part and the imaginary part of the impedance;

[0076] The real part and imaginary part of the impedance are combined with the corresponding frequency values ​​to form an impedance frequency sequence;

[0077] Local electrochemical impedance spectroscopy is constructed based on impedance frequency sequences.

[0078] Specifically, the voltage data obtained by short-time frequency domain transformation is represented in the frequency domain. Frequency domain representation of current data For input, where The time window number, The frequency point number is based on the frequency domain representation of the voltage data. Frequency domain representation of current data Calculate the impedance value at the corresponding frequency point. impedance value The calculation expression is as follows The calculated impedance value By performing complex decomposition, the real part of the impedance is obtained. and the imaginary part of impedance impedance value With the real part of the impedance Imaginary part of impedance The relationship is ;in Using the imaginary unit, calculate the frequency value corresponding to each frequency point. Frequency value The calculation expression is as follows ;in For data sampling frequency, Given the window length corresponding to the time window, the frequency values ​​corresponding to the same frequency point are... Real part of impedance Imaginary part of impedance Combine them in ascending order of frequency to form an impedance frequency sequence. And map the impedance values ​​in the impedance frequency sequence into a function form according to the frequency index. For impedance frequency sequences corresponding to multiple consecutive time windows Perform sliding window averaging to obtain local electrochemical impedance spectroscopy. The expression for sliding window averaging is: ;in For the average number of windows, This is the starting window number for averaging. and The local electrochemical impedance spectroscopy was obtained based on the pre-calibrated operating window length. Used for subsequent adaptive peak enhancement transformation processing.

[0079] Through the aforementioned short-time frequency domain transformation and sliding window averaging, this step extracts the local impedance spectrum reflecting the battery's electrochemical response from the dynamic operating data, providing a steady-state frequency domain data foundation for subsequent feature analysis. Compared to directly using the original time-domain data or the full-band impedance spectrum, this processing effectively suppresses the impact of operating condition fluctuations on feature information and improves the reliability of feature extraction.

[0080] S3. Perform adaptive peak enhancement transformation on the local electrochemical impedance spectrum to obtain characteristic peak distribution data;

[0081] Furthermore, in S3, the step of performing adaptive peak enhancement transformation on the local electrochemical impedance spectrum includes:

[0082] The impedance amplitude sequence is extracted from the impedance frequency sequence of the local electrochemical impedance spectrum, and the impedance amplitude sequence is smoothed.

[0083] The peak response sequence is obtained by calculating the rate of change and the difference between adjacent frequency points based on the smoothed impedance amplitude sequence.

[0084] Calculate the weighting coefficients corresponding to each frequency point based on the impedance amplitude sequence;

[0085] The enhanced peak response sequence is obtained by combining the peak response sequence with the weighting coefficients, and characteristic peak distribution data is generated based on the enhanced peak response sequence.

[0086] Specifically, the impedance frequency sequence corresponding to the local electrochemical impedance spectrum. For input, impedance frequency sequence From frequency value Real part of impedance Imaginary part of impedance Arranged in ascending order of frequency, starting with the impedance frequency sequence. Extract the impedance amplitude sequence corresponding to each frequency point Impedance amplitude sequence The calculation expression is as follows ;in The frequency point number represents the impedance amplitude sequence. Perform sliding window smoothing to obtain the smoothed impedance magnitude sequence. The expression for smoothing the sliding window is: ;in To smooth the window length, Based on a pre-calibrated number of frequency points, the smoothed impedance amplitude sequence is used via a forward differential method. Calculate the rate of change between adjacent frequency points to obtain the first-order difference sequence. First-order difference sequence The calculation expression is as follows For first-order difference sequences Calculate the difference between adjacent points to obtain the second-order difference sequence as the peak response sequence. Peak response sequence The calculation expression is as follows According to the impedance amplitude sequence Calculate the weighting coefficients corresponding to each frequency point. Weighting coefficient The calculation expression is as follows ;in The maximum value of the impedance amplitude sequence is determined by the peak response sequence. With weighting coefficients By performing point-by-point product combination calculations, the enhanced peak response sequence is obtained. Enhanced peak response sequence The calculation expression is as follows This will enhance the peak response sequence. and corresponding frequency value Arrange the data in ascending order of frequency to generate characteristic peak distribution data.

[0087] The core of adaptive peak enhancement transformation lies in approximating the local curvature information by calculating the second-order difference of the impedance amplitude sequence. In electrochemical impedance spectroscopy, characteristic peaks correspond to relaxation frequencies of specific electrochemical processes (such as charge transfer and diffusion) within the battery, where the impedance spectrum curvature changes most drastically. Therefore, the second-order difference response sequence can effectively locate these frequency points of curvature abrupt changes. Furthermore, by introducing amplitude-based weighting coefficients, the response in low-amplitude noise regions can be suppressed while enhancing the curvature change signal, resulting in a final enhanced peak response sequence that exhibits significant peaks only at characteristic frequencies related to key electrochemical processes. This process not only enhances the recognizability of characteristic peaks but also achieves 'adaptive' extraction of characteristic peaks, ensuring stable responses under different aging stages and operating conditions.

[0088] Conventional peak detection methods typically search for local maxima in the impedance spectrum. However, in actual operational data, the impedance spectrum is often superimposed with measurement noise, and the characteristic peak shape may become flat due to battery aging, making direct peak detection prone to false positives. This invention obtains the peak response sequence by calculating the second-order difference, which highlights frequency points with drastic curvature changes. These points correspond to characteristic frequencies dominated by charge transfer or diffusion processes in electrochemical processes. Furthermore, by combining an amplitude weighting coefficient, the response in low-amplitude noise regions is suppressed. Through the above adaptive peak enhancement processing, the enhanced peak response sequence only exhibits significant peaks at characteristic frequencies with clear electrochemical significance, thereby significantly improving the accuracy of characteristic peak identification. The obtained characteristic peak distribution data is used for subsequent construction of restricted search intervals for target characteristic peaks and target characteristic peak identification processing.

[0089] S4. Combining temperature data, state of charge (SOC) data, and historical characteristic peak position data, determine the target characteristic peak position prediction data and construct the restricted search interval of the target characteristic peak.

[0090] Further in S4, the steps of determining the target feature peak prediction data and constructing the restricted search interval for the target feature peak include:

[0091] The historical characteristic peak data are arranged in chronological order to obtain the historical characteristic peak sequence;

[0092] Based on temperature data and state of charge (SOC) data, determine the subset of historical characteristic peak data for the corresponding operating condition range from the historical characteristic peak sequence;

[0093] The peak position change sequence between historical characteristic peaks is calculated based on a subset of historical characteristic peak data, and the target characteristic peak prediction data is calculated by combining temperature data and state of charge (SOC) data.

[0094] The peak position offset range is determined based on historical characteristic peak position data and target characteristic peak position prediction data, and the restricted search interval of the target characteristic peak is determined based on the target characteristic peak position prediction data and the peak position offset range.

[0095] Specifically, using real-time collected temperature data State of Charge (SOC) data And historical characteristic peak position data obtained by identifying target characteristic peaks at historical moments. For input, where The historical data is indexed by its corresponding time sequence number. First, the historical characteristic peak data is analyzed. Arranged in ascending order of time, the historical characteristic peak sequence is obtained. ;in The number of valid time points for historical characteristic peak data, based on the current temperature data. State of Charge (SOC) data In the historical characteristic peak sequence Selected from those with temperatures at The interval, the state of charge (SOC) is in The corresponding data for each interval is used to obtain a subset of historical characteristic peak data for the corresponding operating condition interval. ;in For the preset temperature matching threshold, The preset SOC matching threshold, and All conditions were predetermined through lithium battery operating condition calibration tests. The number of valid data points in the subset of historical characteristic peak data, based on the subset of historical characteristic peak data. Calculate the peak position difference between adjacent data points to obtain the peak position change sequence. Peak position change sequence The Middle The expression for calculating each element is: ;in Combined with peak position change sequence Statistical trends, current temperature data With state of charge data Calculate target feature peak position prediction data Target feature peak position prediction data The calculation expression is as follows ;in Peak position change sequence The mean, For the pre-calibrated temperature compensation coefficient, For the pre-calibrated SOC compensation coefficient, For reference temperature, For reference state of charge, and All parameters are standard operating condition parameters determined by lithium battery calibration tests. The temperature compensation coefficient... With SOC compensation coefficient Calibration was performed using lithium battery lifecycle aging test data. The calibration process is as follows: Under steady-state conditions at different temperatures and SOCs, characteristic peak position data were measured. A multiple linear regression model was constructed with the characteristic peak position as the dependent variable and temperature and SOC as independent variables. The least squares method was used to solve the model. and The optimal estimate minimizes the root mean square error between the model's predicted peak position and the actual measured peak position. This is based on historical characteristic peak position sequences. The difference between each data point and the predicted peak position under the corresponding operating condition is used to calculate the peak position offset range. Peak position shift range The maximum absolute value of the historical difference is used to predict the peak position based on the target characteristic data. and peak position shift range Determine the restricted search interval for the target feature peak. The expression for the restricted search interval is: Compared to peak search across the entire frequency domain, this invention dynamically constructs a restricted search interval based on historical data and current operating conditions, compressing the search range from the entire frequency domain to a local frequency band. This approach not only significantly reduces computational complexity but, more importantly, effectively avoids spurious peak interference introduced by the full-frequency domain search by excluding frequency domain regions unrelated to the target feature peak. This ensures the feature peak identification results remain continuous in time, solving the problem of "intermittent and drastic jumps" in feature peak identification results in existing technologies. The obtained restricted search interval is used for subsequent target feature peak identification processing.

[0096] To further improve the accuracy of target characteristic peak prediction, this embodiment constructs a dynamic prediction model for characteristic peaks based on historical data. This model treats the temporal evolution of characteristic peaks as a dynamic process influenced by battery aging trends, temperature effects, and SOC effects, and its expression is:

[0097] ;

[0098] in: for Predicted peak value of target characteristic at any given time; The baseline trend term based on historical peak sequence is extracted using exponential smoothing or locally weighted regression methods; The peak shift term caused by battery aging is obtained by fitting the full life cycle data using a linear or power function model. For temperature compensation item The state-of-charge compensation term is implemented using polynomial fitting or table lookup. This is the random error term that follows a normal distribution with a mean of zero.

[0099] The parameters in the above model were obtained through offline calibration using lithium battery lifecycle aging test data. The calibration process employed the least squares method or Bayesian inference to determine the optimal coefficients for each compensation term, aiming to minimize the prediction error. In practical applications, the model parameters can be adaptively updated online based on real-time collected operational data to track parameter drift during battery aging, further improving the adaptability of peak position prediction.

[0100] S5. Identify the target feature peak within the restricted search range, extract the target feature peak parameters, and construct the target feature peak parameter time series; for parameter jumps or missing parameters in the target feature peak parameter time series under abnormal working conditions such as drift, overlap, and disappearance, perform continuous tracking and correction processing based on time-series correlation on the target feature peak parameter time series.

[0101] Furthermore, in S5, the step of identifying target feature peaks within the restricted search interval includes:

[0102] Based on the restricted search range of the target feature peak, extract the feature peak distribution data of the corresponding frequency range from the feature peak distribution data and form a feature peak distribution data sequence;

[0103] Traverse each data point in the characteristic peak distribution data sequence in frequency order, and determine the current data point as a candidate characteristic peak when the value of the current data point is greater than that of the adjacent data points.

[0104] Based on the frequency value and peak amplitude of the candidate feature peak, extract and filter the candidate feature peak parameters, and determine the target feature peak parameters;

[0105] The target feature peak parameters obtained at each time point are arranged in chronological order to construct a time series of target feature peak parameters.

[0106] Specifically, using a pre-constructed restricted search interval of the target feature peak. and the characteristic peak distribution data obtained through adaptive peak enhancement transformation. For input, These are the frequency values ​​of each frequency point corresponding to the characteristic peak distribution data. The frequency point number is determined based on the limited search interval. From characteristic peak distribution data Filter out frequency values satisfy All data points are selected and arranged in ascending order of frequency to form a characteristic peak distribution data sequence. , This represents the total number of data points after filtering. For the frequency values ​​of the corresponding data points, traverse the characteristic peak distribution data sequence in ascending frequency order. For each data point in the sequence, the sequence number is... intermediate data points, Determine the value of the current data point. Are they both greater than the values ​​of the previous adjacent data points? The value of the next adjacent data point By determining local maxima, when the determination result is yes, the current data point is identified as a candidate feature peak, and the frequency value corresponding to the candidate feature peak is recorded. and peak amplitude Based on the frequency values ​​corresponding to each candidate characteristic peak and peak amplitude Extract candidate feature peak parameters, including peak frequency, peak amplitude, and peak width. The peak width is calculated using the full width at half maximum (FWHM) of the candidate feature peak. The expression for the FWHM is as follows: , and The candidate feature peak amplitude decreased to the peak amplitude respectively. The right and left frequency values ​​corresponding to 50% of the peak amplitude are sorted according to the peak amplitude from high to low, and the peak amplitude values ​​are selected from those with a preset amplitude threshold. The candidate feature peak parameters are selected, and the selected parameters are determined as the target feature peak parameters, with a preset amplitude threshold. Pre-calibrate using the amplitude statistical characteristics of the characteristic peak distribution data.

[0107] When multiple candidate feature peak parameters are detected within a limited search interval at the same time point, it indicates that feature peak overlap exists within that frequency interval, making it impossible to directly determine the true feature peak through amplitude thresholding. In this case, a screening method based on temporal correlation is adopted: the absolute difference between each candidate feature peak parameter and the frequency value of the most recent valid time point in the historical feature peak sequence is calculated, and the candidate feature peak with the smallest difference is selected as the target feature peak parameter. This process is based on the principle of continuity of feature peaks in the time dimension, meaning that the frequency value of a feature peak will not change drastically between adjacent time points, and the temporal evolution of the true feature peak should remain smooth, while noise spurious peaks do not possess this characteristic. Through temporal correlation screening, the true feature peak can be effectively identified, and the corresponding time points are marked as peak overlap states, providing accurate parameter input for subsequent temporal correction.

[0108] The target feature peak parameters obtained at each consecutive time point are arranged in ascending order of time to construct a time series of target feature peak parameters. , Using time point numbers, the time series of target characteristic peak parameters is obtained. It is used for continuous tracking and sequence correction under subsequent abnormal operating conditions, as well as for the extraction of lithium battery health status characteristics.

[0109] Furthermore, in S5, the steps of performing continuous tracking processing and correction on the time series of target characteristic peak parameters include:

[0110] Traverse adjacent time points in the time series of target feature peak parameters in chronological order, and calculate the rate of change of peak position in the target feature peak parameters corresponding to adjacent time points;

[0111] When the peak position change rate exceeds the preset change range, the corresponding time point will be marked as peak position drift state;

[0112] When multiple target feature peak parameters are detected at the same time point and the difference between the frequency values ​​corresponding to each target feature peak parameter is less than the preset frequency range, the corresponding time point is marked as peak overlap state.

[0113] When no target feature peak parameter is detected within the restricted search range of the target feature peak, the corresponding time point is marked as the peak disappearance state;

[0114] Based on the target characteristic peak parameters at adjacent time points, interpolation correction or neighborhood smoothing is performed on the time series of target characteristic peak parameters marked as peak position drift state, peak overlap state, or peak disappearance state.

[0115] Specifically, the time series of target feature peak parameters obtained through target feature peak identification is used. and the limited search range of the pre-constructed target feature peaks. For input, where For time point number, Includes the peak frequency values ​​of the target feature peaks at each time point. Traverse the time series of target feature peak parameters in ascending order of time. For adjacent time points, the absolute difference between the peak position frequencies of the target feature peaks corresponding to adjacent time points is calculated as the peak position change rate. Peak position change rate The calculation expression is as follows ;in For the first The peak frequency values ​​of the target feature peaks at each time point. For the first The peak position frequency value of the target feature peak corresponding to each time point, and the calculated peak position change rate Compared with the preset peak position change threshold Comparison, when the peak position change rate Greater than the preset peak position change threshold At that time, the first Each time point is marked as a peak position drift state, and a preset peak position change threshold is set. The operating condition calibration test of the lithium battery throughout its entire life cycle is determined in advance, and the first... The target feature peak parameters at each time point are statistically analyzed. When multiple target feature peak parameters are detected at the same time point, and the minimum absolute difference between the peak frequency values ​​corresponding to each target feature peak parameter is less than a preset overlap frequency threshold, the following criteria are applied: When the time point is reached, it is marked as a peak overlap state, and a preset overlap frequency threshold is set. By pre-calibrating the characteristic peak frequency band distribution characteristics of the impedance spectrum of lithium batteries, the first The feature peak identification results at each time point are verified, and when within the limited search range of the target feature peak... When no valid target feature peak parameters are detected, the corresponding time point is marked as a peak disappearance state. Based on the marking results, target feature peak parameters are selected from consecutive valid time points before and after the abnormal time point. Corresponding correction processes are performed on the time series of target feature peak parameters marked as peak position drift, peak overlap, or peak disappearance states. For the time points in peak position drift and peak disappearance states, linear interpolation is used to obtain the corrected target feature peak parameters. The expression for linear interpolation is: ;in The most recent valid time point before the abnormal time point. The most recent valid time point after the abnormal time point. and These represent the peak position frequency values ​​of the target feature peaks at corresponding effective time points. For time points with peak overlap, neighborhood smoothing is applied to obtain the corrected target feature peak parameters. The expression for neighborhood smoothing is as follows: ;in To preset the smooth window length, By pre-calibrating the time series sampling density and then performing interpolation correction and neighborhood smoothing, the corrected target feature peak parameter time series is obtained. Corrected target feature peak parameter time series Used for subsequent extraction of lithium battery health characteristics and calculation of health status estimates.

[0116] S6. Extract health features from the corrected target characteristic peak parameter time series, and calculate the lithium battery health state SOH estimation result according to the preset mapping relationship between health features and SOH.

[0117] Furthermore, in S6, the health characteristics include at least one of the following: the rate of change of the peak position of the target characteristic peak, the rate of change of the peak amplitude of the target characteristic peak, and the rate of change of the peak width of the target characteristic peak.

[0118] Specifically, the corrected target characteristic peak parameter time series is obtained after continuous tracking processing and correction. For input, where The time series of the corrected target characteristic peak parameters is represented by the time point number. Includes the corrected peak position frequency values ​​of the target characteristic peak at each time point. , Corrected target characteristic peak-to-peak amplitude , Peak width of the corrected target feature peak From the corrected target characteristic peak parameter time series Extract at least one health feature, including the rate of change of the peak position of the target feature peak, the rate of change of the peak value amplitude of the target feature peak, the rate of change of the peak width of the target feature peak, and the rate of change of the peak position of the target feature peak. The calculation expression is as follows ;in The initial peak position frequency value of the target characteristic peak obtained by factory calibration under fresh battery conditions, and the rate of change of the peak value amplitude of the target characteristic peak. The calculation expression is as follows ;in The target characteristic peak amplitude is the initial peak value obtained from factory calibration under fresh battery conditions, and the target characteristic peak width variation rate is... The calculation expression is as follows ;in The initial peak width of the target feature peak obtained from factory calibration under fresh battery conditions is used to construct a health feature vector by combining at least one extracted health feature. Based on the mapping relationship between health characteristics obtained through pre-calibrated lithium battery life cycle aging tests and the state of health (SOH) of lithium batteries, the health characteristic vector is... The state of health (SOH) of the lithium battery is estimated by inputting the mapping relationship model. The mapping relationship model uses a preset mapping model, which can be selected from a linear fitting model, a neural network model, or a support vector machine regression model. For example, the expression for the linear fitting model is... ;in , , These are the pre-calibrated weighting coefficients corresponding to the rate of change of the peak position of the target characteristic peak, the rate of change of the peak value amplitude of the target characteristic peak, and the rate of change of the peak width of the target characteristic peak. Based on the calculated SOH estimation results, the obtained lithium battery health state SOH estimation results are used for battery health state assessment and charge / discharge control strategy adjustment of the battery management system.

[0119] In this embodiment, the mapping relationship model between health characteristics and SOH is constructed through the following steps:

[0120] Multiple sets of health feature vectors collected during the battery's full life cycle aging test The training dataset is composed of the corresponding SOH calibration values: ;in, The number of samples;

[0121] Based on the degree of nonlinearity in the relationship between features and SOH, one of the following model structures is selected:

[0122] (1) Linear regression model: This applies to situations where the characteristic and SOH have an approximately linear relationship. , , , Where is the regression coefficient and ϵ is the error term.

[0123] (2) Support Vector Regression Model: ;

[0124] in For the kernel function, a radial basis function can be used. ; For Lagrange multipliers, This is the bias term. This model is suitable for situations with nonlinear relationships and a moderate sample size.

[0125] (3) Neural Network Model: A feedforward neural network with one or more hidden layers is used, and its output layer expression is:

[0126] ;

[0127] in, For activation functions (such as ReLU or Tanh). and These are weights and biases, respectively. This represents the number of neurons in the hidden layer. This model is suitable for scenarios where there are complex interactions between features. Folded cross-validation is used to train the model and optimize its hyperparameters, with root mean square error (RMSE) and mean absolute error (MAE) as evaluation metrics.

[0128] ;

[0129] ;

[0130] in To determine the number of samples in the validation set, These are the model predictions. After training, the model parameters are stored in the battery management system for online SOH estimation.

[0131] In actual operation, when the battery completes a full charge-discharge cycle and meets the SOH calibration conditions, the newly acquired health characteristics and SOH data can be used as incremental samples to fine-tune the model online, so as to maintain the model's ability to track the battery aging characteristics.

[0132] Based on the calculated State of Health (SOH) estimation results, the obtained SOH estimation results of lithium batteries are used for battery health status assessment and charge / discharge control strategy adjustment in the battery management system.

[0133] Example 2: In the second embodiment of the present invention, the present invention provides a lithium battery health state estimation system based on electrochemical impedance spectroscopy characteristic peak tracking, such as... Figure 2 As shown, it includes the following modules:

[0134] The operating condition screening module is used to collect operating data, including voltage data, current data, temperature data, and state of charge (SOC) data, during the operation of lithium batteries. The module filters the collected operating data according to preset operating condition screening rules to obtain local analysis data that meets the screening criteria.

[0135] The impedance construction module is used to perform short-time frequency domain transformation on voltage and current data in local analysis data, calculate frequency domain impedance, and construct local electrochemical impedance spectra.

[0136] The peak enhancement module is used to perform adaptive peak enhancement transformation on the local electrochemical impedance spectrum to obtain characteristic peak distribution data;

[0137] The peak position prediction module is used to combine temperature data, state of charge (SOC) data, and historical characteristic peak position data to determine the target characteristic peak position prediction data and construct the restricted search interval of the target characteristic peak.

[0138] The peak tracking module is used to identify target feature peaks within a limited search range, extract target feature peak parameters, and construct a time series of target feature peak parameters; for abnormal conditions such as drift, overlap, and disappearance of target feature peaks, it performs continuous tracking processing and correction on the time series of target feature peak parameters.

[0139] The SOH estimation module is used to extract health features from the corrected target characteristic peak parameter time series and calculate the SOH estimation result of lithium battery health state based on the preset mapping relationship between health features and SOH.

[0140] During the long-term operation of power batteries in new energy vehicles, the complex and varied driving conditions cause the battery to operate under different temperature environments, charging and discharging currents, and states of charge. This results in continuous changes in the battery's internal electrochemical reaction characteristics, leading to alterations in its internal impedance characteristics over time. In actual operating environments, battery management systems typically rely on operating data such as voltage and current to monitor battery status. However, due to fluctuations in operating conditions, noise interference, and unstable operating states within this data, it is difficult to stably construct impedance information reflecting the battery's internal electrochemical characteristics. Consequently, it is challenging to consistently and stably identify changes in the positions of key characteristic peaks in the impedance spectrum, thus affecting the accuracy and stability of battery health status estimation. To address these issues, this invention employs a lithium battery health status estimation system based on electrochemical impedance spectroscopy characteristic peak tracking, the structure of which is as follows: Figure 2 As shown. The specific implementation process of this system is as follows:

[0141] First, the operating condition screening module acquires voltage, current, temperature, and state of charge (SOC) data collected by the battery management system during the operation of the lithium battery. The acquired operating data is then screened according to preset operating condition screening rules. Relatively stable local analysis data is extracted from the continuous operating data, thereby reducing the impact of operating condition fluctuations on subsequent analysis results. This allows the subsequent impedance calculation process to be based on stable operating state data, improving the reliability of subsequent analysis results.

[0142] Subsequently, the voltage and current data in the local analysis data are transformed into short-time frequency domains using the impedance construction module, and the frequency domain impedance is calculated based on the transformed frequency domain voltage and frequency domain current, thereby constructing a local electrochemical impedance spectrum. This allows the electrochemical reaction characteristics of the battery within a specific time window to be characterized in the frequency domain, providing basic data for subsequent characteristic peak analysis.

[0143] After obtaining the local electrochemical impedance spectrum, an adaptive peak enhancement transformation is performed on the impedance spectrum through the peak enhancement module. This enhances the characteristic peaks in the impedance spectrum that are related to the charge transfer and diffusion processes inside the battery, thereby improving the identifiability of the characteristic peaks in the impedance spectrum and obtaining characteristic peak distribution data, providing a clearer data structure for subsequent characteristic peak identification.

[0144] Based on this, the peak position prediction module combines current temperature data, state of charge (SOC) data, and historical characteristic peak position data to predict the changing trend of the target characteristic peak and construct the corresponding restricted search interval. This concentrates the characteristic peak identification process within the range where characteristic peaks may appear, thereby reducing false detections caused by full-band search and improving the efficiency of characteristic peak localization.

[0145] Subsequently, the peak tracking module performs target feature peak identification on the feature peak distribution data within the limited search range, extracts the corresponding target feature peak parameters and constructs the target feature peak parameter time series. At the same time, for abnormal operating conditions such as peak position drift, peak overlap or peak disappearance that may occur during long-term operation of the target feature peak, the target feature peak parameter time series is continuously tracked and corrected to ensure that the feature peak parameters remain continuous and stable in the time dimension, thereby ensuring that the feature peak changes can truly reflect the internal state change process of the battery.

[0146] Finally, the SOH estimation module extracts health features reflecting battery degradation from the corrected target characteristic peak parameter time series, and calculates the SOH estimation result of the lithium battery health state based on the preset mapping relationship between health features and SOH. This enables continuous assessment of the power battery health state, allowing the battery management system to monitor and manage the battery operating status based on the estimation results, thereby improving battery operating safety and extending the battery system's service life.

[0147] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy, characterized in that, Includes the following steps: S1. Collect operating data of lithium battery during operation, including voltage data, current data, temperature data and state of charge (SOC) data. Filter the collected operating data according to preset operating condition filtering rules to obtain local analysis data that meets the filtering conditions. S2. Perform short-time frequency domain transformation on the voltage and current data in the local analysis data, calculate the frequency domain impedance and construct the local electrochemical impedance spectrum; S3. Perform adaptive peak enhancement transformation on the local electrochemical impedance spectrum to obtain characteristic peak distribution data; S4. Combining temperature data, state of charge (SOC) data, and historical characteristic peak position data, determine the target characteristic peak position prediction data and construct the restricted search interval of the target characteristic peak. S5. Identify the target feature peak within the restricted search range, extract the target feature peak parameters, and construct the target feature peak parameter time series; for parameter jumps or missing parameters in the target feature peak parameter time series under abnormal working conditions such as drift, overlap, and disappearance, perform continuous tracking and correction processing based on time-series correlation on the target feature peak parameter time series. S6. Extract health features from the corrected target characteristic peak parameter time series, and calculate the lithium battery health state SOH estimation result based on the preset mapping relationship between health features and SOH.

2. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 1, characterized in that, In S1, the step of filtering the collected operating data according to preset operating condition filtering rules includes: The running data is divided into multiple time windows according to a preset time length; Calculate the rate of change of current data, the rate of change of voltage data, the amount of change of temperature data, and the amount of change of state of charge (SOC) data within each time window. The rate of change of current data is compared with a preset current rate of change threshold, the rate of change of voltage data is compared with a preset voltage rate of change threshold, the amount of change of temperature data is compared with a preset temperature amount of change threshold, and the amount of change of state of charge (SOC) data is compared with a preset SOC change threshold. When the rate of change of current data is less than the preset threshold for the rate of change of current, the rate of change of voltage data is less than the preset threshold for the rate of change of voltage, the amount of change of temperature data is less than the preset threshold for the amount of change of temperature, and the amount of change of state of charge (SOC) data is less than the preset threshold for the amount of change of SOC, the running data within the corresponding time window will be determined as local analysis data.

3. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 1, characterized in that, In S2, the step of performing short-time frequency domain transformation on the voltage and current data in the local analysis data includes: The voltage and current data in the local analysis data are divided into multiple time windows according to a preset time length; Within each time window, the voltage and current data are weighted using a window function. Frequency domain transformations are performed on the weighted voltage data and the weighted current data respectively to obtain the frequency domain representations of the voltage data and the current data for the corresponding time windows; The frequency domain representations of voltage and current data obtained from each time window are arranged in chronological order to form short-time frequency domain conversion results for voltage and current data.

4. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 3, characterized in that, In S2, the step of calculating the frequency domain impedance and constructing the local electrochemical impedance spectrum includes: Calculate the impedance value at the corresponding frequency point based on the frequency domain representation of voltage data and current data; The impedance value is decomposed to obtain the real part and the imaginary part of the impedance; The real part and imaginary part of the impedance are combined with the corresponding frequency values ​​to form an impedance frequency sequence; Local electrochemical impedance spectroscopy is constructed based on impedance frequency sequences.

5. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 4, characterized in that, In S3, the step of performing adaptive peak enhancement transformation on the local electrochemical impedance spectrum includes: The impedance amplitude sequence is extracted from the impedance frequency sequence of the local electrochemical impedance spectrum, and the impedance amplitude sequence is smoothed. The peak response sequence is obtained by calculating the rate of change and the difference between adjacent frequency points based on the smoothed impedance amplitude sequence. Calculate the weighting coefficients corresponding to each frequency point based on the impedance amplitude sequence; The enhanced peak response sequence is obtained by combining the peak response sequence with the weighting coefficients, and characteristic peak distribution data is generated based on the enhanced peak response sequence.

6. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 1, characterized in that, In S3, the adaptive peak enhancement transformation enhances the intensity of characteristic peak responses related to charge transfer and diffusion processes in the electrochemical impedance spectroscopy by using a second-order difference response based on local curvature weighting.

7. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 1, characterized in that, In S4, the step of determining the target feature peak prediction data and constructing the restricted search interval of the target feature peak includes: The historical characteristic peak data are arranged in chronological order to obtain the historical characteristic peak sequence; Based on temperature data and state of charge (SOC) data, determine the subset of historical characteristic peak data for the corresponding operating condition range from the historical characteristic peak sequence; The peak position change sequence between historical characteristic peaks is calculated based on a subset of historical characteristic peak data, and the target characteristic peak prediction data is calculated by combining temperature data and state of charge (SOC) data. The peak position offset range is determined based on historical characteristic peak position data and target characteristic peak position prediction data, and the restricted search interval of the target characteristic peak is determined based on the target characteristic peak position prediction data and the peak position offset range.

8. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 7, characterized in that, In S5, the step of identifying the target feature peak within the restricted search interval includes: Based on the restricted search range of the target feature peak, extract the feature peak distribution data of the corresponding frequency range from the feature peak distribution data and form a feature peak distribution data sequence; Traverse each data point in the characteristic peak distribution data sequence in frequency order, and determine the current data point as a candidate characteristic peak when the value of the current data point is greater than that of the adjacent data points. Based on the frequency value and peak amplitude of the candidate feature peak, extract and filter the candidate feature peak parameters, and determine the target feature peak parameters; The target feature peak parameters obtained at each time point are arranged in chronological order to construct a time series of target feature peak parameters.

9. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 8, characterized in that, In S5, the step of performing continuous tracking processing and correction on the time series of target feature peak parameters includes: Traverse adjacent time points in the time series of target feature peak parameters in chronological order, and calculate the rate of change of peak position in the target feature peak parameters corresponding to adjacent time points; When the peak position change rate exceeds the preset change range, the corresponding time point is marked as a peak position drift state; When no target feature peak parameter is detected within the restricted search range of the target feature peak, the corresponding time point is marked as the peak disappearance state; When multiple candidate feature peak parameters are detected at the same time point, the target feature peak parameter is selected and determined based on the temporal correlation between each candidate feature peak parameter and the historical feature peak position, and the corresponding time point is marked as peak overlap state; Based on the target characteristic peak parameters of adjacent time points, interpolation correction or neighborhood smoothing is performed on time points marked as peak position drifting state or peak disappearance state.

10. The method for estimating the health status of lithium batteries based on characteristic peak tracking of electrochemical impedance spectroscopy according to claim 1, characterized in that, In S6, the health characteristics include at least one of the following: the change in the peak position of the target characteristic peak, the change rate of the peak value amplitude of the target characteristic peak, and the change in the peak width of the target characteristic peak.

Images