Method and system for detecting lithium precipitation from a battery

Abstract

The application discloses a battery lithium precipitation detection method and system, and belongs to the technical field of battery state detection. The method comprises the following steps: after charging of a target battery is completed, end voltage data in a static period are collected and filtered and denoised to obtain a voltage relaxation curve; relaxation time distribution inversion is performed on the voltage relaxation curve to obtain a relaxation time distribution spectrum; feature parameters related to lithium precipitation are extracted based on the distribution spectrum, the feature parameters comprising a newly-added feature peak in a specific interval, parameter change of an original feature peak and global similarity; temperature compensation and state-of-charge normalization processing are performed on the feature parameters based on temperature and state-of-charge data in the static period; and finally, whether lithium precipitation occurs in the battery is determined based on the processed feature parameters. The application realizes lossless, accurate and high-robustness online detection of the lithium precipitation state of a lithium ion battery by converting the voltage relaxation curve into a relaxation time distribution spectrum with clear physical meaning and combining a working condition compensation mechanism.

Description

Technical Field

[0001] This invention relates to the field of battery state detection technology, specifically to a method and system for detecting lithium plating in batteries. Background Technology

[0002] Under conditions such as fast charging, low-temperature charging, or overcharging, the negative electrode potential of a lithium-ion battery may fall below the deposition potential of metallic lithium, causing lithium ions to deposit on the negative electrode surface in the form of metallic lithium. This phenomenon is called "lithium plating." Lithium plating not only causes a permanent loss of reversible capacity and reduces cycle life, but more seriously, the deposited metallic lithium may grow into dendrites, piercing the separator and causing an internal short circuit, ultimately leading to safety accidents such as thermal runaway. Therefore, achieving rapid, non-destructive, and accurate detection of lithium plating in lithium-ion batteries is crucial for ensuring the safe operation of battery systems, optimizing charging strategies, and extending battery life.

[0003] Currently, lithium plating detection methods for lithium-ion batteries are mainly divided into two categories: destructive testing and non-destructive testing. Destructive testing typically requires disassembling the battery and directly observing the lithium plating morphology on the negative electrode surface through visual inspection, scanning electron microscopy, or nuclear magnetic resonance. Although intuitive and accurate, its destructive nature and time lag prevent it from being applied to online detection. Non-destructive testing methods can be performed without damaging the battery structure, making them a focus of current research and application. Among these, analysis methods based on voltage relaxation characteristics have attracted considerable attention due to their advantages such as ease of operation, lack of complex excitation signals, and ease of deployment in battery management systems.

[0004] Existing lithium plating detection methods based on voltage relaxation, such as a lithium iron phosphate battery lithium plating detection method, obtain the open-circuit voltage curve of a fully charged battery during the relaxation phase, and then perform differential calculations on this curve to obtain a voltage difference curve. Next, based on the ratio of the time corresponding to the peak point of the voltage difference curve to the time when the voltage is fully released, a relaxation time normalization parameter is defined. Finally, based on an empirical relationship model between this parameter and the true value of lithium plating, the amount of lithium plating is estimated.

[0005] However, in practical engineering applications, the above-mentioned existing technical solutions still have the following defects and shortcomings: First, the physical interpretability of the detected features is insufficient. The criterion upon which this method relies—the relaxation time normalization parameter—is a purely mathematical feature extracted from the macroscopic voltage difference curve. This feature lacks a direct and clearly physically meaningful correlation with the physical mechanism of lithium plating, a microscopic electrochemical process. Changes in the voltage curve shape are influenced by multiple factors (such as temperature, state of charge, and degree of aging), and a single normalized time parameter cannot accurately and specifically reflect the lithium plating process itself, leading to insufficient physical basis for the detection results and a high risk of misjudgment.

[0006] Secondly, it has poor adaptability to complex working conditions. While this method mitigates some of the impact through time ratio normalization, its effectiveness has not been rigorously validated across operating conditions. Particularly under conditions of drastic temperature changes or inconsistent state of charge (SOC), the battery's voltage relaxation behavior exhibits significant systematic shifts. Existing solutions lack both a systematic temperature compensation mechanism and consideration of the influence of different SOCs on relaxation characteristics. This results in a substantial decrease in detection accuracy and insufficient robustness when facing the complex and variable operating conditions of actual vehicles or energy storage systems.

[0007] Third, it is impossible to effectively distinguish between lithium and normal aging. Normal battery aging (such as thickening of the solid electrolyte interface film and loss of active materials) can also cause changes in the voltage relaxation curve morphology. This method fails to provide a mechanism to decouple and differentiate between lithium-induced characteristic changes and those caused by normal aging. When a battery ages due to long-term cycling, its voltage differential curve may also change, potentially leading existing technologies to misinterpret it as lithium plating, causing false positive alarms and reducing the specificity and reliability of the test results.

[0008] Fourth, quantitative estimation has limited accuracy and lacks uncertainty assessment. This method relies on a single empirical fitting model between the relaxation time normalization parameter and the true value of lithium plating. The model's development process does not adequately consider the influence of key factors such as battery system, temperature, and rate capability on the mapping relationship, and its generalizability and accuracy in quantitative estimation need improvement. Furthermore, this scheme only outputs an estimated value of lithium plating without providing any information about the confidence interval of this estimate, making it difficult for the battery management system to assess the reliability of the result and failing to provide comprehensive data support for subsequent risk decisions.

[0009] In summary, there is an urgent need for a lithium-ion battery lithium plating detection method that has clear physical meaning, strong adaptability to operating conditions, can effectively distinguish the effects of aging, and can provide high-precision quantitative results and confidence assessment, in order to overcome the many bottlenecks existing in the current technology. Summary of the Invention

[0010] The purpose of this application is to overcome the shortcomings of the above-mentioned background technology and provide a method and system for detecting lithium plating in batteries.

[0011] The technical solution of this application is: a method for detecting lithium plating in batteries, comprising: After the charging process of the target battery is completed, the terminal voltage data of the battery during the resting period is collected and filtered and denoised to obtain the voltage relaxation curve. The relaxation time distribution spectrum is obtained by inverting the voltage relaxation curve; Obtain the first reference spectrum of a healthy battery of the same system as the target battery under reference conditions; Based on the relaxation time distribution spectrum and the first reference spectrum, feature parameters are extracted that include at least whether the relaxation time distribution spectrum has a new feature peak relative to the first reference spectrum within the first time constant interval. Acquire the temperature data of the battery during the resting period, and perform temperature compensation processing on the feature parameters based on the temperature data; Acquire the state of charge (SOC) data of the battery at the initial moment of rest, and perform SOC normalization on the feature parameters based on the SOC data. Based on the characteristic parameters after temperature compensation and state of charge normalization, it is determined whether lithium plating has occurred in the battery.

[0012] According to the battery lithium plating detection method provided in this application, the relaxation time distribution spectrum is obtained. g( ) The methods include: in: —The voltage relaxation curve obtained from the data acquisition; —Relaxation time distribution spectrum; —Relaxation time constant; —Open circuit voltage; The integral model is solved using a regularized inversion algorithm to obtain the relaxation time distribution spectrum. The regularization inversion algorithm includes Tikhonov regularization or Fourier transform filtering deconvolution.

[0013] According to the battery lithium plating detection method provided in this application, the characteristic parameters further include: The parameter changes of the original characteristic peaks in the relaxation time distribution spectrum relative to the first reference spectrum; and, The global similarity between the relaxation time distribution spectrum and the first reference spectrum.

[0014] According to the battery lithium plating detection method provided in this application, the method for temperature compensation processing of characteristic parameters based on temperature data includes: A reference relaxation experiment of a healthy battery with the same system as the target battery was performed in advance. After charging to the same state of charge at the standard charging rate under different temperature conditions, the voltage relaxation curves after the charging was completed were collected. The relaxation time distribution of the collected voltage relaxation curves was inverted to obtain the first reference spectrum library under different temperature conditions. Based on the temperature data, a first reference spectrum matching the current temperature is generated by calling or interpolating from the first reference spectrum library; The first reference spectrum is used as the benchmark for comparing the parameter changes in the feature parameters with the global similarity.

[0015] According to the battery lithium plating detection method provided in this application, the method for normalizing the characteristic parameters based on the state of charge data includes: Determine whether the state of charge data at the initial resting time is within a preset high state of charge range. If yes, proceed directly to the next steps; otherwise: A reference relaxation experiment of a healthy battery with the same system as the target battery was performed in advance. After charging to different states of charge under the same temperature conditions, the voltage relaxation curves of each state of charge were collected. The relaxation time distribution of the collected voltage relaxation curves was inverted to obtain a second reference spectrum library under different states of charge. Based on the second reference spectral library, the sensitivity of the characteristic parameters to changes in state of charge is calculated; Using the aforementioned sensitivity, the feature parameters extracted under the current state of charge are normalized to their corresponding values ​​under a standard state of charge.

[0016] According to the lithium plating detection method for batteries provided in this application, the method for determining whether lithium plating has occurred in the battery includes: if a new characteristic peak higher than the noise threshold is detected within the first time constant interval, it is determined that lithium plating has occurred; If no new characteristic peak is detected, but the parameter change of the original characteristic peak exceeds a preset threshold, it is determined that the battery is aging normally.

[0017] According to the battery lithium plating detection method provided in this application, when lithium plating is determined to have occurred, the peak area of ​​the newly added characteristic peak is calculated; Based on the pre-calibrated quantitative mapping model between the peak area of ​​the newly added characteristic peak and the lithium plating capacity, the estimated value of the lithium plating capacity is output according to the calculated peak area.

[0018] According to the battery lithium plating detection method provided in this application, the method for calculating the peak area of ​​the newly added characteristic peak includes: calculating the peak area of ​​the newly added characteristic peak according to the following formula. in: S—Peak area of ​​newly added characteristic peaks; —Relaxation time distribution spectrum; —The lower limit of the first time constant interval; —The upper limit of the first time constant interval.

[0019] According to the lithium plating detection method provided in this application, the method for calibrating the quantitative mapping model between the peak area of ​​the newly added characteristic peak and the lithium plating capacity includes: using a three-electrode cell of the same system as the target battery, charging under different operating conditions and inducing lithium plating to different degrees; Monitor the negative electrode potential of the battery cell to determine the occurrence of lithium plating and calculate the actual lithium plating capacity; Extract the peak area of ​​the newly added characteristic peaks; Establish a mapping relationship with the newly added characteristic peak area as the independent variable and the actual lithium plating capacity as the dependent variable.

[0020] This application also relates to a battery lithium plating detection system, including: The data acquisition module is used to collect terminal voltage and temperature data during the resting period after the target battery has finished charging. The data preprocessing module is used to process the terminal voltage data to obtain the voltage relaxation curve; The inversion calculation module is used to invert the relaxation time distribution of the voltage relaxation curve to obtain the relaxation time distribution spectrum. The feature extraction module is used to extract, based on the relaxation time distribution spectrum, at least the feature parameters including whether there is a new feature peak in the relaxation time distribution spectrum relative to the first reference spectrum within a first time constant interval. The compensation and normalization module is used to perform temperature compensation and state of charge normalization processing on the feature parameters based on the temperature data and the state of charge data at the start of the resting period. The lithium plating determination module is used to determine whether lithium plating has occurred in the battery based on the processed characteristic parameters.

[0021] The advantages of this application are as follows: 1. This application relates to a method for detecting lithium plating in batteries. This application transforms the collected terminal voltage data (macroscopic electrochemical signal) into a relaxation time distribution spectrum (DRT spectrum) through mathematical inversion. Each peak in this spectrum corresponds to a specific electrochemical or physical process inside the battery (such as charge transfer, diffusion, SEI film impedance, etc.). This correspondence between signal and process allows for diagnosis at the level of electrochemical impedance mechanism, laying the foundation for high reliability and strong interpretability of the detection results.

[0022] This application actively collects these two operating condition parameters and corrects the characteristic parameters, eliminating the interference of temperature fluctuations and SOC differences on the detection results. This ensures that the same set of judgment logic has consistent accuracy and physical benchmarks under different operating conditions such as low temperature and high temperature, full charge and half charge, greatly enhancing the engineering practicality and robustness of the method.

[0023] This application utilizes resting period data to achieve non-invasive online diagnostics. Compared to detection during charging (where high current interference is strong and polarization voltage masks lithium plating signals), voltage changes during resting are driven solely by thermodynamic equilibrium processes, resulting in an extremely high signal-to-noise ratio. This means that this method does not require any additional expensive reference electrodes (three-electrode system), ultrasonic probes, or special excitation current circuits; it can perform in-depth diagnostics solely based on the existing voltage, temperature, and SOC data from the battery management system. This gives the detection method of this application a significant advantage: zero incremental hardware cost and over-the-air upgrade and deployment capability.

[0024] 2. This application specifies the concrete inversion mathematical expression for obtaining the relaxation time distribution spectrum g(τ). This application provides a concrete integral relationship between the voltage relaxation curve V(t) and the relaxation time distribution spectrum g(τ). This not only clarifies the mathematical expression but, more importantly, establishes the physical model basis for the technical solution. The formula shows that the macroscopic voltage response is considered as a linear superposition of countless relaxation processes (RC parallel links) with different time constants τ. g(τ) represents the intensity distribution of these processes. By defining this core algorithmic logic, subsequent quantitative calculations of peak area and peak position have rigorous mathematical and physical support, avoiding the arbitrariness and errors caused by simply relying on empirical algorithms (such as simple differential differentiation).

[0025] Although the direct analytical solution to this expression is a deconvolution, once the model is defined, implementers can use mature regularized inversion algorithms such as Tikhonov regularization and Fourier transform filtering to solve for g(τ). Compared to directly calculating the first or second derivative of V(t) (which would drastically amplify measurement noise), the inversion process based on this integral model naturally includes suppression of high-frequency noise. It can robustly extract weak but crucial relaxation mode information from noisy measured voltage data, ensuring that the small characteristic peaks related to lithium plating are not submerged by noise.

[0026] 3. This application constructs a multi-dimensional judgment logic to achieve precise separation between lithium plating and normal aging. Normal battery aging only changes the morphology of existing peaks, while the dissolution and re-intercalation of metallic lithium introduced by lithium plating generates entirely new interfacial reaction channels, which manifest as new peaks in a specific time constant interval (the first time constant interval) on the DRT spectrum. This is direct evidence of lithium plating. Changes in the original peak parameters correspond to aging monitoring indicators. Monitoring is conducted on peak broadening or peak frequency shifts caused by SEI film thickening, material particle breakage, etc.

[0027] This combined feature extraction scheme decouples lithium plating event detection from long-term health trend tracking. Its advantage lies in the fact that when battery capacity decays due to normal cycling, the system can identify the aging state through changes in the original peaks, without falsely triggering lithium plating alarms due to overall distortion of the voltage curve. This fundamentally solves the problem of traditional voltage threshold methods generating more false alarms with battery aging.

[0028] 4. This application eliminates the coupling effect of temperature and kinetics by establishing a dynamic benchmark for temperature zones. Both lithium plating and lithium-ion diffusion processes are governed by the Arrhenius equation and are extremely sensitive to temperature. The DRT spectrum morphology of the same healthy battery at 25°C and 0°C may be completely different (e.g., the diffusion tail peak shifts significantly to the right and increases at low temperatures). If the spectrum at 25°C is always used as the normal standard to compare with the measured data at 0°C, feature extraction will inevitably be distorted. The method in this application establishes a first reference spectrum library of batteries of the same system at different temperatures in advance, and calls the corresponding benchmark spectrum based on interpolation according to the measured temperature during use. Regardless of whether the battery is charged in the hot summer or the cold winter, the system can find a healthy benchmark that matches its thermodynamic state. This ensures that the calculation of the two key criteria, feature parameter changes and global similarity, is based on the correct comparison origin, avoiding baseline shift errors caused by temperature drift.

[0029] 5. This application constructs a strategy for optimizing the detection window and correcting for low-sensitivity regions. This application specifies an optimized detection logic: prioritizing detection in the high SOC range. This is because lithium plating typically occurs at the end of charging (high SOC), and at high SOC, the graphite anode is in a lithium-rich state. The difference in electrochemical activity between the deposited metallic lithium and the electrolyte and graphite is most pronounced during relaxation. Simultaneously, this application provides a remedial solution for low SOC situations: sensitivity coefficient normalization. Its advantage is that in reality, it's impossible to guarantee a full charge every time the vehicle is parked. If fast charging ends at 60% SOC, the system can still calculate the characteristic decay coefficient at that SOC using a second reference spectral library, effectively converting the measured peak height to the peak height at full charge. This processing ensures the continuity of the lithium plating determination logic. Even if charging is not full, the system can provide an equivalent, comparable risk score, rather than directly abandoning detection or outputting an unreliable result.

[0030] 6. This application clarifies that the addition of new characteristic peaks is a necessary condition for lithium plating, while changes in existing peak parameters are a sufficient characterization of aging. This logical classification has extremely high exclusivity. If a battery has undergone low-temperature, high-rate charging but has not plating lithium, its DRT spectrum may show that the original diffusion peaks have become wider and higher. In this case, relying solely on a single characteristic value may trigger an over-limit alarm. However, according to the detection method of this application, the system will check whether new peaks are generated. If there are no new peaks, and only changes in the original peaks, it is determined to be normal polarization or aging. This judgment logic greatly reduces false alarms caused by increased internal resistance due to low-temperature charging or aging, enabling the lithium plating early warning system to truly possess the low false alarm rate required for industrial applications.

[0031] 7. This application defines a scheme for quantitatively estimating lithium plating capacity after lithium plating occurs. Upgrading from qualitative early warning to quantitative risk assessment, simply knowing that lithium plating has occurred is insufficient. Trace amounts of lithium plating may be completely reversibly eliminated in the next discharge, while severe lithium plating will form dead lithium and puncture the separator. This method achieves graded quantification of lithium plating severity by calculating peak area and mapping it to capacity. The onboard BMS can make differentiated decisions based on this capacity estimate: for trace lithium plating: log recording and appropriately reducing the next charging current; for moderate lithium plating: actively limiting the upper limit of charging power and issuing a maintenance reminder; for severe lithium plating, immediately limiting power and issuing a safety warning. This improvement upgrades the detection function from a binary switch of presence / absence to an analog meter of quantity, providing crucial data support for intelligent battery operation and maintenance.

[0032] 8. This application provides a clear and reproducible quantitative index algorithm. By limiting the integration interval [τ1, τ2], a specific mathematical calculation method for the area of ​​the lithium plating characteristic peak is defined. This avoids calculation disputes caused by peak overlap or baseline drift and ensures the consistency of calculation results between different detection systems or different algorithm implementations.

[0033] 9. This application establishes a precise calibration system based on first-principles calculations. It explicitly uses a three-electrode battery cell as the calibration tool. In the three-electrode system, the voltage of the negative electrode relative to the lithium metal can be monitored independently. Once the negative electrode potential drops below 0V, the onset of lithium deposition can be accurately captured, and the true capacity of deposited lithium can be precisely calculated through coulomb integration. Compared to visually observing the degree of lithium deposition after disassembling the battery (which is imprecise and highly subjective), the three-electrode electrochemical calibration method establishes a direct physical mapping between the DRT spectral signal and the electrochemical capacity of lithium deposition. This ensures the high accuracy and scientific validity of the mapping model. The confidence level of the lithium deposition capacity estimate output by this model is much higher than that of empirical statistical models.

[0034] 10. This application also defines a modular architecture for a battery lithium plating detection system. The hardware and software are decoupled, supporting flexible deployment (compatible with edge computing and cloud computing). The system architecture clearly divides into six modules: data acquisition, preprocessing, inversion calculation, feature extraction, compensation and normalization, and judgment. This system does not rely on disassembly or electrode implantation; the inputs are only voltage, temperature, and SOC, three signals that can be acquired at the cell level, module level, and overall battery pack level. Therefore, this detection system can be used for factory sorting and testing of individual battery cells as well as for after-sales health diagnosis of vehicle battery packs. This cross-level universality is the biggest engineering advantage of this system solution compared to laboratory electrochemical workstation testing solutions. Attached Figure Description

[0035] Figure 1 : Flowchart of the lithium plating detection method for batteries in this application. Detailed Implementation

[0036] The embodiments of this application are described in detail below, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0037] In the description of this application, it should be understood that the terms "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0038] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0039] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] This application relates to a method for detecting lithium plating in batteries. This application uses relaxation time distribution inversion and operating condition compensation to detect lithium plating in batteries. The detection method of this application does not require disassembling the battery or adding additional sensing devices. It can achieve accurate determination of lithium plating by only using the normal operating data collected by the battery management system.

[0041] The battery lithium plating detection method of this application, such as Figure 1 As shown, follow these steps: Step S110: Data acquisition and preprocessing.

[0042] After the charging process of the target battery (such as a single lithium-ion battery cell, module, or entire pack) is completed, it enters a resting period. During this resting period, the battery management system continuously collects terminal voltage data of the battery over time at a preset sampling frequency (e.g., 1Hz). V raw (t) and temperature data T(t) Simultaneously, the state of charge data at the start of the resting period were recorded. SOC start .

[0043] The raw voltage data is filtered and denoised, for example, using moving average filtering or low-pass filtering, to eliminate electromagnetic interference and quantization noise, resulting in a smooth voltage relaxation curve. V(t) .

[0044] Step S120: Relaxation time distribution inversion.

[0045] Using a mathematical inversion algorithm, the voltage relaxation curve V(t) in the time domain is transformed to the time constant domain to obtain the relaxation time distribution spectrum. g(τ) .

[0046] Step S130: Extraction of lithium plating-related characteristic parameters.

[0047] Based on the obtained relaxation time distribution spectrum g(τ) The characteristic parameters related to lithium plating are extracted. These characteristic parameters include at least the presence of new characteristic peaks within a specific first time constant interval, changes in the morphological parameters of existing characteristic peaks, and global spectral similarity.

[0048] Step S140: Temperature-based compensation processing.

[0049] The temperature data T collected in step S110 is obtained, and the feature parameters extracted in step S130 are subjected to temperature compensation processing based on the pre-established temperature-reference spectrum mapping relationship.

[0050] Step S150: Normalization based on state of charge.

[0051] Obtain the state of charge data recorded in step S110 at the start of the resting period. SOC start The characteristic parameters are normalized by state of charge to eliminate differences in charging cutoff. SOC The resulting baseline deviation.

[0052] Step S160: Lithium plating determination.

[0053] Based on the characteristic parameters after temperature compensation and state of charge normalization, a preset logic judgment rule is executed to output a conclusion on whether lithium plating has occurred in the battery.

[0054] The core of the detection method in this application lies in decoupling the macroscopic voltage decay phenomenon into a microscopic electrochemical process spectrum through DRT inversion. Lithium deposition, in its electrochemical essence, belongs to the deposition of metallic lithium on the graphite anode surface and subsequent dissolution and re-intercalation reactions, a process with a specific relaxation time constant. Through the transformation of step S120, this process... g(τ) The spectrum shows a new characteristic peak independent of the original electrochemical processes (such as SEI film impedance and charge transfer). Compared with traditional methods that only observe the slope of the voltage plateau, this method directly captures the true situation of lithium plating, significantly improving the signal-to-noise ratio and physical interpretability of the detection. At the same time, the introduction of compensation mechanisms in steps S140 and S150 ensures the engineering applicability of this method in all climates and the entire charging range.

[0055] In some embodiments of this application, the relaxation time distribution inversion algorithm described in step S120 is defined in detail.

[0056] Obtain the relaxation time distribution spectrum g( ) The methods include: in: —The voltage relaxation curve obtained from the data acquisition; —Relaxation time distribution spectrum; —Relaxation time constant; —Open circuit voltage.

[0057] The collected voltage relaxation curve V(t) As the object of analysis, the voltage relaxation process is decomposed into a superposition of multiple exponential decay processes with different time constants through relaxation time distribution inversion, thus obtaining the relaxation time distribution spectrum. g (t) It satisfies the above integral equation.

[0058] V(t) For the acquired and smoothed terminal voltage data; V ocv This represents the battery's fully balanced open-circuit voltage, due to the limited resting time. V ocvIt is usually used as an unknown parameter in the solution; τ is the relaxation time constant. g(τ) This is the relaxation time distribution function to be determined.

[0059] Since directly solving the above integral equation analytically is an ill-conditioned problem, even small voltage measurement noise can cause... g (t) The solution exhibits significant oscillations. This embodiment employs the Tikhonov regularized Fourier transform deconvolution method for robust solution. The specific steps are as follows: Variable substitution and discretization: Let f=1 / τ The characteristic frequency is represented by logarithmic frequency coordinates. z=ln(f) Time-domain response V(t) Expressed in matrix form V=A*γ+ε ,in A For the kernel matrix, c The discretized distribution vector, e This is the noise term.

[0060] Construct the objective function: Minimize the weighted sum of the residual sum of squares and the regularization penalty term: min{||A*γ、(V、V ocv )|| 2 +λ*||L*γ|| 2 } Wherein, λ is the regularization parameter, which controls the smoothness (automatically optimized by the L-curve method or generalized cross-validation); L is the second-order difference operator matrix, used to penalize the severe jitter of the γ vector, ensuring the smoothness and physical rationality of the spectrum.

[0061] Solving for nonnegativity constraints: Due to the intensity of the relaxation time distribution g(τ) The impedance contribution representing the physical process should always be non-negative. This embodiment uses non-negative least squares (NNLS) combined with the above regularization framework for iterative solution to ensure the output... g(τ) All peaks in the spectrum are positive to avoid the appearance of physical meaningless negative peak artifacts.

[0062] This embodiment clarifies that the voltage response is a superposition integral of countless RC relaxation processes and provides a mathematical means to address the ill-conditioned nature of this inverse problem. By introducing a regularization term, the algorithm can accurately distinguish overlapping electrochemical processes with similar time constants while suppressing measurement noise. For example, the lithium dissolution peak caused by lithium plating (time constant approximately 10) 0 ~10 1 s) and the solid-state diffusion tail peak of the graphite anode (time constant approximately 10) 2 ~10 3 s) It is difficult to distinguish on the time-domain voltage curve, but after processing by this algorithm, g(τ) In the spectrum, the two appear as separate peaks, thus providing a high-resolution spectroscopic basis for feature extraction.

[0063] In a further embodiment of this application, the specific definition and extraction method of the characteristic parameters related to lithium plating in step S130 are refined.

[0064] Based on the electrochemical characteristics of the lithium plating reaction, the relaxation time distribution spectrum... g(τ) The following three types of feature parameters are defined: Parameter 1: Detection of newly added feature peaks within the first time constant interval.

[0065] The first time constant interval is the characteristic time constant interval corresponding to the lithium metal dissolution process, determined through a three-electrode calibration experiment. In this embodiment, the first time constant interval is defined as […]. τ1,τ2 For example, the first time constant interval can be [0.1s, 10s].

[0066] Extraction method: Real-time monitoring of the current situation g(τ) The spectrum is in [ τ1,τ2 Peak values ​​within the range. Set the noise threshold. Noise Th If a peak height exceeding [a certain value] is detected within this interval... Noise Th If an independent peak cannot be matched with a known peak in the health baseline spectrum, it is marked as "a new characteristic peak exists".

[0067] This embodiment uses a peak detection algorithm to analyze the first time constant interval [ τ1,τ2 The system performs a scan to identify local maxima (i.e., the peak height mentioned above). For each detected local maximum, its peak height (the distribution function value corresponding to that point), peak area (the integral of the peak on the time constant axis), and peak position (the time constant corresponding to that peak) are calculated. The peak area is defined as: in: S new —First time constant interval[ t 1 ,t 2 The peak area of ​​the newly added characteristic peak within the [ ]; —Relaxation time distribution spectrum; —The lower limit of the first time constant interval; —The upper limit of the first time constant interval.

[0068] τ1,τ2These represent the lower and upper limits of the first time constant interval, which are essentially the start and end time constants of the peak. The positions corresponding to the local minimum points on both sides of the peak are taken. The peak area is positively correlated with the amount of lithium metal participating in the dissolution reaction, and is a core characteristic parameter for subsequent quantitative estimation of lithium deposition.

[0069] Parameter 2: Parameter changes of the original characteristic peak.

[0070] In addition to the addition of new characteristic peaks, lithium plating also modulates the existing relaxation modes. The changes in interface state and accumulation of by-reaction products caused by lithium plating alter the relaxation characteristics of the charge transfer and diffusion processes. This embodiment extracts relaxation time distribution spectra located in... t3 to t4 The characteristic peaks of the interval (corresponding to the charge transfer process) and t4 The characteristic peaks in the above intervals (corresponding to solid-phase diffusion processes), and τ1<τ2≤τ3<τ4 (For example, it can be set) t1 It is 0.1s. t2 It lasts for 10 seconds. t3 It lasts for 10 seconds. t4 (The timeframe is 100 seconds, but practical applications are not limited to the values ​​mentioned above). Calculate the peak height, peak position, and half-peak width (WHM). Calculate the peak height change rate by comparing it with a reference spectrum (taken from the battery's health status). in: Peak height change rate; —Located in the relaxation time distribution spectrum t3 to t4 Characteristic peaks of the interval; —in the relaxation time distribution spectrum t4 The characteristic peaks in the above intervals.

[0071] Calculate peak position offset: in: Peak position offset; —Located in the relaxation time distribution spectrum t3 to t4 Peak position of the interval; —in the relaxation time distribution spectrum t4 The peak position in the above range.

[0072] Calculate the half-peak width growth rate: in: Half-peak width growth rate; —Located in the relaxation time distribution spectrum t3 to t4 Half-peak width of the interval; —in the relaxation time distribution spectrum t4 The half-peak width of the above interval.

[0073] The above parameters characterize the combined effects of lithium plating and aging on electrode interface kinetics and diffusion kinetics, providing a basis for distinguishing between lithium plating and aging in the future.

[0074] Extraction method: Extract the original characteristic peaks located in the second time constant interval (e.g., 10s, 100s, corresponding to contact impedance and SEI film impedance) and the third time constant interval (e.g., >100s, corresponding to solid-phase diffusion). Calculate the peak height change rate. Peak position offset Dt and half-peak width growth rate ΔW .

[0075] Parameter 3: Global similarity.

[0076] To capture the overall morphological changes in the relaxation time distribution spectrum, this invention introduces a dynamic time warping algorithm to calculate the global similarity between the current spectrum and the reference spectrum. The dynamic time warping algorithm calculates the minimum cumulative distance between two curves by performing non-linear alignment on the time axis. Its advantage lies in its ability to tolerate local offsets on the time constant axis. The similarity index DDTW is defined as: in, g current and g ref These are the current relaxation time distribution spectrum and the first reference spectrum, respectively. t k and t k ′ This represents the time constant point after dynamic alignment. This index comprehensively reflects the differences in peak height, peak position, peak width, and overall distribution morphology of the spectrum, and can serve as auxiliary information for lithium plating criteria, especially when newly added characteristic peaks are not obvious or overlap with existing peaks, providing supplementary judgment. Extraction method: Calculate the current measurement g current Health reference spectrum under the same temperature and SOC conditions g ref The Pearson correlation coefficient or dynamic time-normalized distance between them is used as a global similarity index.

[0077] The above three types of feature parameters constitute a multidimensional feature vector: It is used for subsequent lithium plating determination, aging differentiation and lithium plating amount estimation.

[0078] This embodiment constructs a three-dimensional feature space. The addition of a new peak is a sufficient but not necessary condition for lithium plating, while the broadening of the existing peak is a necessary condition for aging. In the subsequent judgment logic, by combining parameter one and parameter two, spectral distortion caused by lithium can be accurately distinguished from spectral distortion caused by low temperature and aging. For example, under low-temperature conditions, the original diffusion peak (parameter two) may shift to the right and increase in height, but since lithium plating has not occurred, parameter one will not trigger an alarm. This multi-dimensional feature separation mechanism is the key to the extremely high detection specificity (low false alarm rate) of this method.

[0079] In other embodiments of this application, this embodiment describes in detail the specific implementation method of temperature compensation processing based on temperature data in step S140.

[0080] Because the rate constant of the internal chemical reactions of a battery follows the Arrhenius equation, temperature changes cause a systematic shift of the characteristic peaks in the DRT spectrum along the time constant axis. Without compensation, this will lead to inaccurate similarity calculations.

[0081] Step S141: Establish the first reference spectral library (temperature library).

[0082] Healthy batteries with the same material system (same positive electrode, negative electrode, and electrolyte formulation) as the target battery were pre-selected. Different temperature points were set in a constant temperature chamber. T i (e.g., 10°C, 0°C, 10°C, 25°C, 45°C). At each temperature point, the battery is charged to the same target state of charge (e.g., 80% SOC) at a standard rate (e.g., 0.5C). The voltage relaxation curve is collected after resting and then inverted according to the algorithm described in the above embodiment. g(τ,T i ) All g(τ,T i ) The data is stored in the BMS storage device or cloud database to form the first reference spectrum library.

[0083] Step S142: Dynamic retrieval and interpolation of the reference spectrum.

[0084] During actual testing, the average temperature during the current resting period is read. T now .

[0085] Direct call: If T now Equal to a certain spectral library T i (Allowable error ±1°C), directly call the corresponding g(τ,Ti ) As the first reference spectrum g ref T .

[0086] Interpolation generation: If T now Between T i and T i+1 Between these, a first reference spectrum matching the current temperature is generated using bilinear interpolation or time constant translation interpolation based on the Arrhenius relation. g ref T .

[0087] in, g ref T The first reference spectrum to match the current temperature; For the corresponding temperature T i First reference spectrum; For the corresponding temperature T i+1 First reference spectrum; T i and For adjacent test temperature points, and T i <T< .

[0088] Step S143: Apply compensation.

[0089] The generated first reference spectrum g ref T This serves as the calculation benchmark for the changes in the original characteristic peak parameters in the above embodiments, as well as the comparison benchmark for global similarity.

[0090] This embodiment is equivalent to configuring a temperature compensation module for the detection algorithm. It ensures that regardless of whether it's a frigid winter (e.g., charging at 10°C) or a sweltering summer (e.g., charging at 45°C), the normal standard image used for comparing lithium plating characteristics is the standard image at that specific temperature, rather than a fixed 25°C standard image. This fundamentally eliminates the engineering false alarm problem of misjudging a newly added lithium plating peak due to a rightward shift of the diffusion peak caused by low temperature.

[0091] In a preferred embodiment of this application, this embodiment elaborates in detail the specific implementation method of the normalization processing based on the state of charge data in step S150.

[0092] Step S151: Prioritize the determination of the high charge state window.

[0093] Determining the start time of resting SOC start Is the state of charge (SOC) greater than a preset threshold (e.g., 80% SOC)? Lithium plating mainly occurs at the end of charging, and the high SOC range is a high-incidence area for lithium plating. Detection within this range is targeted. Under high SOC, the lithium concentration in the negative electrode is close to saturation, and small changes in SOC have a relatively small impact on relaxation characteristics, resulting in good stability of characteristic parameters. The SOC of the battery management system at the end of charging is usually within this range, facilitating data collection. In practical applications, if the SOC at the end of charging is below 80%, supplementary charging or waiting for the next charging cycle can be used for detection to unify the detection conditions within the high SOC window.

[0094] Case A (SOC≥80%): The characteristic signal is the strongest, and subsequent judgment steps can be performed directly without complex SOC normalization correction.

[0095] Case B (SOC < 80%): Perform the following normalization process.

[0096] Step S152: Establish the second reference spectrum library (SOC gradient library).

[0097] Beforehand, healthy cells from the same system were charged to different discrete SOC points (e.g., 80%, 85%, 90%, 95%, 100%) at a constant temperature (e.g., 25°C). The DRT spectra at different SOCs were obtained by resting and inversion. g(τ,SOC cur ) .

[0098] Step S153: Normalization calculation.

[0099] When the actual detected state of charge differs from the standard detection window (e.g., 100%), a normalization method based on a mapping table is used to adjust the feature parameters. The specific steps are as follows: First, based on the current state of charge... SOC cur Retrieve the reference spectrum of the corresponding charge state from the mapping table. g ref (t,SOC cur ) ; Calculate the difference eigenvector between the current measured spectrum and the reference spectrum F raw Next, based on the characteristic change gradients between adjacent charge state points in the mapping table, the feature vectors are normalized to the standard charge state. SOC std The estimated value is given below. The normalization formula is: in, These are normalized feature vectors; F raw This represents the eigenvector representing the difference between the current measured spectrum and the reference spectrum. The sensitivity of the eigenvector to the state of charge; Standard state of charge; This represents the current state of charge.

[0100] Understandably, when the state of charge is too low (e.g., below 50%), the probability of lithium plating and the significance of the characteristic peaks are greatly reduced. At this time, the error of normalization processing may increase. However, this method can still provide continuous and reliable monitoring in the medium-to-high state of charge range (e.g., above 60% SOC) where lithium plating is more likely to occur.

[0101] This embodiment addresses the detection blind spot problem caused by charging not necessarily reaching full capacity in practical applications. For example, a user might charge from 30% to 70% at a highway service area and then stop. If trace amounts of lithium plating occur at this time, the plating peak might be masked by a stronger charge transfer peak due to the low State of Charge (SOC). Through the sensitivity normalization processing in this embodiment, this weak lithium plating characteristic signal can be equivalently amplified to the amplitude of a fully charged state for evaluation. This allows for continuous monitoring of lithium plating risk across the entire SOC range without changing the user's charging habits.

[0102] In some embodiments of this application, the specific logic threshold for determining whether lithium plating has occurred in step S160 is explicitly given.

[0103] This embodiment employs cascaded decision logic, as detailed below: Level 1: Lithium plating specific detection (high-risk assessment).

[0104] Logical condition: If, within the first time constant interval, the peak signal-to-noise ratio of the spectrum after temperature compensation and SOC normalization is greater than 3... s (i.e., more than 3 standard deviations above the baseline noise fluctuation), and the residual area between the peak and the first reference spectrum in this interval is greater than a preset threshold.

[0105] Output: Lithium plating has occurred.

[0106] Level 2: Normal aging assessment (low risk assessment).

[0107] Logical condition: If no new characteristic peak is detected, but changes in the parameters of existing characteristic peaks are detected (such as the rate of change of full width at half maximum), then... ΔW >20% or peak position offset Dt >15%).

[0108] Output: The battery is determined to be aging normally, and no lithium plating has occurred.

[0109] The peak height, peak position, and half-width at half-maximum (WHM) parameters of the original characteristic peaks in the relaxation time distribution spectrum of the above embodiments are extracted. The current parameters are compared with the reference spectrum (taken from the battery health state) to calculate the changes in peak parameters (including the rate of change of peak height, the shift of peak position, and the rate of increase of WHM). If the above parameters change significantly but no new characteristic peaks are detected, it is determined that normal aging is dominant.

[0110] Level 3: Normal state.

[0111] Logical condition: Neither of the above conditions is met.

[0112] Output: Battery status is normal.

[0113] Alternatively, a combined assessment can be made by constructing a three-dimensional feature vector from the area of ​​newly added feature peaks and the changes in the peak parameters of existing feature peaks. The distribution range of the feature vector under different degrees of lithium plating and aging is established experimentally. During actual testing, the range in which the feature vector of the battery under test falls is used as the judgment criterion: if the feature vector falls into the lithium plating-dominated range, it is determined that lithium plating has occurred; if it falls into the aging-dominated range, it is determined that normal aging has occurred; if it is in the boundary range between the two, it is determined that lithium plating and aging coexist, and the contribution degree of each is output.

[0114] This embodiment clarifies the divergence in the evolutionary paths of lithium plating and aging in the DRT spectrum. Lithium plating is a process of creating a new phase from scratch (spectrally corresponding to the appearance of a new peak); aging is a process of rapid to slow kinetic decay (spectrally corresponding to the broadening and shortening of existing peaks). Through cascaded logic, the system can accurately identify aged batteries that, although exhibiting increased internal resistance and deteriorating voltage curves, have not yet undergone lithium plating. This prevents the BMS from incorrectly limiting charging power due to misjudging lithium plating, thus ensuring the user's driving experience and the battery's usable capacity.

[0115] In a further embodiment of this application, based on the determination of lithium plating in the above embodiment, this embodiment further adds the function of quantitative estimation of lithium plating capacity.

[0116] Quantitative calculations and algorithm supplements involved in the detection method: Step S170: Calculate the peak area of ​​the newly added characteristic peak.

[0117] After confirming the presence of a new lithium plating peak, the relaxation time distribution spectrum was analyzed. g(τ) Above, for the first time constant interval [ τ1,τ 2 Perform numerical integration to calculate the peak area of ​​the newly added characteristic peak. S : in: S —Peak area of ​​newly added characteristic peaks; —Relaxation time distribution spectrum; —The lower limit of the first time constant interval; —The upper limit of the first time constant interval.

[0118] Algorithm implementation details: Discrete calculations are performed using the composite Simpson integral rule. If the spectral baseline is tilted, the baseline is first subtracted using the asymmetric least squares method, and integration is performed only on the net peak area to improve the accuracy of the area calculation.

[0119] Step S180: Mapping lithium plating capacity.

[0120] Based on a pre-calibrated quantitative mapping model of peak area and lithium deposition capacity: in: Q Li Lithium plating capacity; S The peak area of ​​the newly added characteristic peak; k This is the sensitivity coefficient. b This is the intercept.

[0121] The calculated peak area S Substitute the values ​​into the model to output an estimated value for the lithium plating capacity. (Unit: mAh or percentage of total capacity).

[0122] This embodiment achieves a leap from qualitative judgment to quantitative evaluation. The peak area of ​​the lithium plating peak is directly proportional to the reactive area of ​​the lithium metal and electrolyte interface, and the reactive area is strongly correlated with the total amount of deposited lithium metal. This is achieved by outputting a specific lithium plating capacity value (e.g., an estimated lithium plating capacity of 0.15 mAh / cm³). 2 BMS can execute fine-grained degradation strategies. For example, when Q Li When <0.1, only events are recorded; when Q Li When the current exceeds 0.5 ohms, the charging current is actively limited and the user is prompted to bring the battery in for testing. This quantitative and tiered management is the technological foundation for achieving intelligent safety monitoring throughout the battery's entire lifecycle.

[0123] In other embodiments of this application, this embodiment elaborates in detail the specific calibration method of the peak area and lithium plating capacity quantitative mapping model described in the above embodiments.

[0124] Step S191: Prepare a three-electrode cell of the same system.

[0125] A small pouch cell with the exact same positive and negative electrode material formulation, electrolyte and separator as the target battery was fabricated, and a micro lithium metal reference electrode (copper wire plated with lithium) was implanted inside the cell to ensure that the position of the reference electrode does not affect the electric field distribution.

[0126] Step S192: Induce lithium plating at different gradients.

[0127] In a constant temperature chamber, lithium plating behavior of different degrees is induced in the three-electrode cell by changing the charging rate (e.g., 1C, 2C, 3C), charging temperature (e.g., 0°C, 5°C, 10°C) and cutoff voltage.

[0128] Step S193: Calculate the actual lithium plating capacity.

[0129] The potential of the negative electrode relative to the reference electrode is monitored in real time using a data acquisition instrument. E neg .when E neg A value <0V (vs. Li+ / Li) marks the start of lithium deposition. Record the time from the start of lithium deposition. t start Until the end of charging t end Total charge flowing into the battery Q charge Since this stage involves both lithium plating and graphite lithium intercalation, a distinction needs to be made. Actual lithium plating capacity. Q Li The capacity can be obtained through discharge process stripping analysis: that is, during the subsequent small current discharge process, the discharge capacity corresponding to the lithium plating dissolution plateau (approximately 0V vs. Li+ / Li) is monitored. ΔQ This represents the actual lithium plating capacity during the charging process.

[0130] Step S194: DRT peak area extraction and curve fitting.

[0131] The three-electrode cells that have undergone lithium plating were subjected to a static relaxation test, and the peak area of ​​the newly added characteristic peaks was extracted according to the method described in the above embodiment. S i .

[0132] The newly added feature peak area S As the independent variable, the actual lithium plating capacity is used. Q Li Using [variable name] as the dependent variable, a scatter plot was drawn. Experiments verified that the two typically exhibit a linear or quadratic polynomial relationship. The calibration curve equation was obtained by fitting using the least squares method: .

[0133] This embodiment constructs a precise calibration system based on first-principles electrochemistry. The three-electrode system can eliminate interference from the positive electrode and directly read the thermodynamic potential of the negative electrode, thus providing a standard for determining the onset of lithium plating. The mapping model established by this method is far more accurate than visually observing grayish-white deposits after disassembling the battery. Once established, this calibration model can be written into the BMS of all commercially available batteries of the same system, realizing the transformation from precise laboratory measurements to embedded industrial applications.

[0134] A battery lithium plating detection method according to this application is specifically exemplified as follows: Testing scenario: The electric vehicle is fast charged to 85% SOC at 1.5C in an ambient temperature of 15°C and then stops charging, entering a parking sleep state.

[0135] The comprehensive testing process is as follows: Initial condition recording: BMS wakes up the low-voltage sampling circuit and records. SOC start =85%, T =15°C. The static voltage curve was acquired at a frequency of 1Hz for 10 minutes. V raw (t) .

[0136] Data preprocessing: Smoothing curves are obtained by moving average filtering. V(t) .

[0137] Reference Spectrum Retrieval: A query of the first reference spectrum library revealed that no spectrum for 15°C was stored. The stored 10°C and 25°C reference spectra were extracted, and a temporary first reference spectrum adapted to 15°C was generated using the Arrhenius interpolation algorithm. g ref T .

[0138] DRT inversion calculation: Run embedded C code (or cloud-based Python algorithm) and solve using the Tikhonov regularized NNLS algorithm. g(τ) Measured spectrum.

[0139] SOC window judgment and normalization: Judgment SOC start =85%, located within the high SOC detection window, skipping complex sensitivity normalization correction and directly entering feature extraction.

[0140] Multidimensional feature extraction: Scanning the first time constant interval [0.1s, 10s]: a new peak significantly higher than the noise threshold was found.

[0141] contrast g refT This peak has been confirmed as a newly added characteristic peak.

[0142] Monitoring the original diffusion peak: the rate of change of the full width at half maximum (FWHM) only increased by 5%.

[0143] Lithium plating determination and quantitative assessment: Decision: "Level 1 logic" is triggered, indicating that lithium plating has occurred.

[0144] Calculation: Baseline subtraction and Simpson integration are performed on the newly added characteristic peaks to obtain the peak area. S =0.45.

[0145] Lithium plating capacity output: Calls a pre-stored calibration equation Q Li =0.82* S +0.05, the estimated lithium plating capacity is calculated to be 0.42 mAh / cm³. 2 .

[0146] Actuator response: BMS based on 0.42mAh / cm 2 This moderate lithium plating level triggers a secondary protection strategy: the yellow "Battery Maintenance Indicator" light is illuminated on the dashboard, the maximum allowable charging current for the next charge is limited to 0.8C, and the event is uploaded to the cloud monitoring platform.

[0147] In addition, this application also relates to a battery lithium plating detection system, comprising: The data acquisition module is used to collect terminal voltage and temperature data during the resting period after the target battery has finished charging. The data preprocessing module is used to process the terminal voltage data to obtain the voltage relaxation curve; The inversion calculation module is used to invert the relaxation time distribution of the voltage relaxation curve to obtain the relaxation time distribution spectrum. The feature extraction module is used to extract feature parameters related to lithium plating based on the relaxation time distribution spectrum; The compensation and normalization module is used to perform temperature compensation and state of charge normalization processing on the feature parameters based on the temperature data and the state of charge data at the start of the resting period. The lithium plating determination module is used to determine whether lithium plating has occurred in the battery based on the processed characteristic parameters.

[0148] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A method for detecting lithium plating in batteries, characterized in that: include: After the charging process of the target battery is completed, the terminal voltage data of the battery during the resting period is collected and filtered and denoised to obtain the voltage relaxation curve. The relaxation time distribution spectrum is obtained by inverting the voltage relaxation curve; Obtain the first reference spectrum of a healthy battery of the same system as the target battery under reference conditions; Based on the relaxation time distribution spectrum and the first reference spectrum, feature parameters are extracted that include at least whether the relaxation time distribution spectrum has a new feature peak relative to the first reference spectrum within the first time constant interval. Acquire the temperature data of the battery during the resting period, and perform temperature compensation processing on the feature parameters based on the temperature data; Acquire the state of charge (SOC) data of the battery at the initial moment of rest, and perform SOC normalization on the feature parameters based on the SOC data. Based on the characteristic parameters after temperature compensation and state of charge normalization, it is determined whether lithium plating has occurred in the battery.

2. The battery lithium plating detection method according to claim 1, characterized in that: The obtained relaxation time distribution spectrum g( ) The methods include: in: —The voltage relaxation curve obtained from the data acquisition; —Relaxation time distribution spectrum; —Relaxation time constant; —Open circuit voltage; The integral model is solved using a regularized inversion algorithm to obtain the relaxation time distribution spectrum. The regularization inversion algorithm includes Tikhonov regularization or Fourier transform filtering and deconvolution.

3. The battery lithium plating detection method according to claim 1, characterized in that: The feature parameters also include: The parameter changes of the original characteristic peaks in the relaxation time distribution spectrum relative to the first reference spectrum; and, The global similarity between the relaxation time distribution spectrum and the first reference spectrum.

4. The battery lithium plating detection method according to claim 3, characterized in that: The method for temperature compensation processing of feature parameters based on temperature data includes: A reference relaxation experiment of a healthy battery with the same system as the target battery was performed in advance. After charging to the same state of charge at the standard charging rate under different temperature conditions, the voltage relaxation curves after the charging was completed were collected. The relaxation time distribution of the collected voltage relaxation curves was inverted to obtain the first reference spectrum library under different temperature conditions. Based on the temperature data, a first reference spectrum matching the current temperature is generated by calling or interpolating from the first reference spectrum library; The first reference spectrum is used as the benchmark for comparing the parameter changes in the feature parameters with the global similarity.

5. The battery lithium plating detection method according to claim 1, characterized in that: The method for normalizing the characteristic parameters based on the state of charge data includes: Determine whether the state of charge data at the initial resting time is within a preset high state of charge range. If yes, proceed directly to the next steps; otherwise: A reference relaxation experiment of a healthy battery with the same system as the target battery was performed in advance. After charging to different states of charge under the same temperature conditions, the voltage relaxation curves of each state of charge were collected. The relaxation time distribution of the collected voltage relaxation curves was inverted to obtain a second reference spectrum library under different states of charge. Based on the second reference spectral library, the sensitivity of the characteristic parameters to changes in state of charge is calculated; Using the aforementioned sensitivity, the feature parameters extracted under the current state of charge are normalized to their corresponding values ​​under a standard state of charge.

6. The battery lithium plating detection method according to claim 3, characterized in that: The method for determining whether lithium plating has occurred in the battery includes: if a new characteristic peak higher than the noise threshold is detected within the first time constant interval, it is determined that lithium plating has occurred; If no new characteristic peak is detected, but the parameter change of the original characteristic peak exceeds a preset threshold, it is determined that the battery is aging normally.

7. The battery lithium plating detection method according to claim 6, characterized in that: When lithium plating is determined to have occurred, the peak area of ​​the newly added characteristic peak is calculated; Based on the pre-calibrated quantitative mapping model between the peak area of ​​the newly added characteristic peak and the lithium plating capacity, the estimated value of the lithium plating capacity is output according to the calculated peak area.

8. The battery lithium plating detection method according to claim 7, characterized in that: The method for calculating the peak area of ​​the newly added characteristic peak includes: calculating the peak area of ​​the newly added characteristic peak according to the following formula. in: S —Peak area of ​​newly added characteristic peaks; —Relaxation time distribution spectrum; —The lower limit of the first time constant interval; —The upper limit of the first time constant interval.

9. The battery lithium plating detection method according to claim 8, characterized in that: The method for calibrating a quantitative mapping model between the peak area of ​​the newly added characteristic peak and the lithium plating capacity includes: using a three-electrode cell of the same system as the target battery, charging under different operating conditions and inducing lithium plating to different degrees; Monitor the negative electrode potential of the battery cell to determine the occurrence of lithium plating and calculate the actual lithium plating capacity; Extract the peak area of ​​the newly added characteristic peaks; Establish a mapping relationship with the newly added characteristic peak area as the independent variable and the actual lithium plating capacity as the dependent variable.

10. A battery lithium plating detection system, characterized in that: The system is used to operate according to a battery lithium plating detection method as described in any one of claims 1 to 9, including: The data acquisition module is used to collect terminal voltage and temperature data during the resting period after the target battery has finished charging. The data preprocessing module is used to process the terminal voltage data to obtain the voltage relaxation curve; The inversion calculation module is used to invert the relaxation time distribution of the voltage relaxation curve to obtain the relaxation time distribution spectrum. The feature extraction module is used to extract, based on the relaxation time distribution spectrum, at least the feature parameters including whether there is a new feature peak in the relaxation time distribution spectrum relative to the first reference spectrum within a first time constant interval. The compensation and normalization module is used to perform temperature compensation and state of charge normalization processing on the feature parameters based on the temperature data and the state of charge data at the start of the resting period. The lithium plating determination module is used to determine whether lithium plating has occurred in the battery based on the processed characteristic parameters.

Images