Method and system for predicting service life of IDC battery monomer

Abstract

The invention relates to the technical field of life prediction, in particular to an IDC battery monomer life prediction method and system, and the method comprises the following steps: recognizing a load change event in a voltage and current sequence, screening high-frequency disturbance, extracting disturbance characteristics, generating a score, accumulating and calculating a disturbance proportion, analyzing the correlation between the disturbance proportion and capacity attenuation, and defining an anti-interference critical value. And correcting the life attenuation curve according to the critical value, and outputting a life prediction result. According to the invention, through identifying a high-frequency load disturbance event and extracting multiple numerical features corresponding to voltage, current and internal resistance changes, a quantitative scoring system of disturbance intensity is established, boundary conditions of transformation of a capacity state from stability to recession are defined by means of an inflection point identification technology, and a dynamic adjustment mechanism of a capacity stability adjustment factor is combined. And the attenuation gradient of the basic life curve is corrected, so that the prediction process can reflect the dynamic degradation behavior of the battery under the complex working condition, and the real-time performance of life prediction and the accuracy of capacity state estimation are improved.

Description

Technical Field

[0001] This invention relates to the field of lifespan prediction technology, and in particular to a method and system for predicting the lifespan of individual IDC battery cells. Background Technology

[0002] The field of lifespan prediction technology involves lifespan estimation schemes for lithium-ion battery cells in high-reliability power systems such as information data centers. Specifically, it includes methods for modeling and analyzing the remaining lifespan of battery cells based on factors such as battery operation data, aging patterns, and electrochemical performance parameters. This technical field covers core aspects such as battery lifespan modeling, extraction of charge and discharge characteristic parameters, identification of lifespan decay patterns, fusion processing of historical operation data, and lifespan estimation based on machine learning or mathematical statistics methods. It is widely used to improve the stability and predictive maintenance capabilities of power supply systems for IDC equipment.

[0003] Traditional IDC battery cell lifetime prediction methods refer to lifetime assessment approaches used for battery cells applied in data center scenarios to evaluate their future availability and planned replacement timing. These methods typically involve collecting raw operating parameters such as voltage, current, temperature, and internal resistance in the actual operating environment, judging the degree of battery degradation based on set thresholds, or performing interpolation predictions based on battery capacity decay curves under sample operating conditions. Some methods also estimate lifetime by establishing a capacity-cycle count function model under specific environments and operating conditions, using cycle counts as a variable. Furthermore, periodic offline testing under static operating conditions to obtain parameters such as SOC and SOH are often used as a basis for lifetime judgment. Most of these methods rely on empirical regression, threshold triggering mechanisms, or static experimental data, lacking real-time feature mining and comprehensive analysis of the battery's dynamic operating process. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an IDC battery cell lifetime prediction method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for predicting the lifespan of a single IDC battery cell, comprising the following steps:

[0006] S1: Identify load change events in the operating voltage and current sequences of individual IDC battery cells, calculate the load change frequency, filter events whose load change frequency exceeds the frequency limit, and construct a set of high-frequency disturbance events.

[0007] S2: Extract the voltage drop, current amplitude change rate, and short-time increase in internal resistance from the set of high-frequency disturbance events, and calculate and generate a disturbance intensity score;

[0008] S3: Accumulate the disturbance intensity scores within the detection period to obtain the total intensity of high-frequency disturbances, calculate the percentage of the total intensity of high-frequency disturbances to the total intensity of load change events, and generate the high-frequency disturbance intensity percentage.

[0009] S4: Obtain the discharge capacity sequence of IDC battery cells, calculate the capacity decay rate, analyze the Pearson correlation coefficient between the high-frequency disturbance intensity ratio and the capacity decay rate, extract the inflection point where the capacity decay rate turns into an exponential change, and define the anti-disturbance critical value.

[0010] S5: Calculate the capacity stability adjustment factor based on the position of the disturbance rejection critical value in the population disturbance rejection distribution curve, input the capacity stability adjustment factor into the long short-term memory network model to correct the descent gradient of the battery life decay curve provided by the manufacturer, and output the IDC battery cell life prediction result.

[0011] The present invention improves upon this invention by including the following: the high-frequency disturbance event set includes a disturbance occurrence timestamp, instantaneous frequency value of load change, and event index identifier; the disturbance intensity score includes a linearly weighted aggregated value, multi-parameter quantization weight coefficients, and intensity level classification labels; the high-frequency disturbance intensity proportion includes the cumulative total value of high-frequency disturbance intensity, the total intensity value of load change over the entire cycle, and the intensity proportion percentage value; the anti-disturbance critical value includes the proportion value corresponding to the capacity decay inflection point, the initiation slope of exponential change, and the Pearson correlation confidence interval; and the IDC battery cell lifetime prediction result includes the corrected remaining lifetime value, the dynamic decay trajectory of battery capacity, and the prediction result confidence probability.

[0012] The present invention is improved in that the step of obtaining the high-frequency disturbance event set is specifically as follows:

[0013] S111: Identify load change events in the operating voltage and current sequences of IDC battery cells, perform discrete difference operations on the operating voltage and current sequences along the time axis to obtain voltage gradient vectors and current gradient vectors, anchor the sampling time when the magnitude of the voltage gradient vector and current gradient vector exceeds the preset steady-state noise reference value as the disturbance trigger point, retrieve the signal recovery time after the disturbance trigger point as the disturbance termination point, extract the corresponding voltage data segment and current data segment from the original sequence according to the time coordinates of the disturbance trigger point and the disturbance termination point, perform noise removal, cleaning and serialization on the extracted data segments, and generate a load transient response event sequence;

[0014] S112: Call the load transient response event sequence, extract the voltage waveform segment and current waveform segment included in each independent response event in the sequence one by one, perform discrete Fourier transform processing on the voltage waveform segment and current waveform segment, construct a frequency domain energy distribution spectrum, search for the frequency axis coordinate value corresponding to the energy peak in the frequency domain energy distribution spectrum, calculate the relative deviation between the frequency axis coordinate value and the fundamental frequency value and quantize it into the load change frequency parameter of the event, and form a one-dimensional array according to the time order of the event occurrence to generate the event change frequency feature vector;

[0015] S113: Based on the event change frequency feature vector, a preset frequency limit threshold is used as the screening criterion to perform a traversal numerical comparison of the load change frequency parameters included in the vector, identify and lock the target event index number whose load change frequency parameter value is greater than the frequency limit threshold, and retrieve the corresponding voltage and current waveform segments from the load transient response event sequence according to the target event index number, collect and reorganize the retrieved waveform segments and add high-frequency attribute identifier bits to construct a high-frequency disturbance event set.

[0016] The present invention is improved in that the process of setting the steady-state noise reference value is specifically as follows: when the IDC battery is in a stable operating state of constant floating charge or no-load static, the original voltage and current signals within a preset time period under the stable state are collected, a first-order discrete difference operation is performed on the signal data to extract the gradient fluctuation sequence of background noise, statistical analysis is performed on the gradient fluctuation sequence to calculate the mean and standard deviation, and the superposition value of the mean and three times the standard deviation is used as the steady-state noise reference value.

[0017] The process of setting the frequency limit threshold is as follows: by performing electrochemical impedance spectroscopy tests on batteries of the same model, the boundary characteristics between the high-frequency region and the mid-frequency region are traced in the generated Nyquist plot, the topological connection inflection point between the charge transfer impedance semicircular region and the low-frequency Warburg diffusion impedance straight line is located, the characteristic frequency value corresponding to the inflection point is extracted, and it is set as the frequency limit threshold.

[0018] The present invention is improved in that the step of obtaining the disturbance intensity score is specifically as follows:

[0019] S211: Based on the set of high-frequency disturbance events, traverse the voltage response curve and current response curve of each load change event in the set, calculate the absolute difference between steady-state voltage and valley voltage, obtain the voltage drop value, perform time dimension differentiation operation on the current response curve, obtain the current amplitude change rate, perform integration operation on the product of voltage and current within the event window, obtain the energy throughput value, and generate transient electric stress feature vector;

[0020] S212: Call the transient electrical stress feature vector, locate the original data segment in the high-frequency disturbance event set according to the time index corresponding to each element in the vector, identify the ohmic internal resistance step at the event trigger time, obtain the short-term increase value of the internal resistance, simultaneously extract the duration of the event as the event duration, and collect the remaining battery power at the start of the event as the charge state value, and establish the impedance state parameter matrix.

[0021] S213: Call the transient electric stress characteristic vector and impedance state parameter matrix, and convert the voltage drop value, current amplitude change rate, energy throughput value, short-term internal resistance increase value, event duration and charge state value into dimensionless normalized form, and calculate to obtain the disturbance intensity score.

[0022] The present invention is improved in that the formula for obtaining the disturbance intensity score is specifically as follows:

[0023] ;

[0024] in, This represents the disturbance intensity score. This represents the normalized voltage drop value. Represents the rate of change of the normalized current amplitude. Represents the energy weighting coefficient. This represents the short-term increase value of the normalized internal resistance. Represents the normalized energy throughput value. Represents the duration of the influence factor. Duration of the normalization event Represents the numerical value of the charge state.

[0025] The present invention is improved in that the step of obtaining the proportion of high-frequency disturbance intensity is specifically as follows:

[0026] S311: Based on the disturbance intensity score, set the start and end timestamps of the detection period to construct an effective statistical window, retrieve the disturbance intensity score values ​​that fall within the statistical window and perform numerical accumulation on them, aggregate the discretely distributed single disturbance impact intensity into scalar data that characterizes the cumulative damage effect in the high-frequency interval within the period, and generate a high-frequency cumulative stress index.

[0027] S312: Call the load transient response event sequence, traverse all load change events recorded in the sequence and extract the voltage drop and current change rate parameters one by one, calculate the stress intensity of each event, and perform a summation operation on all intensity values ​​within the detection period to quantify the overall load fluctuation energy including low frequency and high frequency, and generate the total load intensity value for the entire cycle.

[0028] S313: Call the high-frequency cumulative stress index and the total load strength value of the whole cycle to construct the ratio relationship between the two. Divide the high-frequency cumulative stress index by the total load strength value of the whole cycle to obtain the dimensionless weighting coefficient. Convert the weighting coefficient into a percentage value and perform precision truncation processing on it to retain the significant numbers and generate the high-frequency disturbance intensity ratio.

[0029] The present invention is improved in that the step of obtaining the anti-interference critical value is specifically as follows:

[0030] S411: Obtain the discharge capacity sequence of the IDC battery cell during its historical operation, perform a first-order differential operation on the discharge capacity sequence with respect to the time dimension to quantify the capacity decay rate, call the high-frequency disturbance intensity ratio, align the capacity decay rate at each detection moment with the high-frequency disturbance intensity ratio at the corresponding moment, calculate the covariance of the two sets of data sequences and divide it by the product of the standard deviations, obtain the Pearson correlation coefficient, filter the data groups whose correlation coefficients meet the preset conditions, and generate decay-disturbance coupled data pairs.

[0031] S412: Call the attenuation-perturbation coupled data pair, project the data pair onto the feature space with the proportion as the horizontal axis and the rate as the vertical axis, use the least squares method to perform piecewise fitting on the scatter distribution in the feature space, set the linear function and the exponential function, find the best fitting breakpoint, calculate the horizontal axis slicing position corresponding to the minimum sum of squared fitting residuals, identify the feature node where the capacity attenuation mode changes from linear accumulation to exponential, and generate the coordinates of the nonlinear abrupt change inflection point.

[0032] S413: Call the coordinates of the nonlinear mutation inflection point, analyze the specific value of the disturbance intensity ratio corresponding to the coordinates on the horizontal axis of the feature space, introduce the preset statistical confidence interval parameter to perform boundary convergence correction on the value, eliminate outlier sample points caused by measurement error or transient interference, and solidify the corrected value as the limit boundary characterizing the battery's ability to maintain linear capacity decay under high-frequency disturbance environment, and generate the anti-disturbance critical value.

[0033] The present invention is improved in that the specific steps for obtaining the IDC battery cell lifetime prediction results are as follows:

[0034] S511: Based on the aforementioned anti-interference critical value, project it onto a preset population anti-interference distribution probability density function coordinate system, use numerical integration to calculate the cumulative distribution probability value of the anti-interference critical value under the distribution curve, quantify the relative anti-interference position of the sample in the population, use the cumulative distribution probability value as an index key to retrieve the corresponding dimensionless weight scalar in a preset nonlinear adjustment coefficient mapping table, and generate a capacity stability adjustment factor.

[0035] S512: Input the capacity stability adjustment factor into the forget gate control logic of the long short-term memory network model, use the factor to perform element-wise weighted calculation on the historical state retention weight matrix of the hidden layer unit of the model, obtain the reset weight parameters, drive the model to dynamically correct the instantaneous decline slope of the battery life decay curve provided by the manufacturer at the current time step according to the reset weight parameters, and generate the corrected decay gradient vector.

[0036] S513: Call the corrected decay gradient vector, use the actual battery capacity at the current moment as the starting boundary condition to perform discrete-time step-by-step integral extrapolation, construct the capacity evolution trajectory covering the entire future life cycle, calculate the time axis coordinate point where the capacity evolution trajectory and the preset scrap capacity cutoff line intersect through the geometric intersection algorithm, and generate IDC battery cell life prediction results.

[0037] An IDC battery cell lifetime prediction system is provided, the system being used to implement the aforementioned IDC battery cell lifetime prediction method, the system comprising:

[0038] The disturbance event identification module identifies load change events in the operating voltage and current sequences of individual IDC battery cells, calculates the load change frequency, filters events whose load change frequency exceeds the frequency limit, and constructs a set of high-frequency disturbance events.

[0039] The disturbance intensity assessment module extracts the voltage drop, current amplitude change rate, and short-term internal resistance increase from the set of high-frequency disturbance events, and calculates and generates a disturbance intensity score.

[0040] The disturbance intensity analysis module accumulates the disturbance intensity scores within the detection period to obtain the total intensity of high-frequency disturbances, calculates the percentage of the total intensity of high-frequency disturbances to the total intensity of load change events, and generates the high-frequency disturbance intensity percentage.

[0041] The anti-interference critical identification module obtains the discharge capacity sequence of IDC battery cells, calculates the capacity decay rate, analyzes the Pearson correlation coefficient between the high-frequency disturbance intensity ratio and the capacity decay rate, extracts the inflection point where the capacity decay rate turns into an exponential change, and defines the anti-interference critical value.

[0042] The lifetime prediction and analysis module calculates the capacity stability adjustment factor based on the position of the disturbance rejection threshold in the population disturbance rejection distribution curve, inputs the capacity stability adjustment factor into the long short-term memory network model to correct the descent gradient of the battery lifetime degradation curve provided by the manufacturer, and outputs the IDC battery cell lifetime prediction result.

[0043] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0044] In this invention, by identifying high-frequency load disturbance events and extracting multiple numerical features of corresponding voltage, current, and internal resistance changes, a quantitative scoring system for disturbance intensity is established. The proportion of disturbance intensity over the entire cycle is calculated, forming a statistical correlation between disturbance impact and capacity decay rate. By using inflection point identification technology to define the boundary conditions for the transition of capacity state from stability to degradation, and combining the dynamic adjustment mechanism of capacity stability adjustment factor, the decay gradient of the basic lifetime curve is corrected, so that the prediction process can reflect the dynamic degradation behavior of the battery under complex operating conditions, thereby improving the real-time performance of lifetime prediction and the accuracy of capacity state estimation. Attached Figure Description

[0045] Figure 1 This is a flowchart of the method of the present invention;

[0046] Figure 2 This is a flowchart illustrating the process of obtaining a set of high-frequency disturbance events according to the present invention;

[0047] Figure 3 This is a flowchart for obtaining the disturbance intensity score in this invention;

[0048] Figure 4 This is a flowchart for obtaining the proportion of high-frequency disturbance intensity in this invention;

[0049] Figure 5 This is a flowchart for obtaining the anti-interference critical value according to the present invention;

[0050] Figure 6 This is a flowchart illustrating the process of obtaining the predicted lifespan of individual IDC battery cells according to the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0052] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention 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, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0053] Please see Figure 1 This invention provides a technical solution, a method for predicting the lifespan of a single IDC battery cell, comprising the following steps:

[0054] S1: Identify load change events in the operating voltage and current sequences of individual IDC battery cells, calculate the load change frequency of each load change event, filter load change events whose load change frequency exceeds a preset frequency limit, and construct a set of high-frequency disturbance events.

[0055] S2: Extract the voltage drop value, current amplitude change rate, and short-time increase value of internal resistance for each load change event in the high-frequency disturbance event set. Input the voltage drop value, current amplitude change rate, and short-time increase value of internal resistance into the linear weighted aggregation function for calculation to generate a disturbance intensity score.

[0056] S3: Perform an accumulation operation on the disturbance intensity scores within the preset detection period to obtain the total intensity of high-frequency disturbances, calculate the percentage of the total intensity of high-frequency disturbances to the total intensity of all load change events, and generate the high-frequency disturbance intensity percentage.

[0057] S4: Obtain the capacity decay rate of IDC battery cells, perform Pearson correlation coefficient analysis on the proportion of high-frequency disturbance intensity and capacity decay rate, extract the inflection point where the capacity decay rate changes from linear to exponential, and define the anti-disturbance critical value.

[0058] S5: Calculate the capacity stability adjustment factor based on the position of the anti-disturbance critical value in the preset group anti-disturbance distribution curve, input the capacity stability adjustment factor into the long short-term memory network model to dynamically correct the descent gradient of the battery life decay curve provided by the manufacturer, and output the IDC battery cell life prediction result.

[0059] The high-frequency disturbance event set includes the disturbance occurrence timestamp, instantaneous frequency value of load change, and event index identifier. The disturbance intensity score includes linearly weighted aggregated values, multi-parameter quantization weight coefficients, and intensity level classification labels. The high-frequency disturbance intensity proportion includes the cumulative total value of high-frequency disturbance intensity, the total intensity value of load change over the entire cycle, and the intensity proportion percentage value. The disturbance immunity critical value includes the proportion value corresponding to the capacity decay inflection point, the initiation slope of exponential change, and the Pearson correlation confidence interval. The IDC battery cell lifetime prediction results include the corrected remaining lifetime value, the dynamic decay trajectory of battery capacity, and the confidence probability of the prediction results.

[0060] Please see Figure 2 The specific steps for obtaining the set of high-frequency disturbance events are as follows:

[0061] S111: Identify load change events in the operating voltage and current sequences of IDC battery cells, perform discrete difference operations on the operating voltage and current sequences along the time axis to obtain voltage gradient vectors and current gradient vectors, anchor the sampling time when the magnitude of the voltage gradient vector and current gradient vector exceeds the preset steady-state noise reference value as the disturbance trigger point, retrieve the signal recovery time after the disturbance trigger point as the disturbance termination point, extract the corresponding voltage data segment and current data segment from the original sequence according to the time coordinates of the disturbance trigger point and the disturbance termination point, perform noise removal, cleaning and serialization on the extracted data segments, and generate a load transient response event sequence;

[0062] The process of setting the steady-state noise benchmark value is as follows: when the IDC battery is in a stable operating state of constant floating charge or no-load static, the original voltage and current signals within a preset time period under the stable state are collected, the first-order discrete difference operation is performed on the signal data, the gradient fluctuation sequence of the background noise is extracted, the gradient fluctuation sequence is statistically analyzed, the mean and standard deviation are calculated, and the superposition value of the mean and three times the standard deviation is used as the steady-state noise benchmark value.

[0063] A single lead-acid battery cell with a nominal voltage of 12V in the IDC server room was selected as the monitoring object. A high-frequency Hall current sensor and a precision resistor voltage divider network were used to capture the operating voltage and current sequences in real time at a sampling frequency of 1000Hz, as shown in Table 1. The raw data segment from 1000ms to 1010ms was extracted, and discrete difference operations were performed on the operating voltage and current sequences along the time axis. The difference between two adjacent sampling points was calculated and divided by the sampling interval (0.001s) to obtain the voltage gradient vector and current gradient vector. 5000 sample points were collected during the steady-state period of the battery's no-load floating charge, and the first-order difference mean of its current fluctuation was calculated to be 0.02A / The standard deviation is 0.01A / s. Based on the principle of superimposing three times the standard deviation on the mean, the steady-state noise reference value is set to 0.05A / s. The sampling time when the magnitude of the current gradient vector reaches 13333A / s (i.e., t=1005ms) is anchored as the disturbance trigger point. The disturbance is then retrieved until the magnitude of the current gradient falls back below 0.05A / s (t=1008ms) as the disturbance termination point. Based on the time coordinates of the disturbance trigger point and the disturbance termination point, the corresponding voltage data segment and current data segment are extracted from the original sequence. The extracted data segments are denoised, cleaned, and serialized to remove single-point glitches and noise and unify the data length, generating a load transient response event sequence.

[0064] Table 1. Data Collection Segments of Individual Battery Cells

[0065] As shown in Table 1, the data demonstrates the voltage drop and current surge before and after load mutation, providing a basic data source for subsequent gradient calculations.

[0066] S112: Call the load transient response event sequence, extract the voltage waveform segment and current waveform segment included in each independent response event in the sequence one by one, perform discrete Fourier transform processing on the voltage waveform segment and current waveform segment, construct a frequency domain energy distribution spectrum, search for the frequency axis coordinate value corresponding to the energy peak in the frequency domain energy distribution spectrum, calculate the relative deviation between the frequency axis coordinate value and the fundamental frequency value and quantize it into the load change frequency parameter of the event, and form a one-dimensional array according to the time order of the event occurrence to generate the event change frequency feature vector;

[0067] Fast Discrete Fourier Transform (FFT) is performed on voltage and current waveform segments. A Hanning window is selected as the windowing function to suppress spectral leakage, converting the time-domain signal into a frequency-domain signal. A frequency-domain energy distribution spectrum is constructed, and the frequency point with the highest energy amplitude is searched in the frequency-domain energy distribution spectrum. The dominant frequency component of the event is identified as 350Hz, which is the frequency axis coordinate value corresponding to the energy peak. The fundamental frequency value of the DC bus where the battery is located is obtained (for IDC DC power supply systems, the fundamental frequency is usually 0Hz or the ripple frequency after rectification, such as 300Hz, which is set to 0Hz here). The absolute difference between 350Hz and 0Hz is calculated as the relative deviation, and this relative deviation is quantized into the load change frequency parameter of the event. The calculated load change frequency parameters are arranged into a one-dimensional array according to the time sequence of the event to generate the event change frequency feature vector.

[0068] S113: Based on the event change frequency feature vector, a preset frequency limit threshold is used as the screening benchmark to perform a traversal numerical comparison of the load change frequency parameters included in the vector, identify and lock the target event index number whose load change frequency parameter value is greater than the frequency limit threshold, and retrieve the corresponding voltage and current waveform segments from the load transient response event sequence according to the target event index number. The retrieved waveform segments are then reassembled and high-frequency attribute identifier bits are added to construct a high-frequency disturbance event set.

[0069] The process of setting the frequency limit threshold is as follows: by performing electrochemical impedance spectroscopy tests on batteries of the same model, the boundary features between the high-frequency region and the mid-frequency region are traced in the generated Nyquist plot, the topological connection inflection point between the charge transfer impedance semicircular region and the low-frequency Warburg diffusion impedance straight line is located, the characteristic frequency value corresponding to the inflection point is extracted, and it is set as the frequency limit threshold.

[0070] For a valve-regulated sealed lead-acid battery with a nominal capacity of 100Ah, an impedance spectrum scan was performed in the range of 100kHz to 0.01Hz using an electrochemical workstation. The intersection of the semicircular arc representing charge transfer impedance (Rct) and the sloping line representing Warburg impedance was located in the Nyquist plot. The frequency corresponding to the inflection point of this intersection was extracted as 200Hz, and 200Hz was set as the frequency limit threshold. The physical meaning of this threshold is to distinguish whether the external disturbance is faster than the diffusion response speed of the internal ions. The elements in the feature vector were judged, and 350Hz, 420Hz, and 800Hz were identified as being greater than 200Hz. The target event index numbers corresponding to these values ​​were locked. Based on the target event index numbers, the corresponding voltage and current waveform segments were retrieved back from the load transient response event sequence. The retrieved waveform segments were aggregated and recombined, and a high-frequency attribute flag "HF_Tag=1" was added to construct a high-frequency disturbance event set.

[0071] Please see Figure 3 The specific steps for obtaining the disturbance intensity score are as follows:

[0072] S211: Based on a set of high-frequency disturbance events, traverse the voltage response curve and current response curve of each load change event in the set, calculate the absolute difference between steady-state voltage and valley voltage, obtain the voltage drop value, perform time dimension differentiation operation on the current response curve to obtain the current amplitude change rate, perform integration operation on the product of voltage and current within the event window to obtain the energy throughput value, and generate transient electric stress feature vector.

[0073] For an event numbered ID_05, its steady-state voltage is read as 13.5V, and the valley voltage at the moment of load impact is 12.8V. The absolute difference between the two is calculated to be 0.7V, and the voltage drop value is obtained. The time dimension differentiation operation is performed on the current response curve. Within 0.003s when the current jumps from 5A to 45A, its average change rate is calculated to be 13333A / s, and the current amplitude change rate is obtained. The trapezoidal integral operation is performed on the product of voltage and current within the event window, and the total energy released during this short-term impact process is calculated to be 29.25Joules, and the energy throughput value is obtained. A transient electrical stress feature vector is generated.

[0074] S212: Call the transient electrical stress feature vector, locate the original data segment in the high-frequency disturbance event set according to the time index corresponding to each element in the vector, identify the ohmic internal resistance step at the event trigger moment, obtain the short-term increase value of the internal resistance, simultaneously extract the duration of the event as the event duration, and collect the remaining battery charge at the start of the event as the charge state value, and establish the impedance state parameter matrix.

[0075] Using Ohm's law, the ratio of the instantaneous voltage drop to the instantaneous current increment at the moment the event is triggered is calculated to obtain the transient DC internal resistance. This transient DC internal resistance is then compared with the steady-state internal resistance before the event (e.g., ...). By performing a difference operation, the internal resistance step quantity is obtained as follows: The short-term increase value of internal resistance is obtained, and the time span during which the current remains at a high level is extracted as 0.05s as the event duration. The remaining capacity percentage reported by the battery management system (BMS) at the start of the event is collected as 85%, i.e. 0.85, as the charge state value, and an impedance state parameter matrix is ​​established.

[0076] S213: Using the transient electrical stress eigenvector and impedance state parameter matrix, the voltage drop, current amplitude change rate, energy throughput, short-term internal resistance increase, event duration, and charge state values ​​are all converted into dimensionless normalized forms using the following formula:

[0077] ;

[0078] The disturbance intensity score is obtained through calculation.

[0079] in, This represents the disturbance intensity score. This represents the normalized voltage drop, obtained by calculating the absolute difference between the steady-state voltage and the valley voltage and dividing it by the battery's rated voltage reference. The normalized rate of change of current amplitude is obtained by performing a differential operation on the current response curve and dividing it by the system's preset maximum rate of change of current threshold. This represents the energy weighting coefficient, which is a preset dimensionless constant. This represents the short-term increase in normalized internal resistance, obtained by identifying the step change in ohmic internal resistance and dividing it by the initial internal resistance value of the battery. The normalized energy throughput value is obtained by integrating the product of voltage and current and dividing by the battery's rated energy capacity. The duration-related factor is a pre-defined dimensionless constant. This represents the normalized event duration, obtained by extracting the event duration and dividing it by a preset standard time unit. This represents the state of charge, which is the ratio of the battery's remaining charge to its total capacity at the start of the collected event.

[0080] formula:

[0081] ;

[0082] The advantage of the formula is that it passes through the first term within the square root. The mechanical tearing stress on the physical structure of the electrode caused by drastic voltage fluctuations and rapid current surges is characterized by the second term. The Joule thermal shock effect, characterized by a sudden increase in internal resistance accompanied by high-energy throughput, is fused through geometric weighting and superimposed with a third term. To reflect the low power state ( Smaller) long-term continuous disturbance ( The (larger) amplification effect on the disruption of chemical equilibrium is shown in Table 2.

[0083] Let's perform the calculation using a real-world example:

[0084] Calculate the normalized voltage drop value : Assume a voltage drop of 0.7V, battery rated voltage 12V. ;

[0085] Calculate the rate of change of normalized current amplitude The rate of change is set at 13333 A / s, and the system's preset maximum tolerance rate of change is 50000 A / s. ;

[0086] Calculate the short-time increase value of normalized internal resistance Take the increase in internal resistance Initial internal resistance , ;

[0087] Calculate the normalized energy throughput value The energy is taken as 150J, and the rated energy capacity (12V 100Ah 3600s) is 4320000J. Since the energy of a single disturbance is extremely small, a normalized scaling factor is used for ease of calculation. After correction ;

[0088] Calculate the duration of normalized events : The duration is 0.05s, and the standard time unit is 1 second. ;

[0089] Substitute into the formula and calculate: Square the first term: ;

[0090] Second product: ;

[0091] Summation within the square root: ;

[0092] The result of taking the square root: ;

[0093] Third calculation: ;

[0094] Final summation: ;

[0095] The calculated result of 0.2324 represents the comprehensive damage intensity caused to the battery by a single high-frequency load disturbance. The larger the value, the more significant the acceleration effect of the disturbance on the battery's lifespan. A disturbance intensity score is generated.

[0096] Table 2. Parameter Setting Table for Disturbance Intensity Scoring Formula

[0097] Please see Figure 4 The specific steps for obtaining the proportion of high-frequency disturbance intensity are as follows:

[0098] S311: Based on the disturbance intensity score, set the start and end timestamps of the detection cycle to construct an effective statistical window, retrieve the disturbance intensity score values ​​that fall within the statistical window and perform numerical accumulation on them, aggregate the discretely distributed single disturbance impact intensity into scalar data that characterizes the cumulative damage effect in the high-frequency interval within the cycle, and generate a high-frequency cumulative stress index.

[0099] The detection period is set to start at 00:00:00 and end at 23:59:59 to construct a 24-hour effective statistical window. All disturbance intensity scores falling within this window are retrieved. It is assumed that three high-frequency disturbance events are captured within this period, with scores of 0.2324, 0.1500, and 0.3100 respectively. These three scores are then summed. The discretely distributed single-instance disturbance impact intensity is aggregated into scalar data characterizing the cumulative damage effect in the high-frequency range within the period, generating a high-frequency cumulative stress index.

[0100] S312: Call the load transient response event sequence, traverse all load change events recorded in the sequence and extract the voltage drop and current change rate parameters one by one, calculate the stress intensity of each event, and perform a summation operation on all intensity values ​​within the detection cycle to quantify the overall load fluctuation energy, including low frequency and high frequency, and generate the total load intensity value for the entire cycle.

[0101] Iterate through all load change events recorded in the sequence, including low-frequency events with frequencies below 200Hz (such as normal charging and discharging switching), extract voltage drop and current change rate parameters one by one, and use linear weighted logic (such as directly summing the normalized voltage and current product) to calculate the stress intensity of each event. Set a total of 100 load changes to occur throughout the day, and calculate the sum of the intensity values ​​of all events as 15.5. Quantify the overall load fluctuation energy, including low and high frequencies, and generate the total load intensity value for the entire cycle.

[0102] S313: Call the high-frequency cumulative stress index and the total intensity value of the full-cycle load to construct the ratio relationship between the two. Divide the high-frequency cumulative stress index by the total intensity value of the full-cycle load to obtain the dimensionless weighting coefficient. Convert the weighting coefficient into a percentage value and perform precision truncation on it to retain the significant numbers and generate the high-frequency disturbance intensity ratio.

[0103] By comparing the high-frequency cumulative stress index (0.6924) with the total strength value of the full-cycle load (15.5), a ratio relationship between the two is established. The high-frequency cumulative stress index is then divided by the total strength value of the full-cycle load to calculate... The dimensionless weighting coefficient is obtained, converted into a percentage value (4.467%), and then truncated to retain two significant figures, resulting in 4.47%. This value directly reflects the proportion of high-frequency disturbance components in the battery load pressure throughout the day, generating the high-frequency disturbance intensity proportion.

[0104] Please see Figure 5 The specific steps for obtaining the anti-disturbance critical value are as follows:

[0105] S411: Obtain the discharge capacity sequence of IDC battery cells during historical operation, perform first-order differential operation on the discharge capacity sequence with respect to the time dimension, quantify the capacity decay rate, call the high-frequency disturbance intensity ratio, time-align the capacity decay rate at each detection moment with the high-frequency disturbance intensity ratio at the corresponding moment, calculate the covariance of the two sets of data sequences and divide it by the product of the standard deviation, obtain the Pearson correlation coefficient, filter the data groups whose correlation coefficients meet the preset conditions, and generate decay-disturbance coupled data pairs;

[0106] The process of selecting data groups whose correlation coefficients meet the preset conditions is as follows: determine the statistical degrees of freedom based on the total number of sample points in the data group; construct a set test logic to test whether there is a linear relationship between variables; calculate the corresponding t-statistic using the Pearson correlation coefficient; call the preset significance level parameter; determine the critical t-value of the two-tailed test in combination with the statistical degrees of freedom; compare the absolute value of the t-statistic with the critical t-value; only when the absolute value of the t-statistic exceeds the critical t-value is the Pearson correlation coefficient determined to be statistically significant, and the corresponding data group is marked as a valid sample and retained; otherwise, it is removed.

[0107] Capacity data from the past 300 days were selected, and the first-order differential operation was performed on the discharge capacity sequence with respect to the time dimension to quantify the capacity decay rate (Ah / day). The proportion of high-frequency disturbance intensity was called, and the capacity decay rate at each detection time was time-series aligned with the proportion of high-frequency disturbance intensity at the corresponding time to obtain 300 data pairs. The covariance of the two data sequences was calculated and divided by the product of the standard deviations to obtain a Pearson correlation coefficient of -0.85. Data groups with correlation coefficients that meet the preset conditions were selected. Based on the total number of sample points (300) in the data group, the statistical degrees of freedom were determined to be 298. A set test logic for testing whether there is a linear relationship between variables was constructed. The corresponding t-statistic value was calculated using the Pearson correlation coefficient and was 27.5. The preset significance level parameter ( The critical t-value for the two-tailed test was determined to be 1.96 based on the statistical degrees of freedom. The absolute value of the t-statistic, 27.5, was compared with the critical t-value of 1.96. It was determined that 27.5 was greater than 1.96, confirming that the Pearson correlation coefficient was statistically significant. The corresponding data set was marked as a valid sample and retained, generating attenuation-perturbation coupled data pairs.

[0108] S412: Call the attenuation-perturbation coupled data pair, project the data pair onto the feature space with the proportion as the horizontal axis and the rate as the vertical axis, use the least squares method to perform piecewise fitting on the scatter distribution in the feature space, set the linear function and the exponential function, find the best fitting breakpoint, calculate the horizontal axis slicing position corresponding to the minimum sum of squared fitting residuals, identify the feature node where the capacity decay mode changes from linear accumulation to exponential, and generate the coordinates of the nonlinear abrupt change inflection point.

[0109] Projecting the data pairs onto a feature space with proportion as the horizontal axis and rate as the vertical axis, the data shows a trend where the capacity decay rate accelerates nonlinearly as the proportion of high-frequency disturbances increases. The least squares method is used to perform piecewise fitting on the scatter distribution within the feature space, and a linear function is defined. With exponential function Find the best fit breakpoint within the percentage range [0%, 20%], calculate the horizontal axis slicing position corresponding to the minimum sum of squared fitting residuals, and find that when the proportion of high-frequency disturbances exceeds 8.5%, the data distribution is more in line with the characteristics of an exponential function. Identify the feature node of the capacity decay mode changing from linear accumulation to exponential, and generate the coordinates of the nonlinear mutation inflection point.

[0110] S413: Call the coordinates of the nonlinear mutation inflection point, analyze the specific value of the disturbance intensity ratio corresponding to the coordinates on the horizontal axis of the feature space, introduce the preset statistical confidence interval parameter to perform boundary convergence correction on the value, remove outlier sample points caused by measurement error or transient interference, and solidify the corrected value as the limit boundary characterizing the battery's ability to maintain linear capacity decay under high-frequency disturbance environment, and generate the anti-disturbance critical value.

[0111] The process of introducing a preset statistical confidence interval parameter to correct the boundary convergence of numerical values ​​and remove outlier sample points caused by measurement errors or transient interference is as follows: Select multiple disturbance intensity percentage values ​​located in the neighborhood of the nonlinear abrupt change inflection point coordinates in the feature space to form a sequence to be corrected; calculate the arithmetic mean and standard deviation of the sequence to be corrected; according to the confidence level corresponding to the preset statistical confidence interval parameter, look up the standard normal distribution table to obtain the coverage factor value; calculate the product of the standard deviation and the coverage factor value to obtain the error tolerance value; superimpose and subtract the error tolerance value based on the arithmetic mean to construct a closed interval for judging numerical validity; traverse the sequence to be corrected; and determine the values ​​that fall outside the closed interval for judging numerical validity as outlier sample points and remove them from the sequence.

[0112] The coordinates of the nonlinear mutation inflection point (8.5%) are used to analyze the specific value of the disturbance intensity percentage corresponding to the coordinates on the horizontal axis of the feature space. Ten disturbance intensity percentage values ​​located in the neighborhood of the nonlinear mutation inflection point coordinates (e.g., 8.0% to 9.0%) are selected in the feature space to form a sequence to be corrected. The arithmetic mean of the sequence to be corrected is calculated to be 8.55%, and the standard deviation is 0.1%. A preset statistical confidence interval parameter (confidence level 95%) is introduced. The coverage factor value is obtained by querying the standard normal distribution table and is 1.96. The product of the standard deviation and the coverage factor value is calculated to obtain the error tolerance value of 0.196%. The error tolerance value is superimposed and subtracted based on the arithmetic mean to construct a closed interval for judging the validity of the numerical value [8.354%, 8.746%]. Outlier sample points caused by measurement errors or transient interference (e.g., a certain point is 9.2%) are removed. The corrected value is solidified as the limit boundary characterizing the battery's ability to maintain linear capacity decay under high-frequency disturbance environment, generating the anti-disturbance critical value.

[0113] Please see Figure 6 The specific steps for obtaining the IDC battery cell lifetime prediction results are as follows:

[0114] S511: Based on the disturbance rejection critical value, it is projected onto the preset population disturbance rejection distribution probability density function coordinate system. The cumulative distribution probability value of the disturbance rejection critical value under the distribution curve is calculated by numerical integration method. The relative disturbance rejection position of the sample in the population is quantified. The cumulative distribution probability value is used as the index key to retrieve the corresponding dimensionless weight scalar in the preset nonlinear adjustment coefficient mapping table and generate the capacity stability adjustment factor.

[0115] Based on the immunity threshold (8.55%), it is projected onto a pre-defined coordinate system of the group immunity distribution probability density function. This distribution function is a Gaussian distribution (mean 8.0%, standard deviation 1.0%) fitted based on historical immunity test data of 5000 batteries in the same batch of the data center. The cumulative distribution probability of the immunity threshold of 8.55% under the distribution curve is calculated using the numerical integration method. The result is 0.7088, which quantifies the relative perturbation resistance position of the sample in the population, indicating that the battery is superior to 70.88% of similar batteries. The cumulative distribution probability value of 0.7088 is used as the index key to retrieve the corresponding dimensionless weight scalar in the preset nonlinear adjustment coefficient mapping table. The mapping table specifies that the coefficient corresponding to the probability value in the interval [0.7, 0.8) is 0.95 (indicating slow decay), and the capacity stability adjustment factor is generated.

[0116] S512: Input the capacity stability adjustment factor into the forget gate control logic of the long short-term memory network model, use the factor to perform element-wise weighted calculation on the historical state retention weight matrix of the hidden layer unit of the model, obtain the reset weight parameters, drive the model to dynamically correct the instantaneous decline slope of the battery life decay curve provided by the manufacturer at the current time step according to the reset weight parameters, and generate the corrected decay gradient vector.

[0117] The capacity stability adjustment factor (0.95) is input into the forgetting gate control logic of the Long Short-Term Memory (LSTM) network model. The original forgetting gate weight matrix of the model... Responsible for controlling the degree to which cell state information from the previous time step is preserved, the historical state retention weight matrix of the hidden layer units in the model is weighted element-wise using a factor of 0.95 to obtain the reset weight parameters, i.e. By appropriately reducing the model's dependence on historical long-term stable data, it becomes more sensitive to trend changes caused by recent high-frequency disturbances. The model is driven to dynamically correct the instantaneous decline slope of the battery life degradation curve provided by the manufacturer at the current time step based on the reset weight parameters, and generate the corrected degradation gradient vector.

[0118] S513: Call the corrected decay gradient vector, use the actual battery capacity at the current moment as the starting boundary condition to perform discrete time step integral extrapolation, construct the capacity evolution trajectory covering the entire future life cycle, calculate the time axis coordinate point where the capacity evolution trajectory and the preset scrap capacity cutoff line intersect through the geometric intersection algorithm, and generate IDC battery cell life prediction results.

[0119] Using the current actual battery capacity (e.g., 85Ah) as the initial boundary condition, a discrete-time step-by-step integral extrapolation operation is performed. The capacity is gradually reduced according to the corrected gradient to construct a capacity evolution trajectory covering the entire future life cycle. The time axis coordinate point where the capacity evolution trajectory intersects with the preset scrap capacity cutoff line (usually 80% of the nominal capacity, i.e., 80Ah) is calculated using a geometric intersection algorithm. The date corresponding to the intersection point is calculated to be December 15, 2026, generating the IDC battery cell life prediction result.

[0120] An IDC battery cell lifetime prediction system is provided. This system is used to implement the aforementioned IDC battery cell lifetime prediction method. The system includes:

[0121] The disturbance event identification module identifies load change events in the operating voltage and current sequences of individual IDC battery cells, calculates the load change frequency, filters events whose load change frequency exceeds the frequency limit, and constructs a set of high-frequency disturbance events.

[0122] The disturbance intensity assessment module extracts the voltage drop, current amplitude change rate, and short-term internal resistance increase from the set of high-frequency disturbance events, and calculates and generates a disturbance intensity score.

[0123] The disturbance intensity analysis module accumulates the disturbance intensity scores within the detection period to obtain the total intensity of high-frequency disturbances, calculates the percentage of the total intensity of high-frequency disturbances to the total intensity of load change events, and generates the high-frequency disturbance intensity percentage.

[0124] The anti-disturbance critical identification module obtains the discharge capacity sequence of IDC battery cells, calculates the capacity decay rate, analyzes the Pearson correlation coefficient between the proportion of high-frequency disturbance intensity and the capacity decay rate, extracts the inflection point where the capacity decay rate turns into an exponential change, and defines the anti-disturbance critical value.

[0125] The lifetime prediction and analysis module calculates the capacity stability adjustment factor based on the position of the disturbance rejection threshold in the population disturbance rejection distribution curve. The capacity stability adjustment factor is then input into the long short-term memory network model to correct the descent gradient of the battery lifetime degradation curve provided by the manufacturer, and outputs the IDC battery cell lifetime prediction results.

[0126] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for predicting the lifetime of a single IDC battery cell, characterized in that, Includes the following steps: S1: Identify load change events in the operating voltage and current sequences of individual IDC battery cells, calculate the load change frequency, filter events whose load change frequency exceeds the frequency limit, and construct a set of high-frequency disturbance events. S2: Extract the voltage drop, current amplitude change rate, and short-time increase in internal resistance from the set of high-frequency disturbance events, and calculate and generate a disturbance intensity score; S3: Accumulate the disturbance intensity scores within the detection period to obtain the total intensity of high-frequency disturbances, calculate the percentage of the total intensity of high-frequency disturbances to the total intensity of load change events, and generate the high-frequency disturbance intensity percentage. S4: Obtain the discharge capacity sequence of IDC battery cells, calculate the capacity decay rate, analyze the Pearson correlation coefficient between the high-frequency disturbance intensity ratio and the capacity decay rate, extract the inflection point where the capacity decay rate turns into an exponential change, and define the anti-disturbance critical value. S5: Calculate the capacity stability adjustment factor based on the position of the disturbance rejection critical value in the population disturbance rejection distribution curve, input the capacity stability adjustment factor into the long short-term memory network model to correct the descent gradient of the battery life decay curve provided by the manufacturer, and output the IDC battery cell life prediction result.

2. The IDC battery cell lifetime prediction method according to claim 1, characterized in that, The high-frequency disturbance event set includes the disturbance occurrence timestamp, instantaneous frequency value of load change, and event index identifier code. The disturbance intensity score includes a linearly weighted aggregate value, multi-parameter quantization weight coefficient, and intensity level classification label. The high-frequency disturbance intensity proportion includes the cumulative total value of high-frequency disturbance intensity, the total intensity value of load change over the entire cycle, and the intensity proportion percentage value. The anti-disturbance critical value includes the proportion value corresponding to the capacity decay inflection point, the initiation slope of exponential change, and the Pearson correlation confidence interval. The IDC battery cell lifetime prediction result includes the corrected remaining lifetime value, the battery capacity dynamic decay trajectory, and the prediction result confidence probability.

3. The IDC battery cell lifetime prediction method according to claim 2, characterized in that, The specific steps for obtaining the set of high-frequency disturbance events are as follows: S111: Identify load change events in the operating voltage and current sequences of IDC battery cells, perform discrete difference operations on the operating voltage and current sequences along the time axis to obtain voltage gradient vectors and current gradient vectors, anchor the sampling time when the magnitude of the voltage gradient vector and current gradient vector exceeds the preset steady-state noise reference value as the disturbance trigger point, retrieve the signal recovery time after the disturbance trigger point as the disturbance termination point, extract the corresponding voltage data segment and current data segment from the original sequence according to the time coordinates of the disturbance trigger point and the disturbance termination point, perform noise removal, cleaning and serialization on the extracted data segments, and generate a load transient response event sequence; S112: Call the load transient response event sequence, extract the voltage waveform segment and current waveform segment included in each independent response event in the sequence one by one, perform discrete Fourier transform processing on the voltage waveform segment and current waveform segment, construct a frequency domain energy distribution spectrum, search for the frequency axis coordinate value corresponding to the energy peak in the frequency domain energy distribution spectrum, calculate the relative deviation between the frequency axis coordinate value and the fundamental frequency value and quantize it into the load change frequency parameter of the event, and form a one-dimensional array according to the time order of the event occurrence to generate the event change frequency feature vector; S113: Based on the event change frequency feature vector, a preset frequency limit threshold is used as the screening criterion to perform a traversal numerical comparison of the load change frequency parameters included in the vector, identify and lock the target event index number whose load change frequency parameter value is greater than the frequency limit threshold, and retrieve the corresponding voltage and current waveform segments from the load transient response event sequence according to the target event index number, collect and reorganize the retrieved waveform segments and add high-frequency attribute identifier bits to construct a high-frequency disturbance event set.

4. The IDC battery cell lifetime prediction method according to claim 3, characterized in that, The process of setting the steady-state noise benchmark value is as follows: when the IDC battery is in a stable operating state of constant floating charge or no-load static, the original voltage and current signals within a preset time period under the stable state are collected, the first-order discrete difference operation is performed on the signal data, the gradient fluctuation sequence of the background noise is extracted, the gradient fluctuation sequence is statistically analyzed, the mean and standard deviation are calculated, and the superposition value of the mean and three times the standard deviation is used as the steady-state noise benchmark value. The process of setting the frequency limit threshold is as follows: by performing electrochemical impedance spectroscopy tests on batteries of the same model, the boundary characteristics between the high-frequency region and the mid-frequency region are traced in the generated Nyquist plot, the topological connection inflection point between the charge transfer impedance semicircular region and the low-frequency Warburg diffusion impedance straight line is located, the characteristic frequency value corresponding to the inflection point is extracted, and it is set as the frequency limit threshold.

5. The IDC battery cell lifetime prediction method according to claim 4, characterized in that, The specific steps for obtaining the disturbance intensity score are as follows: S211: Based on the set of high-frequency disturbance events, traverse the voltage response curve and current response curve of each load change event in the set, calculate the absolute difference between steady-state voltage and valley voltage, obtain the voltage drop value, perform time dimension differentiation operation on the current response curve, obtain the current amplitude change rate, perform integration operation on the product of voltage and current within the event window, obtain the energy throughput value, and generate transient electric stress feature vector; S212: Call the transient electrical stress feature vector, locate the original data segment in the high-frequency disturbance event set according to the time index corresponding to each element in the vector, identify the ohmic internal resistance step at the event trigger time, obtain the short-term increase value of the internal resistance, simultaneously extract the duration of the event as the event duration, and collect the remaining battery power at the start of the event as the charge state value, and establish the impedance state parameter matrix. S213: Call the transient electric stress characteristic vector and impedance state parameter matrix, and convert the voltage drop value, current amplitude change rate, energy throughput value, short-term internal resistance increase value, event duration and charge state value into dimensionless normalized form, and calculate to obtain the disturbance intensity score.

6. The IDC battery cell lifetime prediction method according to claim 5, characterized in that, The specific formula for obtaining the disturbance intensity score is as follows: ； in, This represents the disturbance intensity score. This represents the normalized voltage drop value. Represents the rate of change of the normalized current amplitude. Represents the energy weighting coefficient. This represents the short-term increase value of the normalized internal resistance. Represents the normalized energy throughput value. Represents the duration of the influence factor. Duration of the normalization event Represents the numerical value of the charge state.

7. The IDC battery cell lifetime prediction method according to claim 6, characterized in that, The specific steps for obtaining the proportion of high-frequency disturbance intensity are as follows: S311: Based on the disturbance intensity score, set the start and end timestamps of the detection period to construct an effective statistical window, retrieve the disturbance intensity score values ​​that fall within the statistical window and perform numerical accumulation on them, aggregate the discretely distributed single disturbance impact intensity into scalar data that characterizes the cumulative damage effect in the high-frequency interval within the period, and generate a high-frequency cumulative stress index. S312: Call the load transient response event sequence, traverse all load change events recorded in the sequence and extract the voltage drop and current change rate parameters one by one, calculate the stress intensity of each event, and perform a summation operation on all intensity values ​​within the detection period to quantify the overall load fluctuation energy including low frequency and high frequency, and generate the total load intensity value for the entire cycle. S313: Call the high-frequency cumulative stress index and the total load strength value of the whole cycle to construct the ratio relationship between the two. Divide the high-frequency cumulative stress index by the total load strength value of the whole cycle to obtain the dimensionless weighting coefficient. Convert the weighting coefficient into a percentage value and perform precision truncation processing on it to retain the significant numbers and generate the high-frequency disturbance intensity ratio.

8. The IDC battery cell lifetime prediction method according to claim 7, characterized in that, The specific steps for obtaining the anti-interference critical value are as follows: S411: Obtain the discharge capacity sequence of the IDC battery cell during its historical operation, perform a first-order differential operation on the discharge capacity sequence with respect to the time dimension to quantify the capacity decay rate, call the high-frequency disturbance intensity ratio, align the capacity decay rate at each detection moment with the high-frequency disturbance intensity ratio at the corresponding moment, calculate the covariance of the two sets of data sequences and divide it by the product of the standard deviations, obtain the Pearson correlation coefficient, filter the data groups whose correlation coefficients meet the preset conditions, and generate decay-disturbance coupled data pairs. S412: Call the attenuation-perturbation coupled data pair, project the data pair onto the feature space with the proportion as the horizontal axis and the rate as the vertical axis, use the least squares method to perform piecewise fitting on the scatter distribution in the feature space, set the linear function and the exponential function, find the best fitting breakpoint, calculate the horizontal axis slicing position corresponding to the minimum sum of squared fitting residuals, identify the feature node where the capacity attenuation mode changes from linear accumulation to exponential, and generate the coordinates of the nonlinear abrupt change inflection point. S413: Call the coordinates of the nonlinear mutation inflection point, analyze the specific value of the disturbance intensity ratio corresponding to the coordinates on the horizontal axis of the feature space, introduce the preset statistical confidence interval parameter to perform boundary convergence correction on the value, eliminate outlier sample points caused by measurement error or transient interference, and solidify the corrected value as the limit boundary characterizing the battery's ability to maintain linear capacity decay under high-frequency disturbance environment, and generate the anti-disturbance critical value.

9. The IDC battery cell lifetime prediction method according to claim 8, characterized in that, The specific steps for obtaining the IDC battery cell lifetime prediction results are as follows: S511: Based on the aforementioned anti-interference critical value, project it onto a preset population anti-interference distribution probability density function coordinate system, use numerical integration to calculate the cumulative distribution probability value of the anti-interference critical value under the distribution curve, quantify the relative anti-interference position of the sample in the population, use the cumulative distribution probability value as an index key to retrieve the corresponding dimensionless weight scalar in a preset nonlinear adjustment coefficient mapping table, and generate a capacity stability adjustment factor. S512: Input the capacity stability adjustment factor into the forget gate control logic of the long short-term memory network model, use the factor to perform element-wise weighted calculation on the historical state retention weight matrix of the hidden layer unit of the model, obtain the reset weight parameters, drive the model to dynamically correct the instantaneous decline slope of the battery life decay curve provided by the manufacturer at the current time step according to the reset weight parameters, and generate the corrected decay gradient vector. S513: Call the corrected decay gradient vector, use the actual battery capacity at the current moment as the starting boundary condition to perform discrete-time step-by-step integral extrapolation, construct the capacity evolution trajectory covering the entire future life cycle, calculate the time axis coordinate point where the capacity evolution trajectory and the preset scrap capacity cutoff line intersect through the geometric intersection algorithm, and generate IDC battery cell life prediction results.

10. An IDC battery cell lifetime prediction system, characterized in that, The system is used to implement the IDC battery cell lifetime prediction method according to any one of claims 1-9, the system comprising: The disturbance event identification module identifies load change events in the operating voltage and current sequences of individual IDC battery cells, calculates the load change frequency, filters events whose load change frequency exceeds the frequency limit, and constructs a set of high-frequency disturbance events. The disturbance intensity assessment module extracts the voltage drop, current amplitude change rate, and short-term internal resistance increase from the set of high-frequency disturbance events, and calculates and generates a disturbance intensity score. The disturbance intensity analysis module accumulates the disturbance intensity scores within the detection period to obtain the total intensity of high-frequency disturbances, calculates the percentage of the total intensity of high-frequency disturbances to the total intensity of load change events, and generates the high-frequency disturbance intensity percentage. The anti-interference critical identification module obtains the discharge capacity sequence of IDC battery cells, calculates the capacity decay rate, analyzes the Pearson correlation coefficient between the high-frequency disturbance intensity ratio and the capacity decay rate, extracts the inflection point where the capacity decay rate turns into an exponential change, and defines the anti-interference critical value. The lifetime prediction and analysis module calculates the capacity stability adjustment factor based on the position of the disturbance rejection threshold in the population disturbance rejection distribution curve, inputs the capacity stability adjustment factor into the long short-term memory network model to correct the descent gradient of the battery lifetime degradation curve provided by the manufacturer, and outputs the IDC battery cell lifetime prediction result.

Images