A battery fault early warning and diagnosis method, system, device and storage medium

Abstract

The application relates to the technical field of battery detection, and specifically provides a battery fault early warning and diagnosis method, which comprises the following steps: acquiring the operation parameters of a battery pack in real time, processing the operation parameters based on preset judgment conditions, generating corresponding pretreatment signals according to noise environment states, extracting space-time features to perform weak signal enhancement processing, and generating an enhanced feature set; performing space-time fusion calculation based on the enhanced feature set according to a preset weight relationship to obtain a fault energy accumulation value; dynamically correcting a fault judgment threshold value according to a battery health state and real-time environment parameters; and outputting a graded early warning signal when the fault energy accumulation value exceeds the dynamically corrected fault judgment threshold value. Through capturing weak fault features of the battery pack, combining space domain feature extraction and signal enhancement technology of temperature difference and voltage distribution, and improving the detection sensitivity of early hidden faults, the time for fault detection is shortened, the false alarm rate is reduced, and the like in the lithium battery safety early warning scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery detection, in particular to a battery fault early warning and diagnosis method and system. BACKGROUND

[0002] As an environmentally sensitive, noise-sensitive complex nonlinear system, lithium ion batteries have numerous fault types, hidden fault characteristics, and great individual differences. The fault diagnosis of lithium ion batteries faces the dual challenges of high misdiagnosis rate and high missed diagnosis rate, and it is extremely difficult to accurately find and locate faults. In addition, the development speed of lithium ion battery faults is extremely fast. For example, internal short circuit of the battery will cause the battery to release a large amount of energy and generate heat in a short time, and even burn and explode. In order to avoid such vicious accidents, the battery management system (BMS) must have the ability to find and identify early faults, shorten the fault diagnosis time, and realize safety early warning to minimize the risk caused by battery faults.

[0003] Most of the existing fault diagnosis algorithms use threshold diagnosis method. When the fault causes the battery abnormal characteristic value to break through the preset threshold, the fault diagnosis system will alarm. In practical application, the traditional diagnosis algorithm often adopts a relatively conservative high threshold setting to avoid fault misdiagnosis caused by environmental noise. However, at the same time, the risk of missing early faults due to hidden characteristics is also high. If the threshold is lowered to reduce the fault omission rate, the lower threshold will be frequently touched by abnormal characteristic values caused by environmental noise, causing frequent misdiagnosis, affecting the normal operation of the battery system. It can be seen that the traditional fault diagnosis method based on threshold cannot solve the contradiction between misdiagnosis and missed diagnosis.

[0004] Therefore, the present application provides a battery fault early warning and diagnosis method to solve the above problems. SUMMARY

[0005] In order to overcome the shortcomings of the prior art, the present application provides a battery fault early warning and diagnosis method and system to solve the problems in the prior art.

[0006] One embodiment of the present application provides a battery fault early warning and diagnosis method, comprising the following steps:

[0007] Real-time acquisition of the operating parameters of the battery pack, branch processing of the operating parameters based on the preset environmental noise intensity judgment condition; wherein, if it is judged as a strong noise environment, noise stripping processing is performed on the operating parameters to generate a preprocessed signal; if it is judged as a non-strong noise environment, the operating parameters are taken as the preprocessed signal;

[0008] extracting spatio-temporal features from the preprocessed signal, the extracted spatio-temporal features including signal energy features of a preset target frequency band, cell group monomer temperature difference distribution features and monomer voltage distribution features; performing weak signal enhancement processing on the extracted spatio-temporal features to generate an enhanced feature set;

[0009] performing spatio-temporal fusion calculation on the spatio-temporal features in the enhanced feature set according to a preset weight relationship to obtain a fault energy accumulation value;

[0010] dynamically correcting a fault determination threshold according to a battery health state and real-time environmental parameters;

[0011] outputting a graded early warning signal when the fault energy accumulation value exceeds the dynamically corrected fault determination threshold.

[0012] In one of the embodiments, the step of dynamically correcting the fault determination threshold according to the battery health state and the real-time environmental parameters includes:

[0013] performing primary determination based on historical noise intensity records of an environment where the battery group is located, and directly determining that the environment is a non-strong noise environment if an average noise intensity in a past continuous preset time length is lower than a preset safety noise threshold;

[0014] otherwise, performing secondary determination, performing spectrum analysis on the currently obtained operating parameters, extracting noise component energy higher than a preset demarcation frequency, and determining that the environment is a strong noise environment if the noise component energy exceeds a dynamic noise threshold;

[0015] only when the secondary determination output is a strong noise environment, performing noise stripping processing to generate the preprocessed signal.

[0016] In one of the embodiments, the weak signal enhancement processing includes:

[0017] a) high frequency signal enhancement path:

[0018] performing frequency domain decomposition on the signal energy features of the preset target frequency band to extract a frequency band component with a preset bandwidth and a center frequency point determined by the real-time state of the battery;

[0019] calculating a dynamic gain value based on a ratio of a signal amplitude of the frequency band component to an environmental noise amplitude and applying the dynamic gain value to the frequency band component to obtain a high frequency enhanced component;

[0020] b) low frequency trend maintaining path:

[0021] performing sliding average filtering on the cell group monomer temperature difference distribution features and the monomer voltage distribution features, wherein a filtering window length is inversely proportional to a battery charge-discharge rate to obtain a low frequency enhanced feature;

[0022] c) Cross-scale fusion:

[0023] The high-frequency enhancement component and the low-frequency enhancement feature are weighted and superimposed by a preset fusion coefficient to generate the enhanced feature set.

[0024] In one embodiment, in the step of performing spatio-temporal fusion calculation according to a preset weight relationship based on the spatio-temporal features in the enhanced feature set to obtain the fault energy accumulation value, the preset weight relationship is a dynamic weight relationship, and the dynamic weight relationship is dynamically adjusted according to real-time working state parameters of the battery pack, wherein:

[0025] The real-time working state parameters include at least one of ambient temperature, battery charge / discharge rate, and battery state of charge;

[0026] The dynamic weight relationship includes a time domain feature weight, a first spatial domain feature weight, and a second spatial domain feature weight, wherein the time domain feature is a signal energy feature, the first spatial domain feature is a battery monomer temperature difference distribution feature, and the second spatial domain feature is a battery monomer voltage distribution feature; and the dynamic adjustment of the dynamic weight relationship satisfies:

[0027] When the ambient temperature is lower than a temperature threshold, the weight of the first spatial domain feature is increased;

[0028] When the charge / discharge rate is higher than a rate threshold, the weight of the time domain feature is increased;

[0029] The second spatial domain feature weight is kept within a preset threshold range.

[0030] In one embodiment, in the step of performing spatio-temporal fusion calculation according to a preset weight relationship based on the spatio-temporal features in the enhanced feature set to obtain the fault energy accumulation value, the calculation method of the fault energy accumulation value is:

[0031] The spatio-temporal features in the enhanced feature set are correspondingly mapped to the characteristic values and reference values of the fault energy calculation formula, and the fault energy accumulation value is calculated by the following formula:

[0032]

[0033] wherein, is the time The battery pack target feature value after weak signal enhancement, the target feature value is selected from at least one of the signal energy feature, the battery monomer temperature difference distribution feature, and the battery monomer voltage distribution feature; is the normal working condition reference value corresponding to the target feature value; is the feature sampling interval; is the cumulative calculation time length;

[0034] and the selection of the target characteristic value and the assignment of the corresponding adapt to the dynamic weight relationship, specifically satisfying:

[0035] When the ambient temperature is lower than the temperature threshold, the battery pack monomer temperature difference distribution characteristic is preferentially selected as for calculation, and the weight proportion of the monomer temperature difference normal baseline value is correspondingly increased;

[0036] When the charge-discharge rate is higher than the rate threshold, the signal energy characteristic is preferentially selected as for calculation, and the weight proportion of the signal energy normal baseline value is correspondingly increased.

[0037] In one embodiment, the setting of the normal working condition baseline value further includes:

[0038] Collecting battery pack historical operation data for preprocessing to obtain standardized feature data applied to a machine model as input data;

[0039] Training the battery individual health state model using a machine learning algorithm on the standardized feature data obtained after preprocessing to obtain a trained battery individual health state model, which receives real-time operation parameters of the battery pack and outputs a personalized normal working condition baseline interval dynamically adapted to the battery life cycle; wherein the personalized normal working condition baseline interval is dynamically updated with the number of battery cycles and the degree of aging;

[0040] When it is monitored that the health state dispersion degree of the monomer battery in the same batch of battery packs exceeds a preset threshold, the adaptive correction of the personalized normal working condition baseline interval is triggered, and the correction process is combined with real-time working state parameters for multidimensional dynamic adjustment.

[0041] In one embodiment, after the step of performing spatio-temporal fusion calculation on the spatio-temporal features in the enhanced feature set according to the preset weight relationship to obtain the fault energy accumulation value, the following steps are further included:

[0042] Obtaining historical fault data and real-time working state parameters, combining the calculation result of the fault energy accumulation value, and performing data standardization preprocessing to generate a multivariate time series feature data set;

[0043] Based on the standardized multivariate time series feature data set, a multivariate time series prediction model is constructed using a machine learning algorithm, and the multivariate time series prediction model is used to predict the fault energy growth rate in the future period;

[0044] The predicted failure energy growth rate is compared with the dynamically modified failure determination threshold change gradient in real time, and when the predicted failure energy growth rate exceeds the product of a preset safety factor and the failure determination threshold change gradient, an early warning mechanism is triggered;

[0045] According to the relative deviation degree of the predicted failure energy growth rate and the failure determination threshold change gradient, a plurality of early warning levels are divided, and the remaining safe operation time is calculated based on the current failure energy accumulation value, the dynamically modified failure determination threshold, and the predicted failure energy growth rate.

[0046] The application also relates to a battery failure warning and diagnosis system, comprising:

[0047] A data acquisition module is configured to acquire operation parameters of a battery pack in real time, perform branch processing on the operation parameters based on a preset environmental noise intensity determination condition, and generate preprocessed signals.

[0048] A noise stripping module is configured to perform noise stripping processing on the operation parameters to generate preprocessed signals when it is determined that the environment is a strong noise environment.

[0049] A feature extraction module is configured to extract spatiotemporal features from the preprocessed signals, wherein the extracted spatiotemporal features include signal energy features of a preset target frequency band, battery pack monomer temperature difference distribution features, and monomer voltage distribution features; and perform weak signal enhancement processing on the extracted spatiotemporal features to generate an enhanced feature set.

[0050] A fusion calculation module is configured to perform spatiotemporal fusion calculation on the spatiotemporal features in the enhanced feature set according to a preset weight relationship to obtain a failure energy accumulation value.

[0051] A dynamic modification module is configured to dynamically modify a failure determination threshold according to a battery health state and real-time environmental parameters.

[0052] An early warning output module is configured to output a graded early warning signal when the failure energy accumulation value exceeds the dynamically modified failure determination threshold.

[0053] The application also relates to a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above battery failure warning and diagnosis method when executing the computer program.

[0054] The application also relates to a computer readable storage medium storing a computer program, wherein the computer program is executable on a processor to implement the steps of the above battery failure warning and diagnosis method.

[0055] The battery fault early warning and diagnosis method and system provided by the above embodiments have the following beneficial effects:

[0056] 1. The branch processing is adapted to ambient noise to strip interference signals, and the weak fault characteristics of the battery pack are accurately captured under strong noise; the spatial feature extraction and signal enhancement technology of temperature difference and voltage distribution is combined to improve the detection sensitivity of early hidden faults; the fault energy accumulation value is calculated by time and space fusion to quantify the fault deterioration trend and predict the risk trajectory; finally, the battery health state and environmental parameters are dynamically corrected based on the threshold to realize accurate output of the graded early warning signal, which fundamentally solves the technical difficulties of high misdiagnosis rate, high missed diagnosis rate and response lag of traditional methods, and shortens the fault detection time, reduces the false alarm rate, etc. in the lithium battery safety warning scene.

[0057] 2. In one of the embodiments, by introducing a dynamic weight relationship, the time domain feature, the first spatial feature, and the second spatial feature weight are dynamically adjusted according to the real-time working state parameters of the battery pack (environmental temperature, battery charge and discharge rate, battery state of charge, etc.), effectively solving the problem that the traditional fixed weight diagnosis method cannot accurately match the importance of fault characteristics under different working conditions. In a low temperature environment, the weight of the battery pack single cell temperature difference distribution feature is increased to capture potential faults caused by uneven temperature; in a high charge and discharge rate working condition, the weight of the signal energy feature is increased to accurately identify abnormal energy fluctuations under high load. This dynamic weight mechanism breaks the limitations of the traditional threshold diagnosis method, avoids misdiagnosis and missed diagnosis caused by environmental changes or working condition differences, makes the fault diagnosis result more consistent with the actual operation state of the battery, and significantly improves the accuracy and reliability of fault diagnosis.

[0058] 3. In one of the embodiments, by deeply integrating the enhanced feature set and the fault energy calculation formula, the weak abnormal characteristics in the battery operation process are converted into quantifiable and accumulative fault energy values, solving the problem that the traditional threshold diagnosis method cannot effectively evaluate early hidden fault characteristics. Combined with the dynamic weight relationship, the appropriate target feature is selected for fault energy accumulation calculation under different working conditions, so that the fault energy accumulation value can truly reflect the fault development trend. Compared with the traditional diagnosis method, this calculation method realizes the leap from "static threshold judgment" to "dynamic energy accumulation analysis", accurately captures the feature changes at the early stage of the fault, significantly shortens the fault diagnosis time, reduces the misdiagnosis rate and missed diagnosis rate caused by feature misjudgment, provides a more scientific and accurate quantitative basis for early fault warning of lithium batteries, and effectively improves the safety and reliability of the battery management system.

[0059] 4、In one of the embodiments, the individual health state model is trained by collecting battery pack historical operation data, and a personalized normal working condition reference interval dynamically adapted to the battery aging degree is generated, breaking through the bottleneck that the traditional fixed threshold is difficult to match the individual differences of the battery. The reference interval combines with the dynamic weight rule (such as preferentially selecting the single cell temperature difference feature at low temperature), realizing the cooperative correction of the reference value and the feature weight: when the dispersion degree of the single cell health state exceeds the threshold value, the reference interval is adjusted in multiple dimensions based on the real-time working state parameters (environmental temperature, charge-discharge rate), ensuring that the temperature difference feature tolerance is improved at low temperature working condition, and the voltage abnormal sensitivity is enhanced at high rate working condition, thereby reducing the misjudgment rate.

[0060] 5、In one of the embodiments, based on the fault energy accumulation value (combining the dynamically selected feature value and the weight), a multivariate time series prediction model is constructed, realizing the leap from real-time diagnosis to trend prediction: through real-time comparison of the predicted fault energy growth rate and the gradient change of the dynamically corrected threshold value, when the predicted value exceeds the safety threshold value, a hierarchical early warning is triggered. At the same time, combined with the personalized reference interval correction mechanism, the accurate remaining safe operation time is output. This method converts the fault energy accumulation value into a quantifiable risk trajectory, providing sufficient disposal window period for malignant faults such as thermal runaway, effectively improving the advance and reliability of the battery system safety warning. BRIEF DESCRIPTION OF DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, brief descriptions will be given below for the drawings needed to be used in the embodiments or prior art descriptions. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of the drawings shown.

[0062] Figure 1 A flowchart of a battery fault warning and diagnosis method provided by an embodiment of the present application;

[0063] Figure 2 A principle block diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be described below in a clear and complete manner with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0065] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, motion condition, etc. between the components in a certain posture, and if the certain posture changes, the directional indications will also change accordingly.

[0066] In addition, if the embodiments of the present application involve descriptions such as "first", "second", etc., the descriptions of "first", "second", etc. are only for description purposes and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first" and "second" can explicitly or implicitly include at least one of the features. In addition, "and / or" or "and / or" appearing throughout the text means that the three parallel schemes are included, for example, "A and / or B" includes A scheme, or B scheme, or A and B simultaneously satisfy the scheme. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of the ordinary skilled in the art, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor within the scope of protection required by the present application.

[0067] Referring to Figure 1 One embodiment of the present application provides a battery fault early warning and diagnosis method, comprising the following steps:

[0068] S10, real-time acquisition of the operating parameters of the battery pack, branch processing of the operating parameters based on a preset environmental noise intensity determination condition;

[0069] If it is determined that it is a strong noise environment, noise stripping processing is performed on the operating parameters to generate a preprocessed signal; if it is determined that it is a non-strong noise environment, the operating parameters are taken as the preprocessed signal;

[0070] S20, extracting the spatiotemporal features of the preprocessed signal, the extracted spatiotemporal features including signal energy features of a preset target frequency band, battery pack monomer temperature difference distribution features and monomer voltage distribution features; performing weak signal enhancement processing on the extracted spatiotemporal features to generate an enhanced feature set;

[0071] S30, based on the spatiotemporal features in the enhanced feature set, performing spatiotemporal fusion calculation according to a preset weight relationship to obtain a fault energy accumulation value;

[0072] S40, dynamically correcting the fault determination threshold value according to the battery health state and real-time environmental parameters;

[0073] S50, when the fault energy accumulation value exceeds the dynamically corrected fault determination threshold value, outputting a graded early warning signal.

[0074] In this embodiment, the interference signal is stripped by the adaptive branch processing of environmental noise, and the weak fault features of the battery pack are accurately captured under strong noise. The spatial feature extraction and signal enhancement technology of temperature difference and voltage distribution are combined to improve the detection sensitivity of early hidden faults. The time and space fusion calculation of fault energy accumulation value is used to quantify the fault deterioration trend and predict the risk trajectory. Finally, based on the dynamic correction threshold of battery health state and environmental parameters, the accurate output of graded warning signals is realized, which fundamentally solves the technical difficulties of high misdiagnosis rate, high missed diagnosis rate and response lag of traditional methods, and shortens the fault detection time and reduces the false alarm rate in the lithium battery safety warning scene.

[0075] As described in step S10 above, the running parameters of the battery pack are collected in real time by the battery management system (BMS), including the internal state parameters such as the current signal, single voltage and temperature of the battery pack; based on the preset environmental noise intensity judgment condition (such as voltage fluctuation standard deviation > 50mV or temperature sampling noise > 2℃), the running parameters are processed by branch: if it is determined that it is a strong noise environment (such as current impact at vehicle start or sensor noise in low temperature environment), the wavelet transform threshold denoising algorithm is used to perform noise stripping processing on the running parameters, and the effective signal features are retained to generate preprocessed signals; if it is determined that it is not a strong noise environment (such as normal working condition in smooth driving), the running parameters are directly used as preprocessed signals. Through the noise branch processing mechanism, the signal distortion caused by strong noise environment is avoided, and the data reliability is improved.

[0076] As described in step S20 above, the time and space features of the preprocessed signals are extracted, including:

[0077] Time domain feature: calculate the signal energy feature of the preset target frequency band (such as 5-100Hz) (avoid <5Hz low frequency noise), capture the weak electric signal fluctuation accompanied by the internal chemical reaction of the battery;

[0078] Spatial feature: extract the single cell temperature difference distribution feature (such as the area ratio of single cell temperature difference > 5℃) and the single cell voltage distribution feature (such as the number of single cells with voltage difference > 100mV).

[0079] The method of empirical mode decomposition (EMD) combined with adaptive threshold is used for weak signal enhancement processing of time and space features, and the background noise interference is suppressed to generate an enhanced feature set containing energy features, temperature difference features and voltage features.

[0080] Through time and space feature fusion and weak signal enhancement, the hidden features of early battery faults (such as voltage micro-fluctuation and local overheating temperature difference caused by micro-short circuit) are converted into identifiable dominant features.

[0081] As described in step S30 above, based on the spatio-temporal features in the enhanced feature set, spatio-temporal fusion calculation is performed according to the preset weight relationship (such as time energy feature weight 40%, temperature difference distribution feature weight 35%, voltage distribution feature weight 25%). The weight relationship is dynamically adjusted according to the battery health state (such as increasing the temperature difference feature weight to 45% for an aging battery), and the fault energy accumulation value is obtained by weighted summation, which quantifies the energy release degree of the internal fault of the battery.

[0082] Through dynamic weight spatio-temporal fusion, the one-sidedness of single feature diagnosis is solved, and multi-dimensional fault feature collaborative representation is realized.

[0083] As described in step S40 above, the fault judgment threshold is dynamically corrected according to the battery health state (SOH) and real-time environmental parameters (such as environmental temperature -20℃, charge-discharge rate 2C). When SOH<80%, the initial threshold is lowered by 15% to improve the fault sensitivity of the aging battery; when the environmental temperature is <0℃, the temperature-related threshold is corrected in combination with the Arrhenius equation to compensate for the influence of low temperature on the performance of the battery.

[0084] This breaks the limitations of traditional fixed threshold, makes the fault judgment standard real-time adapt to the actual health state of the battery and working conditions, and reduces the missed diagnosis rate.

[0085] As described in step S50 above, when the fault energy accumulation value exceeds the dynamically corrected fault judgment threshold, a three-level graded early warning mechanism is triggered:

[0086] First-level warning (energy accumulation value>1.2×threshold): through the vehicle-mounted battery management system (BMS), the fault prompt is pushed to the instrument panel, and the upper limit of the charge-discharge current is adjusted;

[0087] Second-level warning (energy accumulation value>1.5×threshold): the battery thermal management system is started to start liquid cooling circulation, and the charge-discharge power is limited;

[0088] Third-level warning (energy accumulation value>2.0×threshold): based on the voltage distribution feature, the abnormal single cell is located (such as the 37th single cell voltage deviating from the mean value by 25%), and the main loop is simultaneously triggered.

[0089] Through the graded early warning, the quantitative management of fault risk is realized, different disposal strategies are provided for faults of different severity, and the system safety is improved.

[0090] In an embodiment, it is assumed that a certain brand of electric vehicle is equipped with NCM811 power battery pack (200 series-parallel structure), and when driving in a cold region at -10℃, the battery management system (BMS) performs fault early warning through the following process:

[0091] Step S10: Real-time acquisition of battery pack operating parameters, discovery of voltage fluctuation standard deviation reaching 80mV (strong noise threshold exceeding 50mV), determination of low-temperature strong noise environment, and use of wavelet transform to strip noise from voltage signal and retain true voltage change trend.

[0092] Step S20: Extraction of signal energy features of target frequency band (20-80Hz), discovery of energy value being 30% higher than normal working condition; and detection of cell temperature difference distribution, with cell Nos. 35-40 having temperature difference reaching 7°C (threshold value exceeding 5°C), and enhancement of weak temperature difference features through EMD algorithm.

[0093] Step S30: Calculation of fault energy cumulative value according to preset weight (time domain energy 40%, temperature difference 35%, and voltage 25%), with current value being 0.78 (initial threshold value).

[0094] Step S40: Dynamic correction of threshold value to 0.9 (initial threshold value is lowered by 10%).

[0095] Step S50: When vehicle climbing causes charge-discharge rate to rise to 2.5C, fault energy cumulative value quickly rises to 1.1x0.9 =0.99 , triggering level-1 early warning: BMS lowers charging current from 150A to 100A, and displays "battery local overheating risk, suggest reducing load" on instrument panel.

[0096] Through the whole-process fault diagnosis mechanism, early warning is triggered in advance when the battery appears early local overheating, the response time is shortened compared with the traditional fixed threshold scheme, the fault missed detection caused by noise interference in low-temperature environment is avoided, and the battery safety of electric vehicles in extreme working conditions is ensured.

[0097] In one of the embodiments, in step S10, the operating parameters of the battery pack are acquired in real time, and the operating parameters are processed based on a preset environmental noise intensity determination condition, specifically including:

[0098] S11, performing primary determination, based on historical noise intensity records of the environment where the battery pack is located, if the average noise intensity in the past continuous preset time length is lower than a preset safe noise threshold value, then it is directly determined as a non-strong noise environment;

[0099] S12, otherwise, performing secondary determination, performing frequency spectrum analysis on the currently acquired operating parameters, extracting noise component energy higher than a preset demarcation frequency, and if the noise component energy exceeds a dynamic noise threshold value, then it is determined as a strong noise environment;

[0100] ​wherein, only when the secondary determination output is a strong noise environment, a noise stripping process is performed to generate the preprocessed signal.

[0101] In the present embodiment, by adopting a two-stage determination mechanism, a preliminary decision on the environmental noise state is first made through a time history noise intensity evaluation to avoid misjudgment of transient interference; secondly, a high-frequency strong noise component is accurately captured through a frequency-division scanning noise determination of operating parameters; finally, noise stripping processing is started only when the secondary verification confirms a strong noise environment, thereby realizing on-demand allocation of computing resources. The problems of power waste and interference residue caused by traditional single-stage determination are avoided at the root, thereby providing lightweight real-time decision support for fault diagnosis of power batteries in high-noise scenarios.

[0102] As described in step S11 above, based on a battery pack historical operation database, environmental noise intensity records (such as voltage fluctuation standard deviation, temperature sampling noise) in the past continuous preset time length (such as 1 hour) are real-time retrieved, and the average noise intensity is calculated. If the average value is lower than the preset safe noise threshold (such as voltage fluctuation standard deviation < 50 mV), it is directly determined as a non-strong noise environment, and the secondary determination process is skipped.

[0103] Through statistical analysis of historical noise, rapid screening is realized, redundant calculation on low-noise environments is avoided, the time consumption of single noise determination is shortened, and the real-time performance of the system is improved.

[0104] As described in step S12 above, when the primary determination fails (such as average noise intensity ≥ 50 mV), a dynamic noise threshold needs to be modified to accurately identify a strong noise environment.

[0105] The significance of dynamic noise threshold modification lies in:

[0106] The noise energy (high-frequency noise component) of the battery is affected by two core factors:

[0107] Charge-discharge rate (C): when high-rate charging and discharging, the current density in the battery changes sharply, which will cause electrode polarization and intensified side reactions, and the noise energy will soar.

[0108] Ambient temperature: in low-temperature environment, the viscosity of electrolyte increases, and the ion migration speed slows down, which makes the electrode interface prone to "local polarization", and the noise energy significantly increases.

[0109] If a fixed threshold (such as ) is used to determine a strong noise environment, there may be:

[0110] When high-rate / low-temperature, the actual noise energy is much higher than → it will be misjudged as "strong noise", and noise stripping will be triggered excessively (resulting in loss of useful signals). When low-rate / high-temperature, the actual noise energy is much lower than → Will miss "strong noise" (cause false alarm of fault diagnosis).

[0111] Therefore, the fast Fourier transform (FFT) spectrum analysis is performed on the current operating parameters (voltage, temperature signal), the high-frequency noise component energy higher than the preset demarcation frequency (such as 100 Hz) is extracted, the dynamic noise threshold is self-adaptively adjusted according to the real-time working condition parameters, and the noise threshold is dynamically corrected through the following formula:

[0112] Dynamic noise threshold = basic threshold × charge-discharge rate correction term × ambient temperature correction term.

[0113] Specifically:

[0114]

[0115] If the high-frequency noise component energy exceeds the dynamic threshold, it is determined that it is a strong noise environment, and the wavelet transform noise stripping processing is triggered.

[0116] Through spectrum analysis, mechanical vibration noise (low frequency) and electrochemical reaction noise (high frequency) are distinguished, and the threshold is automatically adjusted downward at-10℃ low temperature environment, avoiding the missed judgment of pulse noise at cold start.

[0117] In an available embodiment, it is assumed that a certain brand of electric vehicle is equipped with NCM811 power battery pack (200 string-parallel structure), and when driving in a cold region at-10℃:

[0118] Primary judgment stage:

[0119] Retrieve the noise record of the battery pack in the past 1 hour, find that the voltage fluctuation standard deviation during low-temperature start-up stage reaches 75mV, but decreases to 40mV during smooth driving, the overall average noise intensity is 55mV (≥50mV safety threshold), the primary judgment fails, and the secondary judgment is entered.

[0120] Secondary judgment stage:

[0121] Perform FFT analysis on the current operating parameters, extract the high-frequency noise component energy above 100Hz as 36m At this time, the charge-discharge rate is 2.0C, the ambient temperature is-10℃, and the dynamic noise threshold is calculated:

[0122] Since 36m > , it is determined that it is a strong noise environment, and the noise stripping processing is triggered, and the wavelet transform threshold denoising algorithm is used to perform noise stripping processing on the operating parameters.

[0123] In one of the embodiments, the weak signal enhancement processing in step S20 includes:

[0124] a) High-frequency signal enhancement path:

[0125] The signal energy characteristics of the preset target frequency band are frequency domain decomposed, and a frequency band component with a preset bandwidth and a center frequency point determined by the real-time state of the battery is extracted;

[0126] Based on the ratio of the signal amplitude of the frequency band component to the ambient noise amplitude, a dynamic gain value is calculated and applied to the frequency band component to obtain a high-frequency enhancement component;

[0127] b) Low-frequency trend retention path:

[0128] The battery pack cell temperature difference distribution characteristics and cell voltage distribution characteristics are subjected to a moving average filter, wherein the filter window length is inversely proportional to the battery charge-discharge rate, to obtain a low-frequency enhancement feature;

[0129] c) Cross-scale fusion:

[0130] The high-frequency enhancement component and the low-frequency enhancement feature are weighted and superimposed according to a preset fusion coefficient to generate the enhancement feature set.

[0131] In this embodiment, by adopting a cross-scale dual-path enhancement mechanism, the characteristic frequency band is dynamically locked based on the real-time state of the battery, and the high-frequency fault signal is amplified in a targeted manner; combined with the dynamic adaptive filtering of the low-frequency feature sliding window, the slow decay trend characteristics of the battery are completely retained; finally, the enhancement feature set is output through weighted fusion, which breaks through the noise interference limit while maintaining the distinguishability of the full-scale fault mode.

[0132] Specifically:

[0133] a) High-frequency signal enhancement path (solves the problem of environmental noise masking high-frequency fault signals).

[0134] The time domain signal is converted into frequency domain energy distribution through fast Fourier transform (FFT), focusing on the preset target frequency band (such as 10-100Hz) that is strongly related to battery faults, which usually contains fault characteristic signals such as electrode contact failure and electromagnetic noise generated by side reactions. The high-frequency fault characteristics in the mixed signal are separated from the time domain environmental noise, avoiding the interference of low-frequency noise such as environmental vibration and motor electromagnetic interference.

[0135] According to the real-time state parameters of the battery (such as SOC, SOH, temperature), through the electrochemical model (such as PNGV equivalent circuit model, porous electrode theory model), combined with the real-time state parameters of the battery (SOC, SOH, temperature), the characteristic frequency is dynamically calculated. For example, based on the PNGV model, the relationship between electrode polarization impedance and frequency is derived, and when the battery ages (SOH decreases) and causes the SEI film to thicken, the characteristic frequency of the polarization impedance will drift from 50Hz to 30Hz. Through the model, the fault sensitive frequency under the current working condition is solved in real time, ensuring the accuracy of the frequency band component extraction. Alternatively, a machine learning algorithm (such as random forest, long short-term memory network LSTM) is used, with historical battery data (fault characteristic frequency under different SOC, SOH, temperature) as the training set, to build a "battery state-characteristic frequency" mapping model. Real-time input of current battery state parameters (such as SOC=85%, SOH=78%, temperature 25°C), through the pre-trained model, the characteristic frequency under the corresponding working condition (such as 42Hz) is output. This algorithm automatically learns the correlation between battery state changes and characteristic frequency drift, without the need for manual parameter adjustment. Taking this frequency as the center, the frequency band component of a predetermined bandwidth (such as ±5Hz) is intercepted, and the fault sensitive signal is accurately locked. Breakthrough the limitations of traditional fixed frequency analysis, dynamically adapt to the characteristic frequency shift caused by battery state changes, avoid missing fault signals. Unlike traditional methods that fix the characteristic frequency (which is prone to missing faults due to battery state changes), this scheme dynamically calculates the fault sensitive frequency under the current working condition through an electrochemical model (such as PNGV) or a machine learning algorithm (such as LSTM). For example, when the battery SOH decreases from 90% to 70%, the characteristic frequency drifts from 50Hz to 35Hz, and the model can capture this change in real time, ensuring that the frequency band component extraction covers the real fault signal.

[0136] Real-time calculation of the ratio of fault signal amplitude to environmental noise amplitude in the target frequency band, if the ratio > 1 (such as 3mV fault signal vs 2mV noise), then the signal is amplified by the ratio (such as gain = 10 x 1.5 = 15 times); if the ratio < 1, then the signal is maintained or suppressed. Through dynamic gain adjustment, the real fault signal is directionally enhanced. Avoid the problem of synchronous amplification of noise caused by fixed gain, and still highlight the weak fault signal in strong noise environment (such as when the motor starts).

[0137] b) Low-frequency trend maintenance path (solves the problem of low-frequency fault signal being masked by working condition fluctuations).

[0138] For low-frequency slowly changing signals such as single cell temperature difference (normal temperature difference <0.5℃) and single cell voltage difference (normal voltage difference <50mV), a sliding average filtering algorithm is used to smooth the data. This algorithm calculates the average of N consecutive sampling points to suppress transient fluctuations (such as voltage spikes during fast charging) during charging and discharging. It retains the slow changing trend of temperature difference and voltage (such as a 0.3℃ temperature difference increase lasting 10 minutes), and filters high-frequency noise interference under normal working conditions.

[0139] The filter window length is dynamically adjusted according to the charging and discharging rate (formula: , 60 seconds as the reference length). For example, the window length is shortened to 20 seconds at 3C high-rate charging to quickly respond to transient fluctuations, and lengthened to 120 seconds at 0.5C low-rate to avoid excessive smoothing of the true trend. This solves the shortcomings of traditional fixed window filtering, such as trend lagging at high rates and noise remaining at low rates, and dynamically adapts the filtering effect to the working conditions.

[0140] c) Cross-scale fusion (solving the problem of high / low frequency signal segmentation).

[0141] According to the battery fault type (such as thermal runaway, overcharge, micro short circuit), the fusion weight is preset (such as high frequency 0.6, low frequency 0.4), or the weight is dynamically adjusted in real time according to the diagnosis demand. The enhanced feature set containing full-scale fault information is generated by weighting and merging the high-frequency enhanced fault features (such as 45mV high-frequency noise) and low-frequency trend signals (such as 0.8℃ temperature difference increase). This breaks through the one-sidedness of single frequency band analysis, integrates high-frequency early fault features and low-frequency fault development trend, and provides complete and accurate input data for subsequent fault energy calculation.

[0142] In one embodiment, in step S30, the preset weight relationship is a dynamic weight relationship, which is dynamically adjusted according to real-time working state parameters of the battery pack, wherein:

[0143] The real-time working state parameters include at least one of ambient temperature, battery charging and discharging rate, and battery state of charge;

[0144] The dynamic weight relationship includes time domain feature weight, first spatial domain feature weight, and second spatial domain feature weight, wherein the time domain feature is signal energy feature, the first spatial domain feature is battery pack single cell temperature difference distribution feature, and the second spatial domain feature is battery pack single cell voltage distribution feature; the dynamic adjustment of the dynamic weight relationship satisfies:

[0145] When the ambient temperature is lower than a temperature threshold, the weight of the first spatial domain feature is increased;

[0146] When the charging and discharging rate is higher than a rate threshold, the weight of the time domain feature is increased;

[0147] The second spatial feature weight remains within a preset threshold range.

[0148] In this embodiment, by introducing a dynamic weighting relationship, the weights of time-domain features, first spatial-domain features, and second spatial-domain features are dynamically adjusted based on the real-time operating state parameters of the battery pack (ambient temperature, battery charge / discharge rate, battery state of charge, etc.). This effectively solves the problem that traditional fixed-weight diagnostic methods cannot accurately match the importance of fault features under different operating conditions. In low-temperature environments, increasing the weight of the temperature difference distribution feature of individual battery cells can detect potential faults caused by uneven temperature in advance. Under high charge / discharge rate conditions, increasing the weight of the signal energy feature can accurately identify abnormal energy fluctuations under high load. This dynamic weighting mechanism breaks through the "one-size-fits-all" limitation of traditional threshold diagnostic methods, avoiding misdiagnosis and missed diagnosis due to environmental changes or differences in operating conditions. This makes the fault diagnosis results more consistent with the actual operating state of the battery, significantly improving the accuracy and reliability of fault diagnosis.

[0149] Specifically, real-time operating status parameters affect weight allocation in the following ways:

[0150] Ambient temperature: reflects the thermal environment in which the battery pack operates. In low-temperature environments, the rate of internal chemical reactions in the battery decreases, which can easily lead to localized temperature anomalies and affect battery life.

[0151] Charge / discharge rate: This reflects the magnitude of the battery's charge / discharge current. High charge / discharge rates will exacerbate internal polarization of the battery, generating more heat and voltage fluctuations.

[0152] Battery State of Charge (SOC): Indicates the current charge level of the battery. Different SOC ranges have different internal resistance and thermal characteristics, resulting in different fault manifestations.

[0153] For the adjustment strategy of dynamic weights:

[0154] (1) Time-domain feature weights (signal energy features).

[0155] Technical issue: During high-rate charging and discharging (such as fast charging), the internal reaction of the battery is intense, and the energy characteristics (time-domain characteristics) of voltage / temperature fluctuations can better reflect changes in battery state.

[0156] Adjustment logic:

[0157] When the charge / discharge rate exceeds the rate threshold (e.g., 2C), the weight of time-domain features is automatically increased. By increasing the proportion of signal energy features, the sensitivity to transient faults (e.g., lithium plating, poor contact) at high rates is enhanced.

[0158] Mathematical expression: Time-domain feature weights = base weights × (1 + ... × (charge / discharge rate - rate threshold) is a gain coefficient.

[0159] (2) First spatial feature weight (single cell temperature difference distribution feature).

[0160] Technical problem: In low temperature environments (such as <0℃), the temperature difference between each single cell in the battery pack intensifies, easily causing local overcooling or thermal runaway.

[0161] Adjustment logic:

[0162] When the ambient temperature is lower than the temperature threshold (such as 5℃), automatically increase the first spatial feature weight. By amplifying the influence of the single cell temperature difference distribution feature, focus on monitoring the temperature imbalance problem under low temperature.

[0163] Mathematical expression: First spatial feature weight = base weight × (1 + × (temperature threshold - ambient temperature)), is a gain coefficient.

[0164] (3) Second spatial feature weight (single cell voltage distribution feature).

[0165] Technical problem: Single cell voltage difference is a key indicator of battery pack consistency, but excessive attention may lead to neglect of other fault features.

[0166] Adjustment logic:

[0167] Limit the second spatial feature weight within a preset threshold range (such as 0.3-0.5) to ensure stable monitoring of battery pack consistency, while avoiding too high weight masking other fault signals.

[0168] Mathematical expression: Second spatial feature weight = min (max (base weight × f (SOC), lower limit value), upper limit value), f (SOC) is an adjustment function related to SOC.

[0169] Synergistic mechanism of dynamic weight:

[0170] Through real-time working state parameter dynamic adjustment of each feature weight, realize multi-dimensional accurate monitoring of battery faults:

[0171] Low temperature scenario: Increase the single cell temperature difference distribution feature weight to preferentially identify local performance degradation caused by low temperature.

[0172] High rate scenario: Increase the signal energy feature weight to focus on capturing transient abnormalities under high current.

[0173] Full working condition range: Stabilize the single cell voltage distribution feature weight to continuously monitor the battery pack consistency.

[0174] Compared with the traditional fixed weight method, the dynamic weight allocation mechanism can more flexibly adapt to the real-time working state of the battery pack, significantly improve the recognition of fault features under different working conditions, and provide more reliable feature input for the fault type determination in the subsequent step S40.

[0175] In one of the embodiments, in step S30, the calculation method of the fault energy accumulation value is:

[0176] The space-time features in the enhanced feature set are correspondingly mapped to the feature values and reference values of the fault energy calculation formula, and the fault energy accumulation value is calculated by the following formula:

[0177]

[0178] Among them, is the time Next, the battery pack target feature value after weak signal enhancement is selected from at least one of the signal energy feature, the battery pack single cell temperature difference distribution feature, and the battery pack single cell voltage distribution feature. is the normal working condition reference value corresponding to the target feature value; is the feature sampling interval; is the cumulative calculation time length;

[0179] Moreover, the selection of the target feature value and the assignment of the corresponding are associated with the dynamic weight and are adapted accordingly, specifically satisfying:

[0180] When the ambient temperature is lower than the temperature threshold, the battery pack single cell temperature difference distribution feature is preferentially selected as for calculation, which correspondingly increases the weight proportion of the single cell temperature difference normal reference value;

[0181] When the charge-discharge rate is higher than the rate threshold, the signal energy feature is preferentially selected as for calculation, which correspondingly increases the weight proportion of the signal energy normal reference value.

[0182] In this embodiment, by deeply integrating the enhanced feature set with the fault energy calculation formula, the weak abnormal features in the battery operation process are converted into quantifiable and accumulative fault energy values, solving the problem of early hidden fault features that are difficult to evaluate effectively in traditional threshold diagnosis methods. Combined with the dynamic weight relationship, the appropriate target features are selected under different working conditions for fault energy accumulation calculation, so that the fault energy accumulation value can truly reflect the fault development trend. Compared with the traditional diagnosis method, this calculation method realizes the leap from "static threshold judgment" to "dynamic energy accumulation analysis", can accurately capture feature changes at the early stage of failure, greatly shortens the fault diagnosis time, reduces the misdiagnosis rate and missed diagnosis rate caused by feature misjudgment, provides a more scientific and more accurate quantitative basis for early fault warning of lithium batteries, and effectively improves the safety and reliability of the battery management system.

[0183] Specifically, the degree of battery feature deviation from the normal state is quantified in the form of "energy accumulation". The greater the deviation and the longer the duration, the higher the fault energy, which intuitively reflects the severity and development trend of the fault.

[0184] The adaptation rule of the target feature and the reference value, the selection of the target feature and the assignment of the normal working condition reference value are linked with the dynamic weight relationship, and the specific adaptation logic is as follows:

[0185] (1) When the ambient temperature is lower than the threshold value (e.g. <5℃).

[0186] Under low temperature environment, the battery pack monomer temperature difference distribution feature is more sensitive to fault (such as local cold start caused by temperature difference anomaly). Therefore:

[0187] Prioritize selecting monomer temperature difference distribution features as target features (such as the maximum temperature difference between monomers);

[0188] Dynamic adjustment : Increase the weight proportion of "monomer temperature difference normal reference value" (such as relaxing the normal temperature difference upper limit from 0.5℃ to 0.8℃, compensating for natural temperature difference fluctuations under low temperature).

[0189] This avoids misjudging normal temperature difference fluctuations under low temperature environment as faults, while strengthening the identification of real temperature difference faults (such as coolant blockage).

[0190] (2) When the charge and discharge rate is higher than the threshold value (e.g. >2C).

[0191] Under high-rate charging and discharging, signal energy features (such as voltage fluctuation energy) can better reflect internal polarization, lithium precipitation and other faults in the battery. Therefore:

[0192] Prioritize selecting signal energy features as target features (signal energy value of target frequency band);

[0193] Dynamic adjustment : Increase the weight proportion of "signal energy normal baseline value" (such as increasing the normal energy upper limit from 30 to 50 , adapt to the normal energy fluctuation under high magnification).

[0194] In order to avoid misjudgment caused by normal energy fluctuation under high magnification, and accurately capture the fault caused by high magnification (such as energy mutation caused by diaphragm puncture).

[0195] Synergistic mechanism of energy accumulation calculation:

[0196] Through "dynamic selection of characteristics + adaptation of baseline value", the fault energy accumulation calculation is realized:

[0197] Working condition self-adaptation: focus on temperature difference in low temperature scene and energy in high magnification scene, solve the one-sidedness of traditional single feature calculation;

[0198] Baseline value dynamic compensation: adjust the normal baseline according to the working condition, avoid misjudgment caused by environmental / working condition fluctuation;

[0199] Energy quantization intuitiveness: use integral form to accumulate deviation, convert "instantaneous fault feature" into "traceable energy index", which is convenient for subsequent fault threshold judgment (step S40).

[0200] In one of the embodiments, the setting of the normal working condition baseline value further comprises:

[0201] Collect historical operation data of the battery pack for preprocessing to obtain standardized feature data applied to the machine model as input data;

[0202] The standardized feature data obtained after preprocessing is used to train the battery individual health state model by machine learning algorithm, and the trained battery individual health state model is obtained, which receives the operation parameters of the battery pack in real time and outputs the personalized normal working condition baseline interval dynamically adapted to the battery life cycle; wherein the personalized normal working condition baseline interval is dynamically updated with the number of battery cycles and the degree of aging;

[0203] When it is monitored that the health state dispersion of single batteries in the same batch of battery packs exceeds the preset threshold, the adaptive correction of the personalized normal working condition baseline interval is triggered, and the correction process is combined with real-time working state parameters for multidimensional dynamic adjustment.

[0204] In this embodiment, the individual health state model is trained by collecting battery pack historical operation data, and a personalized normal working condition reference interval dynamically adapted to the battery aging degree is generated, breaking through the bottleneck that the traditional fixed threshold is difficult to match the individual differences of the battery. The reference interval combines dynamic weight rules (such as preferentially selecting the single cell temperature difference feature at low temperature), realizing the cooperative correction of the reference value and the feature weight: when the dispersion degree of the single cell health state exceeds the threshold, the reference interval is adjusted in multiple dimensions based on the real-time working state parameters (environmental temperature, charge and discharge rate), ensuring that the temperature difference feature tolerance is improved at low temperature working condition, and the voltage abnormal sensitivity is enhanced at high rate working condition, thereby reducing the misjudgment rate.

[0205] Specifically, for the setting of the normal working condition reference value, it is realized by "historical data driving + model dynamic adaptation + dispersion trigger correction", and the specific logic is as follows:

[0206] (1) Historical data preprocessing: constructing standardized input.

[0207] Collect the full life cycle historical operation data of the battery pack (covering different working conditions, health states, voltage, temperature, charge and discharge rate, etc.), and preprocess by the following steps:

[0208] Abnormal value filtering: eliminate abnormal data caused by sensor failure, extreme working condition (such as short circuit protection);

[0209] Normalization processing: map different dimension features (such as voltage 0-5V, temperature -20-60℃) to the same scale (such as 0-1);

[0210] Time sequence feature construction: extract the feature change trend in the sliding window (such as the standard deviation of voltage fluctuation in 5 minutes).

[0211] In this way, the problems of "dimension heterogeneity, uneven distribution, and noise" in the original data are solved, and high-quality standardized feature data is generated, providing reliable input for machine learning model training.

[0212] (2) Machine learning training: dynamically adapted health state model.

[0213] The preprocessed standardized feature data is input into the machine learning algorithm (such as LSTM, random forest), and the battery individual health state model is trained:

[0214] Model input: real-time operation parameters (voltage, temperature, charge and discharge rate, cycle number, etc.);

[0215] Model output: personalized normal working condition reference interval (such as the normal voltage fluctuation range and temperature difference range of a single cell battery);

[0216] Dynamic updating mechanism: the model adjusts the reference interval in real time according to the number of battery cycles and the degree of aging (SOH decay). For example, after 1000 cycles, the upper limit of normal voltage fluctuation is relaxed from 20mV to 30mV.

[0217] Breakthrough traditional "fixed reference value" defects, make normal working condition reference adapt to battery aging and cycle number, solve the problem of "misjudgment / omission caused by the same reference for new and old batteries".

[0218] (3) Dispersity trigger correction: multi-dimensional dynamic adjustment.

[0219] Real-time calculation of the health state dispersity of single cells in the same batch of battery pack (such as the standard deviation of each single cell SOH), when the dispersity exceeds the preset threshold (such as SOH standard deviation > 10%):

[0220] Trigger condition: determine the deterioration of battery pack consistency (such as premature aging of some single cells);

[0221] Correction mechanism: combined with real-time working state parameters (environmental temperature, charge-discharge rate), multi-dimensional adjustment of individual normal working condition reference interval:

[0222] In low temperature environment, narrow single cell temperature difference reference interval (strengthen the identification of temperature difference abnormality);

[0223] When high-rate charging and discharging, expand the voltage fluctuation reference interval (tolerate normal fluctuation under high rate).

[0224] Solve the problem of "battery pack consistency deterioration, individual reference failure", trigger adaptive correction through dispersity monitoring, make the reference interval adapt to individual aging, and compatible with batch difference and real-time working condition.

[0225] Through the whole process of "historical data preprocessing → machine learning modeling → dispersity trigger correction", a dynamic, individualized and adaptive normal working condition reference system is established:

[0226] Historical data driven to ensure that the reference fits the characteristics of the whole life cycle of the battery;

[0227] Machine learning model to realize dynamic updating of reference with aging and cycle number;

[0228] Dispersity trigger correction solves the problem of reference failure after batch consistency deterioration.

[0229] This scheme provides accurate and dynamic normal working condition reference for fault energy accumulation calculation (assignment) of , so that fault diagnosis not only adapts to individual differences of batteries, but also compatible with complex working condition fluctuations, fundamentally improving the accuracy and robustness of fault warning.

[0230] In one of the embodiments, after step S30, the method further comprises the following steps:

[0231] Obtaining historical failure data and real-time working state parameters, combining the calculation result of the failure energy accumulation value, performing data standardization preprocessing, and generating a multivariate time series feature dataset;

[0232] Based on the standardized multivariate time series feature dataset, a multivariate time series prediction model is constructed using a machine learning algorithm, and the multivariate time series prediction model is used to predict the failure energy growth rate in the future period;

[0233] The predicted failure energy growth rate is compared with the dynamically corrected failure judgment threshold change gradient in real time, and when the predicted failure energy growth rate exceeds the product of the preset safety factor and the failure judgment threshold change gradient, an advanced warning mechanism is triggered;

[0234] According to the relative deviation degree of the predicted failure energy growth rate and the failure judgment threshold change gradient, a multi-level warning level is divided, and the remaining safe operation time is calculated based on the current failure energy accumulation value, the dynamically corrected failure judgment threshold and the predicted failure energy growth rate.

[0235] In this embodiment, based on the failure energy accumulation value (integrating dynamically selected feature values and weights), a multivariate time series prediction model is constructed to realize the leap from real-time diagnosis to trend prediction: by comparing the predicted failure energy growth rate with the gradient change of the dynamically corrected threshold in real time, when the predicted value exceeds the safety threshold, a hierarchical advanced warning is triggered. At the same time, combined with the individualized reference interval correction mechanism, the accurate remaining safe operation time is output. This method converts the failure energy accumulation value into a quantifiable risk trajectory, providing sufficient disposal window period for malignant failures such as thermal runaway, effectively improving the advancement and reliability of the battery system safety warning.

[0236] Specifically, after step S30, the advanced warning mechanism is added by "multivariate time series analysis + dynamic threshold comparison + remaining life prediction", and the specific logic is as follows:

[0237] (1) Multivariate time series feature construction.

[0238] Data collection: integrating historical failure data (such as overcharge, thermal runaway cases), real-time working state parameters (temperature, charge-discharge rate, SOC) and the failure energy accumulation value calculated in step S30;

[0239] Standardization preprocessing: mapping heterogeneous data (such as temperature -20~60℃, failure energy 0~1000J) to a unified scale (such as [-1,1]);

[0240] Time-series feature generation: Extract feature change trends (e.g., fault energy growth rate, temperature gradient) in sliding windows (e.g., 5 minutes) to construct a multivariate time-series dataset.

[0241] By converting "static fault indicators" into "dynamic time-series features", the time correlation of fault evolution is captured, providing structured input for machine learning prediction.

[0242] (2) Fault energy growth rate prediction.

[0243] Model construction: Use LSTM, Transformer, and other time-series prediction algorithms to train multivariate time-series prediction models based on standardized datasets.

[0244] Model output: Predict the fault energy growth rate in the future period (e.g., 30 minutes) .

[0245] Dynamic adaptation: Multivariate time-series prediction model parameters are updated in real time with battery aging (SOH decline) and working condition changes (temperature fluctuations), ensuring prediction accuracy.

[0246] By anticipating potential trends in fault energy, the traditional "threshold trigger type warning" lag problem is solved.

[0247] (3) Dynamic threshold comparison and early warning trigger

[0248] Dynamic threshold gradient: According to the dynamic weight relationship, real-time calculation of the change gradient of the fault judgment threshold (such as the threshold decline rate accelerating at low temperatures);

[0249] Safety factor product: Preset safety factor (such as 1.2), define the safety boundary as · .

[0250] Early warning trigger condition: When the predicted fault energy growth rate > · , trigger early warning.

[0251] Through the dynamic comparison of "predicted growth rate vs. threshold change gradient", early warning is given before the fault energy approaches the threshold, leaving a disposal time window.

[0252] (4) Multi-level warning level division and remaining life calculation.

[0253] i) Warning level division:

[0254]

[0255] ii) Residual safe operation time calculation:

[0256]

[0257] wherein, is the current fault determination threshold, is the current fault energy accumulation value.

[0258] The "risk degree visualization" is realized through the hierarchical early warning, and a differentiated response strategy is supported. The residual life prediction quantifies the safety boundary, and provides an accurate time reference for maintenance decision.

[0259] Through the whole chain of "time sequence feature extraction → growth rate prediction → dynamic threshold comparison → risk quantification", an advanced, hierarchical and quantifiable early warning system is constructed.

[0260] The multivariate time sequence model captures the fault evolution law, and breaks through the limitations of traditional single variable early warning;

[0261] The dynamic threshold comparison mechanism adapts to complex working conditions and battery aging, and avoids false positives / false negatives;

[0262] The residual life calculation provides a time window for preventive maintenance, and reduces the safety risk.

[0263] The scheme cooperates with the aforementioned dynamic weight relationship and fault energy accumulation calculation depth, forms a complete diagnosis link from "feature enhancement → fault quantification → risk prediction", and significantly improves the safety and reliability of the battery system.

[0264] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0265] In one of the embodiments, a battery fault early warning and diagnosis system is provided, which corresponds to the battery fault early warning and diagnosis method in the above embodiment. The battery fault early warning and diagnosis system comprises:

[0266] A data acquisition module is configured to acquire the operating parameters of the battery pack in real time, branch process the operating parameters based on a preset environmental noise intensity determination condition, and generate a preprocessed signal;

[0267] A noise stripping module is configured to determine when it is a strong noise environment, and execute noise stripping processing on the operating parameters to generate a preprocessed signal;

[0268] a feature extraction module configured to extract spatio-temporal features from the preprocessed signal, the extracted spatio-temporal features including signal energy features of a preset target frequency band, cell group monomer temperature difference distribution features, and monomer voltage distribution features; and perform weak signal enhancement processing on the extracted spatio-temporal features to generate an enhanced feature set;

[0269] a fusion calculation module configured to perform spatio-temporal fusion calculation on the spatio-temporal features in the enhanced feature set according to a preset weight relationship to obtain a fault energy accumulation value;

[0270] a dynamic correction module configured to dynamically correct a fault determination threshold according to a battery health state and real-time environmental parameters;

[0271] a pre-warning output module configured to output a graded pre-warning signal when the fault energy accumulation value exceeds the dynamically corrected fault determination threshold.

[0272] The specific limitations of the battery fault pre-warning and diagnosis system can be seen from the limitations of the battery fault pre-warning and diagnosis method described above, and will not be repeated here. Each module in the battery fault pre-warning and diagnosis system described above can be realized by software, hardware, or a combination thereof, in whole or in part. Each module described above can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.

[0273] In one embodiment, a computer device is provided, which can be a server, and its internal structure diagram can be as shown in Figure 2 The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store data, perform data processing, and perform data analysis. The network interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a battery fault pre-warning and diagnosis method.

[0274] In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor implements a battery fault pre-warning and diagnosis method when executing the computer program.

[0275] In one of the embodiments, a computer readable storage medium is provided, and a computer program is stored on the computer readable storage medium, and the computer program is executed by a processor to implement a battery fault early warning and diagnosis method.

[0276] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the processes of the above-mentioned embodiments can be included. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0277] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0278] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A battery failure early warning and diagnosis method, characterized in that, The method comprises the following steps: Real-time acquisition of the operating parameters of the battery pack, branch processing of the operating parameters based on a preset environmental noise intensity determination condition; wherein, if it is determined that it is a strong noise environment, noise stripping processing is performed on the operating parameters to generate a preprocessed signal; if it is determined that it is not a strong noise environment, the operating parameters are taken as the preprocessed signal; Extracting the spatio-temporal features of the preprocessed signal, the extracted spatio-temporal features including signal energy features of a preset target frequency band, battery pack cell temperature difference distribution features and cell voltage distribution features; performing weak signal enhancement processing on the extracted spatio-temporal features to generate an enhanced feature set; the weak signal enhancement processing including: a) High-frequency signal enhancement path: Frequency domain decomposition is performed on the signal energy features of the preset target frequency band to extract a frequency band component with a preset bandwidth and a center frequency determined by the real-time state of the battery; Based on the ratio of the signal amplitude of the frequency band component to the environmental noise amplitude, a dynamic gain value is calculated and applied to the frequency band component to obtain a high-frequency enhanced component; b) Low-frequency trend maintaining path: Sliding average filtering is performed on the battery pack cell temperature difference distribution features and the cell voltage distribution features, wherein the filtering window length is inversely proportional to the battery charge-discharge rate to obtain a low-frequency enhanced feature; c) Cross-scale fusion: The high-frequency enhanced component and the low-frequency enhanced feature are weighted and superimposed according to a preset fusion coefficient to generate the enhanced feature set; Based on the spatio-temporal features in the enhanced feature set, spatio-temporal fusion calculation is performed according to a preset weight relationship to obtain a fault energy accumulation value; According to the battery health state and real-time environmental parameters, the fault determination threshold is dynamically corrected; When the fault energy accumulation value exceeds the dynamically corrected fault determination threshold, a graded early warning signal is output.

2. The battery failure early warning and diagnosis method of claim 1, wherein, In the step of real-time acquisition of the operating parameters of the battery pack, branch processing of the operating parameters based on a preset environmental noise intensity determination condition, specifically comprising: Performing primary determination based on the historical noise intensity record of the environment where the battery pack is located, if the average noise intensity in the past continuous preset time length is lower than the preset safety noise threshold, it is directly determined as a non-strong noise environment; Otherwise, performing secondary determination, performing frequency spectrum analysis on the currently acquired operating parameters, extracting the noise component energy higher than the preset demarcation frequency, if the noise component energy exceeds the dynamic noise threshold, it is determined as a strong noise environment; Only when the secondary determination output is a strong noise environment, noise stripping processing is performed to generate the preprocessed signal.

3. The battery failure early warning and diagnosis method of claim 1, wherein, In the step of based on the spatio-temporal features in the enhanced feature set, spatio-temporal fusion calculation is performed according to a preset weight relationship to obtain a fault energy accumulation value, the preset weight relationship is a dynamic weight relationship, which is dynamically adjusted according to the real-time working state parameters of the battery pack, wherein: The real-time working state parameters include at least one of environmental temperature, battery charge-discharge rate and battery state of charge; The dynamic weight relationship includes a time domain feature weight, a first space domain feature weight and a second space domain feature weight, wherein the time domain feature is a signal energy feature, the first space domain feature is a battery pack monomer temperature difference distribution feature, and the second space domain feature is a battery pack monomer voltage distribution feature; dynamic adjustment of the dynamic weight relationship satisfies: When the ambient temperature is lower than a temperature threshold, the weight of the first space domain feature is increased; When the charge-discharge rate is higher than a rate threshold, the weight of the time domain feature is increased; The second space domain feature weight is kept within a preset threshold range.

4. The battery failure early warning and diagnosis method of claim 3, wherein, In the step of performing spatio-temporal fusion calculation according to a preset weight relationship based on the spatio-temporal features in the enhanced feature set to obtain a fault energy accumulation value, the calculation method of the fault energy accumulation value is: The spatio-temporal features in the enhanced feature set are correspondingly mapped to characteristic values and reference values of a fault energy calculation formula, and the fault energy accumulation value is calculated through the following formula: wherein, is the time Next, the battery pack target characteristic value after weak signal enhancement is selected from at least one of a signal energy characteristic, a battery pack single cell temperature difference distribution characteristic, and a battery pack single cell voltage distribution characteristic; is a normal working condition reference value corresponding to the target characteristic value; is a characteristic sampling interval; is a cumulative calculation duration; And, the selection of the target feature value and the assignment of the corresponding value are adaptively fitted with the dynamic weight relationship, specifically satisfying: When the ambient temperature is lower than the temperature threshold value, the battery pack cell temperature difference distribution characteristics are preferentially selected as The weight proportion corresponding to the cell temperature difference normal baseline value is calculated. When the charge-discharge rate is higher than the rate threshold value, the signal energy feature is preferentially selected as The weight proportion corresponding to the normal reference value of the signal energy is calculated.

5. The battery failure early warning and diagnosis method of claim 4, wherein, The setting of the normal working condition reference value further includes: Collecting battery pack historical operation data for preprocessing to obtain standardized feature data applied to a machine model as input data; Training a battery individual health state model by using a machine learning algorithm on the standardized feature data obtained after preprocessing to obtain a trained battery individual health state model, the battery individual health state model receiving operation parameters of the battery pack in real time and outputting a personalized normal working condition reference interval dynamically adapted to a battery life cycle; wherein the personalized normal working condition reference interval is dynamically updated with the number of battery cycles and the aging degree; When it is monitored that the health state dispersion degree of monomer batteries in the same batch of battery packs exceeds a preset threshold, the adaptive correction of the personalized normal working condition reference interval is triggered, and the correction process is combined with real-time working state parameters for multidimensional dynamic adjustment.

6. The battery failure early warning and diagnosis method of claim 4, wherein, After the step of performing spatio-temporal fusion calculation according to a preset weight relationship based on the spatio-temporal features in the enhanced feature set to obtain a fault energy accumulation value, the following steps are further included: Obtaining historical fault data and real-time working state parameters, combining the calculation result of the fault energy accumulation value, and performing data standardization preprocessing to generate a multivariate time series feature data set; Based on the standardized multivariate time series feature data set, a multivariate time series prediction model is constructed by using a machine learning algorithm, and the multivariate time series prediction model is used to predict the fault energy growth rate in a future period; The predicted fault energy growth rate is compared with the dynamically corrected fault judgment threshold change gradient in real time, and when the predicted fault energy growth rate exceeds the product of a preset safety factor and the fault judgment threshold change gradient, an advanced warning mechanism is triggered; According to the relative deviation degree of the predicted fault energy growth rate and the fault judgment threshold change gradient, a multistage warning level is divided, and the remaining safe operation time is calculated based on the current fault energy accumulation value, the dynamically corrected fault judgment threshold and the predicted fault energy growth rate.

7. A battery failure warning and diagnosis system for implementing the steps of a battery failure warning and diagnosis method according to any one of claims 1 to 6, characterized in that, It includes: A data acquisition module is configured to acquire operation parameters of a battery pack in real time, branch process the operation parameters based on a preset environmental noise intensity judgment condition, and generate preprocessed signals. A noise stripping module is configured to determine whether the running parameter is in a strong noise environment, and perform noise stripping processing on the running parameter to generate a preprocessed signal; A feature extraction module is configured to extract spatial-temporal features from the preprocessed signal, and the extracted spatial-temporal features include signal energy features of a preset target frequency band, cell group monomer temperature difference distribution features, and monomer voltage distribution features; The extracted spatial-temporal features are subjected to weak signal enhancement processing to generate an enhanced feature set; A fusion calculation module is configured to perform spatial-temporal fusion calculation on the spatial-temporal features in the enhanced feature set according to a preset weight relationship to obtain a fault energy accumulation value; A dynamic correction module is configured to dynamically correct a fault determination threshold according to a battery health state and real-time environmental parameters; An early warning output module is configured to output a graded early warning signal when the fault energy accumulation value exceeds the dynamically corrected fault determination threshold.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the steps of the battery fault early warning and diagnosis method according to any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the steps of the battery fault early warning and diagnosis method according to any one of claims 1-6.

Images