Lithium battery voltage out-of-control prevention method and system based on data scene self-adaption

Abstract

The invention provides a lithium battery voltage out-of-control prevention method and system based on data scene self-adaption, and belongs to the technical field of lithium battery voltage protection. Comprising the following steps: collecting a plurality of parameter tuples related to self-discharge of a battery cell, and selecting a key judgment parameter set from the parameter tuples; carrying out full-process testing on the abnormal calibration battery cells and the normal battery cells in a mixed manner, and generating a calibration data set through process data acquisition; constructing an initial judgment model according to a data scene of the calibration data set; a monitoring model based on a time sequence function is adopted to monitor the parameter fluctuation condition during self-discharging of the battery cell in real time, and when the parameter fluctuation reaches the alarm condition of the monitoring model, the initial judgment model is trained to obtain a judgment model for prediction of the out-of-control voltage of the lithium battery. According to the method, the risk of robustness collapse of a model system for lithium battery voltage out-of-control prediction caused by data trend progressive can be solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium battery voltage protection, and in particular relates to a method and system for preventing lithium battery voltage runaway based on data scenario adaptation. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Lithium-ion batteries are the core power source of electric vehicles, and their module voltage consistency is crucial to system safety. Voltage runaway (i.e., unbalanced cell voltages within a series-connected battery pack) can lead to localized overcharging / over-discharging of cells, causing serious problems such as irreversible capacity degradation, a sudden drop in cycle life, and deterioration of storage performance. Therefore, preventing voltage runaway in lithium-ion batteries is a highly worthy research topic.

[0004] However, existing technologies for preventing lithium battery voltage runaway generally have some technical problems, such as: (1) Existing methods mainly rely on single-dimensional data (such as static voltage threshold) or simple rule combinations to judge voltage out of control. Under this judgment method: on the one hand, misjudgment may be caused by production fluctuations (such as normal battery cells are mistakenly labeled as out of control); on the other hand, it has poor adaptability to complex working conditions and there is a possibility of over-judgment (such as misjudging temporary fluctuations as out of control) and missed judgment (such as delayed response to progressive micro-short circuits).

[0005] (2) Some existing methods also adopt a one-time modeling strategy and combine multi-dimensional data to judge voltage out of control. Under this judgment method: on the one hand, the mutual exclusion effect between multi-dimensional data may further reduce the reliability of the judgment, and when multi-dimensional parameters are introduced to improve the accuracy, there is a lack of effective feature screening mechanism, and redundant parameters will interfere with the stability of the model; on the other hand, the static model cannot adapt to the data drift in the manufacturing process (such as material batch differences, equipment aging), resulting in prediction distortion after long-term use.

[0006] (3) Existing methods also consider enhancing the stability of the discriminant model by introducing an adaptive adjustment mechanism. However, this adaptive method relies on manual experience to set the adjustment rules and cannot respond to time series fluctuations in the production environment (such as seasonal temperature and humidity changes) in real time. Moreover, it lacks a closed-loop verification mechanism, which may cause secondary deviations during the correction operation. Summary of the Invention

[0007] In order to overcome the above-mentioned shortcomings of the existing technology, the present invention provides a lithium battery voltage runaway prevention method and system based on data scenario adaptation, which can solve the risk of robustness collapse of the model system used for lithium battery voltage runaway prediction due to the accumulation of data trends.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a method for preventing lithium battery voltage runaway based on data scenario adaptation.

[0009] The lithium battery voltage runaway prevention method based on data scenario adaptation includes: Collect multiple parameter tuples related to cell self-discharge and select key judgment parameter sets from them; Make and calibrate abnormal cells and mix them with normal cells, collect process data corresponding to the key judgment parameter set, and generate a calibration data set; According to the data scenario of the obtained calibration data set, an initial discrimination model is constructed. Specifically: if the construction is for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if the construction is for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filter method to construct an initial discrimination model; A monitoring model based on time series function is used to monitor the parameter fluctuations of the battery cell during self-discharge in real time; When the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set within the alarm temporal state is extracted to generate a revised data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage out of control prediction.

[0010] Furthermore, calibrated abnormal cells are produced and mixed with normal cells, and process data corresponding to the key judgment parameter set is collected, including: mixing the calibrated abnormal cells with the normal cells for full process testing, applying physical marks to the calibrated abnormal cells, and making them undergo formation, aging, and capacity separation processes synchronously with the normal cells.

[0011] Furthermore, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discriminant model, including: first, dividing the calibration data set into abnormal subsets and normal subsets, and performing robust normalization on the abnormal subsets and normal subsets; then, performing nonlinear dimensionality reduction using an exponential kernel function and constructing a classification function; and using the constructed classification function as the initial discrimination model.

[0012] Furthermore, the normalized calibration data set is filtered based on the Kalman filtering method to construct an initial discrimination model, including: first, dividing the calibration data set into an abnormal subset and a normal subset, and performing Z-Score normalization on the abnormal subset and the normal subset; then, taking the normal subset as the main data and the abnormal subset as the noise data, and dividing them into multiple segments according to the time series; finally, constructing the state equation and covariance update matrix based on Kalman filtering to form the initial discrimination model.

[0013] Furthermore, a monitoring model based on a timing function is used to monitor parameter fluctuations during self-discharge of a battery cell in real time, including: using a timing function to collect parameters of the fluctuation signal, the timing function being composed of a conventional timing monitoring capture function based on an addition mode and a periodic change monitoring capture function; wherein, the conventional timing monitoring capture function uses confirmation under an infinite continuation timing as a window for opening confirmation, and the periodic change monitoring capture function uses confirmation under a periodic continuation timing as a window for opening confirmation.

[0014] Furthermore, the alarm conditions of the monitoring model include: dividing the data of the key judgment parameter set into multiple modules, calculating the monitoring value and quantile interval of each module; when the monitoring value corresponding to the parameter fluctuation exceeds the quantile interval, the alarm condition of the monitoring model is met.

[0015] Furthermore, in the process of obtaining the determination model for predicting lithium battery voltage runaway, the update of the determination model only modifies the weight parameters of the initial determination model, and the function structure and coefficients of the monitoring model remain fixed.

[0016] A second aspect of the present invention provides a lithium battery voltage runaway prevention system based on data scenario adaptation.

[0017] The lithium battery voltage out-of-control prevention system based on data scenario adaptation includes: The key judgment parameter set determination module is configured to: collect multiple parameter tuples related to the self-discharge of the battery cell and select a key judgment parameter set from them; The calibration module is configured to: produce and calibrate abnormal cells and mix them with normal cells, collect process data corresponding to the key judgment parameter set, and generate a calibration data set; The initial discrimination model construction module is configured to: construct an initial discrimination model based on the data scenario of the obtained calibration data set. Specifically, if the construction is for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if the construction is for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filter method to construct an initial discrimination model; The monitoring module is configured to: use a monitoring model based on a time series function to monitor parameter fluctuations of the battery cell during self-discharge in real time; The lithium battery voltage out-of-control prevention module is configured as follows: when the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set within the alarm time state is extracted to generate a corrected data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage out-of-control prediction. A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the method for preventing lithium battery voltage out-of-control based on data scenario adaptation as described in the first aspect of the present invention.

[0018] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and runnable on the processor. When the processor executes the program, it implements the steps in the lithium battery voltage out-of-control prevention method based on data scenario adaptation as described in the first aspect of the present invention.

[0019] One or more of the above technical solutions have the following beneficial effects: (1) The present invention collects multiple parameter tuples related to the self-discharge of the battery cell and calculates the covariance between each parameter tuple and the self-discharge data; selects a key judgment parameter set from the parameter tuple according to the covariance value, that is, based on the covariance quantitative analysis, it selects multiple core parameters that are strongly correlated with self-discharge, which can eliminate the interference of redundant dimensions; then, the sample library is constructed using the selected key judgment parameter set, which can ensure that the data used for subsequent training is more focused on effective features. Therefore, compared with the existing technology, the present invention can achieve multi-dimensional collaborative anti-interference judgment through this key judgment parameter screening mechanism, thereby avoiding the occurrence of problems such as misjudgment, over-judgment and missed judgment.

[0020] (2) Based on the calibration data set, the present invention constructs an initial judgment model for voltage out of control by using preset judgment rules and loss functions. At the same time, based on the key judgment parameter set, a monitoring model independent of the initial judgment model is constructed. The initial judgment model is used to process real-time classification, and a time series function is used based on the independent monitoring model to capture data drift. On the one hand, the present invention no longer uses a one-time modeling model as the only judgment standard, but continuously optimizes the judgment model after analyzing the time series data through the monitoring model. Therefore, even if multi-dimensional data is introduced, it will not affect the stability of the model. On the other hand, it can also make the judgment model better adapt to the data offset phenomenon.

[0021] (3) When parameter fluctuations reach the alarm condition of the monitoring model, the present invention extracts new data from the key judgment parameter set within the alarm time state to generate a corrected data set and inputs it into the initial judgment model for training to obtain a judgment model for lithium battery voltage runaway prediction. The entire process achieves completely self-adaptive correction without human intervention, namely, through additive mode fluctuation perception and quantile triggering mechanism, real-time response to time series fluctuations and avoidance of secondary deviations.

[0022] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0024] Figure 1 This is a flow chart of a method for preventing lithium battery voltage runaway based on data scenario adaptation in Example 1 of the present invention.

[0025] Figure 2 This is a schematic diagram of the logic for generating the initial determination model in the first embodiment of the present invention.

[0026] Figure 3 This is a logical diagram of generating the monitoring model in the first embodiment of the present invention. DETAILED DESCRIPTION

[0027] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0028] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0029] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0030] Example 1 This embodiment discloses a method for preventing lithium battery voltage runaway based on data scenario adaptation.

[0031] like Figure 1 As shown, the lithium battery voltage runaway prevention method based on data scenario adaptation includes: Step S1: collecting multiple parameter tuples related to cell self-discharge and selecting a key judgment parameter set from them; Step S2: Prepare and calibrate abnormal cells and mix them with normal cells, collect process data corresponding to the key judgment parameter set, and generate a calibration data set; Step S3: construct an initial discrimination model based on the data scenario of the obtained calibration data set. Specifically, if the model is constructed for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if the model is constructed for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filter method to construct an initial discrimination model; Step S4: using a monitoring model based on a time series function to monitor parameter fluctuations during self-discharge of the battery cell in real time; Step S5: When the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set in the alarm state is extracted to generate a revised data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage runaway prediction.

[0032] Based on the above process, the present invention can address the risk of robustness collapse in the model system used for lithium battery voltage runaway prediction due to data trend progression. To facilitate understanding of the technical solution of the present invention, the specific implementation method of the technical solution of the present invention is further explained and illustrated below.

[0033] In step S1, multiple parameter tuples related to the self-discharge of the battery cell are collected, and the covariance between each parameter tuple and the self-discharge data is calculated; and a key judgment parameter set is selected from the parameter tuple according to the covariance value.

[0034] Collect multiple parameter tuples related to the self-discharge of the battery cell, that is, collect the parameters that can be collected automatically in the entire manufacturing chain. By exploring the collection points, the parameters that may be related to the self-discharge of the battery cell in the entire manufacturing process are marked and named ;in, Indicates the measures taken in relation to the self-discharge of the battery cell. tuple of parameters.

[0035] The parameter tuple The corresponding self-discharge data Perform covariance calculation, that is: ; in, represents the covariance calculation, Indicates the The first parameter in the tuple A specific data point, that is, the parameter Observation values ​​at each sampling moment or cell sample; Indicates the Self-discharge data of battery cell samples.

[0036] After calculating the covariance between different parameter tuples and their corresponding self-discharge data, a key determination parameter set is selected from the parameter tuples based on the covariance values. As an optional embodiment, the first 10 to 15 parameter tuples with the largest covariance values ​​can be selected as the key determination parameter set.

[0037] In step S2, calibrated abnormal cells are prepared; the calibrated abnormal cells are mixed with normal cells to perform full-process testing, and process data corresponding to the key judgment parameter set are collected to generate a calibration data set.

[0038] The production and calibration of abnormal cells is to produce a batch of defective samples of the defective types of cells that have been found and mark them. In this embodiment, the abnormal cells produced mainly include various defective types such as diaphragm damage, electrode dislocation, and foreign matter intrusion.

[0039] After the abnormal cells are manufactured, it is necessary to confirm the bad identification and collect data. The abnormal cells with markings are mixed with normal cells to conduct normal testing of the whole process. The mixed cells are used as normal products to do the whole process production to obtain the parameters of the whole process, that is: physical identification is applied to the abnormal cells with markings, so that they undergo the formation, aging and capacity separation processes synchronously with the normal cells to obtain the data of all cells. .

[0040] Based on step S2, the present invention mixes abnormal battery cells with normal battery cells according to the production ratio, so that they operate synchronously in the series circuit. This can better capture the voltage fluctuation characteristics of abnormal battery cells in a real series environment. For example, after the abnormal battery cells with damaged diaphragms are mixed, their voltage drop will be restrained by the adjacent normal battery cells. Different from the "isolated fluctuation" in individual tests, this data enables the model to more accurately identify out-of-control phenomena in actual production.

[0041] In step S3, an initial judgment model is constructed based on the data scenario of the obtained calibration data set. Specifically: if it is constructed for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if it is constructed for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filtering method to construct an initial discrimination model. In the actual implementation process, if any of the conditions of the number of samples being less than 1,000 or the dimension of the key judgment parameter being less than 10 dimensions is met, it is considered to belong to a small data volume scenario; otherwise, it is considered to belong to a large data volume scenario. This can be achieved specifically through the following methods: Step S3-1: When constructing the initial judgment model, if it is constructed for a small data volume scenario, first, the calibration data set is divided into abnormal subsets. and normal subset , and robustly normalize the abnormal subset and the normal subset; then, perform nonlinear dimensionality reduction through the exponential kernel function and construct a classification function; use the constructed classification function as the initial judgment model (high discrimination ability), as follows Figure 2 shown.

[0042] 1) Perform robust normalization on the calibration dataset, namely: ; in, represents the value after robust normalization, Indicates the original value, Indicates the median value of the series data; Represents the interquartile range of the series data, that is, the upper quartile - the lower quartile.

[0043] 2) Subset the anomaly and normal subset Make distinctions.

[0044] First, assume a plane, namely: ; in, represents the plane vector of the assumed plane, represents the intercept.

[0045] Then, based on the assumed plane, a data model is constructed, namely: .

[0046] 3) For abnormal subsets and normal subset Perform the transformation and determine the plane function based on the transformed parameters.

[0047] In the actual production process, most of the data is nonlinear. Therefore, it is necessary to perform kernel function operations such as dimensionality increase and deformation on the data (abnormal subsets and normal subsets) to solve: ; ; in, It represents a constant, which is used to confirm the coefficient of constraining misallocation. In actual application, the specific value of this constant can be set according to the actual situation. Indicates the interval float parameters; represents the kernel function, that is, the conversion function used for data transformation; Indicates the sample label, abnormal cells are marked as +1, and normal cells are marked as -1, which is used for classification training in supervised learning.

[0048] As an optional embodiment, when setting the conversion function, a polynomial solution or an exponential solution may be adopted, that is: A. Polynomial solution ; in, Represents the multiplication factor, which is used to adjust the size of the equation. represents the constant term used for intercept adjustment, Represents the polynomial order used to control the complexity of the scheme.

[0049] B. Exponential solution .

[0050] When selecting a kernel function, the exponential scheme is preferred, as it is more flexible and has strong discrimination capabilities. When computing resources are limited, the polynomial scheme can be selected. After performing parameter transformation based on the kernel function, the transformed parameters are substituted into the plane function to determine the plane function.

[0051] By mapping low-dimensional nonlinear data into a high-dimensional linearly separable space using an exponential kernel function, the model can uncover hidden nonlinear correlations (such as the implicit relationship between voltage fluctuation rate and cell defects) in small samples, creating clear boundaries between previously overlapping normal and abnormal samples in the high-dimensional space. This nonlinear dimensionality reduction capability gives the model high resolution, resolving the existing problem of linear models failing to fit nonlinear relationships. Furthermore, by constructing a classification function based on a "key decision parameter set" (strongly correlated parameters identified through covariance screening), combined with the exponential kernel function's amplification of core features, the model can focus on valid features (such as self-discharge voltage difference and internal resistance change rate) within a limited sample, enhancing the characteristic differences between normal cells and cells with calibrated abnormalities (such as separator damage and electrode misalignment). This combination of "feature focusing and nonlinear amplification" enables the model to accurately distinguish subtle anomalies (such as the tiny voltage fluctuations associated with early-stage micro-shorts) even with small data volumes, thereby addressing the existing problem of poor model stability due to mutually exclusive or redundant multi-dimensional parameters.

[0052] Step S3-2: When constructing the initial judgment model, if it is constructed for a large data volume scenario, first, the calibration data set is divided into abnormal subsets. and normal subset , and perform Z-Score normalization on the abnormal subset and the normal subset; then, the normal subset is used as the main data and the abnormal subset as the noise data, and divided into multiple segments according to the time series; finally, the state equation and covariance update matrix are constructed based on the Kalman filter to form an initial judgment model based on the non-measurement parameter type of big data (high anti-interference ability), as follows Figure 2 shown.

[0053] When the number of calibration datasets changes or increases, this strategy can easily lead to overfitting due to dimensionality explosion. Under this premise, dimensionality selection is made, and the basis for dimension selection is still the confirmation of covariance correlation. This upfront confirmation of correlation and comprehensive consideration of dimensions can prevent data corruption caused by crude single- and multi-dimensional approaches. Furthermore, during data collection, the battery cell may experience external changes (such as regular 4M changes). The entire system state is divided into a hidden state X (unknown) and an observed state Z (with the total data display set). Based on this, the initial judgment model can be constructed using the following method.

[0054] 1) Perform standard deviation normalization on the calibration data set, namely: ; in, represents the value normalized by the standard deviation, represents the mean, Represents standard deviation.

[0055] 2) Define and distinguish between abnormal subsets and normal subsets.

[0056] Defining a normal subset The data is the main data , define the abnormal subset The data is noisy data , master data With noisy data Together they form the total data .

[0057] Master Data Split according to the time sequence of the starting data, that is, divide the main data into ;in, Indicates the first Master data; For noisy data Segment according to the time sequence of the starting data, that is, divide the noise data into ;in, Indicates the first Noisy data; For total data Split according to the time sequence of the starting data, that is, divide the total data into ;in, Indicates the first Total data.

[0058] 3) Construct prediction function and prediction covariance.

[0059] according to State and control quantity under time to predict The state of the time, and then construct the prediction function, namely: ; in, Indicates based on The state of the moment The predicted state at the moment (such as the estimated value of the battery cell voltage), which is the prior estimate in the Kalman filter; Indicates in Total data The state conversion mode under , that is, the transformation matrix or vector state; express An externally known transformation matrix in temporal form, express The true state of the data under time.

[0060] The covariance prediction model used to predict covariance can be expressed as: ; in, express The covariance matrix of the predicted state at each moment is used to measure the uncertainty of the predicted value. The larger the matrix element, the greater the error fluctuation of the corresponding state. express For noisy data under temporal conditions The covariance of the predicted values.

[0061] On this basis, the Total data Expressed as: ;in, express The observation matrix under time is the Gaussian distribution model corresponding to the main data.

[0062] 4) Construct update function and prediction covariance.

[0063] according to The observed and predicted values ​​at time t, calculated The estimated state value at the moment is: ; in, express The weights of predicted values ​​and observed values ​​under different time states.

[0064] After the covariance is updated, it can be expressed as: ; ; in, express.

[0065] Thus, the probability covariance matrix , updated weight value , the optimal state estimation of the entire system at each moment , completed by the superposition of prediction and update Tense and Temporal link. Through the temporal link of "prediction-update", the Kalman filter finally outputs the optimal state estimate at each moment and the covariance matrix , together they constitute the basic parameters of the initial discrimination model. On this basis, the model can realize voltage out-of-control discrimination in the following ways: for new battery cell time series data, use the state equation (dynamic model based on temporal link) to predict its parameter evolution trend; combine the covariance matrix to determine whether the deviation between the actual observation value and the predicted value exceeds the "normal fluctuation range" (defined by the normal subset time series law of the calibration data set); if the deviation exceeds the range, it is determined to be a voltage out-of-control risk. Therefore, Tense and The temporal link reflects the adaptability of the initial discriminant model to complex working conditions.

[0066] Step S3-2 can solve the core problems of existing technologies in large data scenarios, such as "parameter scale interference, neglect of timing rules, insufficient adaptation to data drift, and weak noise interference resistance" by adapting to the normalization, timing segmentation, and Kalman filter dynamic modeling of the timing characteristics of big data. Ultimately, it constructs an initial judgment model with high anti-interference ability, providing a reliable foundation for the precise prevention of lithium battery voltage out of control.

[0067] In step S4, a monitoring model based on a time series function is used to monitor parameter fluctuations during self-discharge of the battery cell in real time.

[0068] By constructing a monitoring function for time series, the resulting set of key judgment parameters is used to build an automated adaptive model, namely the monitoring model. The monitoring model presupposes the perception of correction demand, i.e., by constructing a time-related function to perceive process fluctuations, converting the fluctuation characteristics into correction signals to initiate corrections, which are then used to confirm alarms for subsequent data. After an alarm occurs, the temporal data is automatically fitted, and the process returns to step S4. Upon completion, the judgment is re-performed; and the monitoring model remains unchanged. The existence of an adaptive model can avoid the false interference of periodic regular fluctuations on the entire model, and the control of nodes can also further reduce the possibility of missing truly bad nodes.

[0069] The monitoring model adopts a timing function to monitor parameter fluctuations in real time, including: using a timing function to collect parameters of the fluctuation signal, and the timing function is composed of a conventional timing monitoring capture function based on the addition mode and a periodic change monitoring capture function; among them, the conventional timing monitoring capture function uses confirmation under infinite duration timing as the window for opening confirmation, and the periodic change monitoring capture function uses confirmation under periodic duration timing as the window for opening confirmation.

[0070] As a result, the monitoring model is independent of the initial decision model, capturing data fluctuations solely through the timing function. Triggering corrections only adjusts the decision model weights, without changing the monitoring function structure. For example, if the decision model drifts due to equipment aging, the monitoring model can still identify the anomaly through timing fluctuations and trigger corrections, forming a closed "monitoring-correction" loop. This makes the system more robust than existing coupled models.

[0071] The conventional timing monitoring capture function and periodic change monitoring capture function based on the addition mode can be expressed as: ; in, Represents a timing function in addition mode, Indicates the regular timing monitoring capture function, Indicates the periodic change monitoring and capture function.

[0072] Conventional timing monitoring capture function The confirmation window is opened with the infinite time sequence, that is: ; in, Represents a trigger function, which is 1 when the time exceeds the confirmation window and 0 otherwise; Indicates the rate of change compared to the median value of the data in the previous iteration of the model.

[0073] Periodic change monitoring capture function The confirmation window is opened with the periodic continuous timing confirmation, that is: ; in, represents the harmonic frequency, Indicates the period value; in this embodiment, the harmonic frequency ; The cycle value is one year, that is . 、 They represent the periodic functions The cosine and sine coefficients of the subharmonics are obtained by least squares fitting and are used to quantify the amplitude and phase of the periodic fluctuations. 、 satisfy: .

[0074] In step S5, when the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set in the alarm time state is extracted to generate a revised data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage out of control prediction.

[0075] The monitoring model's alarm conditions involve dividing the data from the key decision parameter set into multiple modules and calculating the monitoring value and quantile interval for each module. When the monitoring value corresponding to the parameter fluctuation exceeds the quantile interval, the monitoring model's alarm condition is reached. This eliminates the need for quantile calculations to rely on assumptions such as Gaussian distributions, making it particularly suitable for asymmetric distributions of lithium battery parameters (such as the right-skewed distribution of self-discharge rate). This improves robustness compared to fixed thresholds based on mean-standard deviation.

[0076] Specifically, by segmenting the original data forming the model in time series, Modules (controlled within 30~50) perform timing functions Sampling to form a trajectory, namely: ; in, Indicates the The timing monitoring function value of a module is used to reflect the fluctuation trajectory of the module parameters over time.

[0077] For each predicted point, calculate the trajectory and score, that is: ; ; ; in, Indicates triggering the correction signal upper limit, Indicates the lower limit of triggering correction signal. Figure 3 As shown, when the monitoring value corresponding to the parameter fluctuation exceeds the quantile interval (that is, it exceeds the upper and lower limit intervals of the trigger correction signal after verification by the monitoring model), it is considered that the trigger condition has been met, and the new data of the key judgment parameter set in the alarm time state can be extracted to generate a correction data set (that is, the current mass production data + marked data) and input into the initial judgment model for training to obtain a new judgment model for lithium battery voltage out of control prediction, that is, it can be judged in the subsequent according to the weight collection rule. After determining the correction need, the random judgment model formed by the trigger correction area data is used as the new judgment model.

[0078] It's also important to note that when developing the decision model for predicting lithium battery voltage runaway, updates to the decision model only modify the weight parameters of the initial decision model, while the function structure and coefficients of the monitoring model remain fixed. Maintaining a fixed monitoring model ensures consistent alarm criteria. For example, regardless of how the decision model is updated, the monitoring model consistently fits temperature and humidity fluctuations on an annual basis, thus avoiding confusion in alarm logic caused by model adjustments.

[0079] Example 2 This embodiment discloses a lithium battery voltage runaway prevention system based on data scenario adaptation.

[0080] The lithium battery voltage out-of-control prevention system based on data scenario adaptation includes: The key judgment parameter set determination module is configured to: collect multiple parameter tuples related to the self-discharge of the battery cell, calculate the covariance between each parameter tuple and the self-discharge data; and select the key judgment parameter set from the parameter tuple according to the covariance value; The calibration module is configured to: produce calibration abnormal cells; mix the calibration abnormal cells with normal cells to perform full process testing, and collect process data corresponding to the key judgment parameter set to generate a calibration data set; The initial determination model building module is configured to: construct an initial determination model for voltage out of control according to the obtained calibration data set through preset determination rules and loss functions; The monitoring model construction module is configured to: construct a monitoring model independent of the initial judgment model based on the obtained key judgment parameter set, wherein the monitoring model uses a time series function to monitor parameter fluctuations in real time; The lithium battery voltage runaway prevention module is configured to: when the parameter fluctuation reaches the alarm condition of the monitoring model, extract new data of the key judgment parameter set within the alarm time state to generate a corrected data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage runaway prediction. Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0081] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for preventing lithium battery voltage runaway based on data scenario adaptation as described in the first embodiment of the present disclosure.

[0082] Example 4 The purpose of this embodiment is to provide an electronic device.

[0083] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for preventing lithium battery voltage runaway based on data scenario adaptation as described in the first embodiment of the present disclosure are implemented.

[0084] The steps involved in the apparatuses of Examples 2, 3, and 4 above correspond to those of Method Example 1. For detailed implementations, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any method of the present invention.

[0085] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, and thus stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0086] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A method for preventing lithium battery voltage out of control based on data scenario adaptation, characterized in that: include: Collect multiple parameter tuples related to cell self-discharge and select key judgment parameter sets from them; Make and calibrate abnormal cells and mix them with normal cells, collect process data corresponding to the key judgment parameter set, and generate a calibration data set; According to the data scenario of the obtained calibration data set, an initial discrimination model is constructed. Specifically: if the construction is for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if the construction is for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filter method to construct an initial discrimination model; A monitoring model based on time series function is used to monitor the parameter fluctuations of the battery cell during self-discharge in real time; When the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set within the alarm temporal state is extracted to generate a revised data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage out of control prediction.

2. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: A monitoring model based on a timing function is used to monitor parameter fluctuations during self-discharge of a battery cell in real time, including: using a timing function to collect parameters of the fluctuation signal, wherein the timing function is composed of a conventional timing monitoring capture function based on an addition mode and a periodic change monitoring capture function; wherein the conventional timing monitoring capture function uses confirmation under an infinite continuation timing as a window for opening confirmation, and the periodic change monitoring capture function uses confirmation under a periodic continuation timing as a window for opening confirmation.

3. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: Make calibrated abnormal cells and mix them with normal cells, and collect process data corresponding to the key judgment parameter set, including: mixing the calibrated abnormal cells with normal cells for full process testing, applying physical identification to the calibrated abnormal cells, and making them undergo the formation, aging, and capacity separation processes synchronously with the normal cells.

4. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: The normalized calibration dataset is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discriminant model, including: first, the calibration dataset is divided into abnormal subsets and normal subsets, and robust normalization is performed on the abnormal subsets and normal subsets; then, nonlinear dimensionality reduction is performed using an exponential kernel function, and a classification function is constructed; the constructed classification function is used as the initial discriminant model.

5. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: The normalized calibration data set is filtered based on the Kalman filtering method to construct an initial discrimination model, including: first, dividing the calibration data set into an abnormal subset and a normal subset, and performing Z-Score normalization on the abnormal subset and the normal subset; then, taking the normal subset as the main data and the abnormal subset as the noise data, and dividing them into multiple segments according to the time series; finally, constructing the state equation and covariance update matrix based on the Kalman filtering to form the initial discrimination model.

6. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: The alarm conditions of the monitoring model include: dividing the data of the key judgment parameter set into multiple modules, calculating the monitoring value and quantile interval of each module; when the monitoring value corresponding to the parameter fluctuation exceeds the quantile interval, the alarm condition of the monitoring model is met.

7. The method for preventing lithium battery voltage out of control based on data scenario adaptation according to claim 1, characterized in that: In the process of obtaining the judgment model for predicting lithium battery voltage runaway, the update of the judgment model only modifies the weight parameters of the initial judgment model, and the function structure and coefficients of the monitoring model remain fixed.

8. The lithium battery voltage out-of-control prevention system based on data scenario adaptation is characterized by: include: The key judgment parameter set determination module is configured to: collect multiple parameter tuples related to the self-discharge of the battery cell and select a key judgment parameter set from them; The calibration module is configured to: produce and calibrate abnormal cells and mix them with normal cells, collect process data corresponding to the key judgment parameter set, and generate a calibration data set; The initial discrimination model construction module is configured to: construct an initial discrimination model based on the data scenario of the obtained calibration data set. Specifically, if the construction is for a small data volume scenario, the normalized calibration data set is subjected to nonlinear dimensionality reduction processing using an exponential kernel function to construct an initial discrimination model; if the construction is for a large data volume scenario, the normalized calibration data set is filtered based on the Kalman filter method to construct an initial discrimination model; The monitoring module is configured to: use a monitoring model based on a time series function to monitor parameter fluctuations of the battery cell during self-discharge in real time; The lithium battery voltage out-of-control prevention module is configured as follows: when the parameter fluctuation reaches the alarm condition of the monitoring model, new data of the key judgment parameter set within the alarm time state is extracted to generate a corrected data set and input it into the initial judgment model for training to obtain a judgment model for lithium battery voltage out-of-control prediction.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for preventing lithium battery voltage out of control based on data scenario adaptation are implemented as described in any one of claims 1 to 7.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the lithium battery voltage out-of-control prevention method based on data scenario adaptation are implemented as described in any one of claims 1 to 7.

Images