Intelligent grading early warning method for thermal runaway of battery

Abstract

The invention relates to a battery thermal runaway intelligent grading early warning method, and relates to the technical field of battery thermal runaway early warning, and the method comprises the steps: loading a pre-control parameter and an environment temperature of a target cell, and constructing a cell dynamic portrait; obtaining a first battery cell temperature extreme value through a battery cell temperature extreme value prediction model; when the first battery cell temperature extreme value is smaller than the thermal runaway temperature threshold value, loading the abnormal compression time sequence information and the cycle index of the target battery cell, and constructing a battery cell static portrait; when the abnormal compression time sequence information is not empty, or / and the cycle index is greater than a cycle index threshold value, retrieving a sample set of the battery cells of the same model; counting the proportion of the thermal runaway samples in the same type of battery cell sample set, and setting the proportion as a target battery cell thermal runaway risk coefficient; and executing the grading early warning strategy. The problems that thermal runaway is difficult to accurately predict, early warning lags and potential safety hazards exist due to the fact that traditional battery thermal runaway management depends on a battery pack temperature sensor to collect temperature and early warning is executed only when the temperature reaches a preset value are solved.

Description

Technical Field

[0001] This application relates to the field of battery thermal runaway early warning, and in particular to an intelligent graded early warning method for battery thermal runaway. Background Technology

[0002] With the increasing demands for battery safety in fields such as new energy storage and electric vehicles, accurate early warning of battery thermal runaway has become a key technical requirement for ensuring equipment stability and avoiding safety accidents.

[0003] Currently, traditional battery thermal runaway management methods do not adequately address the impact of complex internal chemical reactions, physical processes, and external environmental temperature fluctuations. They cannot overcome the limitations of relying solely on battery pack temperature sensors to collect temperature data and only issuing warnings when preset values ​​are reached. This not only makes it difficult to accurately predict thermal runaway but also increases safety risks during battery use due to delayed warnings. Summary of the Invention

[0004] This application provides an intelligent graded early warning method for battery thermal runaway, which improves upon the shortcomings of traditional battery thermal runaway management, which relies on battery pack temperature sensors to collect temperature data and only executes early warnings when the temperature reaches a preset value. This results in thermal runaway being difficult to predict accurately, having delayed early warnings, and posing safety hazards.

[0005] The embodiments of this application disclose the following technical solutions: This application provides a method for intelligent graded early warning of battery thermal runaway, the method comprising: Load the pre-control parameters and ambient temperature of the target battery cell to construct a dynamic profile of the battery cell; The cell temperature extreme value prediction model is used to process the cell dynamic profile to obtain the first cell temperature extreme value. The cell temperature extreme value prediction model is obtained by training the cell dynamic historical profile set with zero abnormal pressure and a cycle number less than or equal to the cycle number threshold and the cell temperature extreme value detection set using machine learning. When the extreme temperature of the first cell is less than the thermal runaway temperature threshold, the abnormal pressure timing information and cycle number of the target cell are loaded to construct a static profile of the cell. When the abnormal pressure timing information is not empty, or / and the number of cycles is greater than the number of cycles threshold, retrieve a sample set of the same type of battery cell that satisfies the static profile and the dynamic profile of the battery cell; The proportion of thermal runaway samples in the same type of battery cell sample set is set as the thermal runaway risk coefficient of the target battery cell; Based on the thermal runaway risk coefficient of the target battery cell, a graded early warning strategy corresponding to the coefficient is implemented.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application proposes an intelligent graded early warning method for battery thermal runaway. By constructing dynamic and static dual profiles of the battery cell, training a battery cell temperature extreme value prediction model, quantifying the thermal runaway risk coefficient, and executing graded early warning strategies in a coordinated manner, it achieves accurate prediction and differentiated management of battery thermal runaway risk. First, the pre-control parameters and ambient temperature of the target battery cell are loaded to construct a dynamic profile of the battery cell reflecting its real-time operating status. Then, a temperature extreme value prediction model trained with data from battery cells with no pressure damage and low cycle loss is used to process the dynamic profile of the battery cell to obtain the first extreme temperature value of the battery cell. If the first extreme temperature value of the battery cell is less than the thermal runaway threshold, the abnormal pressure timing information and cycle count of the battery cell are loaded to construct a static profile of the battery cell. When there is an abnormal pressure record of the battery cell or the cycle count exceeds the threshold, the sample set of the same model of battery cell is retrieved and the proportion of thermal runaway samples is counted and set as the thermal runaway risk coefficient. Finally, based on the predefined risk threshold, prompt information, maintenance prompts, and warnings for prohibited operations are sent for low, medium, and high risk levels, respectively. If the first extreme temperature value of the battery cell exceeds the thermal runaway threshold, the thermal runaway risk coefficient is directly configured to 1 and the highest level warning is activated.

[0007] This technical solution addresses the problems of traditional thermal runaway management, such as overemphasizing the overall state of the battery pack while neglecting the detailed analysis of individual cells and issuing warnings only when temperatures are abnormal, through steps including dynamic profiling to analyze the real-time operating status of battery cells, static profiling to analyze long-term damage and loss, temperature extreme value prediction model to predict temperature boundaries in advance, risk quantification supported by a sample set of battery cells of the same model, and differentiated management through a graded early warning strategy. It provides technical support for the safe operation and efficient maintenance of battery systems. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart illustrating a battery thermal runaway intelligent graded early warning method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating the implementation of a graded early warning strategy based on the thermal runaway risk coefficient of the target battery cell, as provided in an embodiment of this application. Detailed Implementation

[0010] This application provides a battery thermal runaway intelligent hierarchical early warning method to solve the technical problem in the prior art that battery thermal runaway management relies on battery pack temperature sensors to collect temperature and only executes early warning when the temperature reaches a preset value, which makes it difficult to accurately predict thermal runaway and poses safety hazards.

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0013] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0014] Examples, as shown in the appendix Figure 1 As shown, this application provides a smart graded early warning method for battery thermal runaway, the method comprising the following steps: S110: Load the pre-control parameters and ambient temperature of the target battery cell to construct a dynamic profile of the battery cell; In this embodiment of the application, in order to break through the limitations of traditional reliance on overall battery pack temperature monitoring, and to provide a data basis for subsequent temperature extreme value prediction by analyzing the real-time operating characteristics of the battery cell, it is necessary to first load the pre-control parameters of the target battery cell and the ambient temperature, and then construct a dynamic profile of the battery cell to achieve a refined characterization of the current operating state of the battery cell.

[0015] Specifically, the scope of the target cell's pre-control parameters must first be clarified. These pre-control parameters cover the core performance parameters such as the rated voltage, rated capacity, and internal resistance reference value calibrated at the time of manufacture. They are the basis for reflecting the inherent properties of the cell and judging whether its operation deviates from the normal range.

[0016] Meanwhile, real-time ambient temperature data of the area where the target cell is located is collected from an ambient temperature sensor deployed separately for the target cell within the battery pack. Fluctuations in ambient temperature directly affect the rate of chemical reaction and heat dissipation efficiency inside the cell, and are key external factors causing changes in cell temperature.

[0017] Furthermore, after acquiring the pre-control parameters and ambient temperature data, the two types of data need to be synchronously integrated. During the integration process, it is necessary to ensure the consistency of the data timestamps, that is, each set of pre-control parameters and the corresponding ambient temperature data should correspond to the same monitoring time, to avoid discrepancies between the dynamic profile of the battery cell and the actual operating state of the battery cell due to time deviations.

[0018] At the same time, the validity of the collected data is verified. For example, it is checked whether the pre-control parameters are within the reasonable error range of the factory calibration, and whether there are jump values ​​in the ambient temperature data caused by sensor abnormalities. If data abnormalities are found, the data re-sampling mechanism is immediately triggered to obtain accurate data from the backup sensor to ensure the reliability of the data source for the construction of the cell profile.

[0019] Furthermore, the pre-control parameters of the target battery cell, after being verified by timestamp synchronization and validity screening, are associated and bound with real-time ambient temperature data to form a dataset that can fully reflect the inherent performance benchmark and external environmental influence of the battery cell at a specific moment. Based on this dataset, a dynamic profile of the battery cell is constructed to ensure that the profile can accurately reflect the current basic operating status and environmental adaptability of the target battery cell.

[0020] For example, for a certain type of ternary lithium battery cell, its pre-control parameters include a rated capacity of 200Ah and an internal resistance benchmark value of 80mΩ. The ambient temperature of the cell at a certain moment is collected as 25℃. After the above data is correlated and integrated, the constructed cell dynamic profile can initially present the real-time status of "the ternary lithium battery cell is currently operating at a rated capacity of 200Ah and an internal resistance of 80mΩ as the benchmark performance at 25℃". If the ambient temperature rises to 35℃ in the future, the updated cell dynamic profile will reflect the change in ambient temperature in a synchronous manner.

[0021] The constructed dynamic profile of the battery cell is not a static data set, but is updated in real time with the monitoring cycle. Every preset monitoring time interval (e.g., 1 minute), the latest pre-control parameters and real-time ambient temperature of the target battery cell are automatically reloaded, and the dynamic profile of the battery cell is iteratively updated to ensure that the profile always remains consistent with the current operating state of the battery cell. This provides accurate dynamic data support for the subsequent accurate calculation of the first battery cell temperature extreme value through the battery cell temperature extreme value prediction model.

[0022] S120: The cell temperature extreme value prediction model is used to process the cell dynamic profile to obtain the first cell temperature extreme value. The cell temperature extreme value prediction model is obtained by training the cell dynamic historical profile set with zero abnormal pressure and a cycle number less than or equal to the cycle number threshold and the cell temperature extreme value detection set using machine learning. In this embodiment of the application, in order to accurately obtain the extreme temperature value that the target cell may reach in the future and provide a reliable basis for subsequent risk assessment, it is necessary to process the dynamic profile of the cell through a cell temperature extreme value prediction model obtained through specific training, so as to achieve a forward-looking assessment of the thermal state of the cell.

[0023] Specifically, the preset cell model is first obtained by reading the target cell's factory identification and the cell model information stored in the battery BMS system.

[0024] Furthermore, the obtained preset cell model is input into the threshold calibration table. Through the preset "cell model-threshold" association mapping relationship in the table, the predefined pressure threshold and cycle number threshold corresponding to the cell model are accurately located, so as to clarify the standard for subsequent screening of cell dynamic historical profiles and judging abnormal pressure states.

[0025] Furthermore, data from battery cells with zero abnormal pressure exposures and a cycle count less than or equal to a cycle count threshold are selected. Abnormal pressure exposure requires that the number of times the pressure monitoring value is greater than or equal to the pressure threshold be zero, ensuring that the selected battery cells are not affected by pressure damage. Controlling the cycle count within the threshold ensures that the battery cells are not excessively aged, thus eliminating interference factors and ensuring that the data reflects the temperature characteristics of the battery cells under normal operating conditions.

[0026] Furthermore, according to the preset cell model, obtain the dynamic historical profile of the cell that meets the above standards to form a dynamic historical profile set of the cell; at the same time, with the same preset cell model and dynamic historical profile as constraints, collect the corresponding initial cell temperature extreme value detection value set, perform central tendency analysis on the set, extract the first cell temperature extreme value detection value that can represent the stable temperature performance of the cell, and integrate them to form a cell temperature extreme value detection value set.

[0027] Furthermore, machine learning was used to train a cell temperature extreme value prediction model on a dynamic historical image set of the battery cells and a set of detected extreme temperature values. During training, parameters from the dynamic historical image set were used as input features, and the corresponding detected extreme temperature values ​​were used as output labels. The model parameters were iteratively optimized to ensure that the model could accurately predict extreme temperature values ​​based on the dynamic state of the battery cells.

[0028] Finally, the completed dynamic profile of the battery cell is input into the trained battery cell temperature extreme value prediction model. The model calculates and outputs the first battery cell temperature extreme value by analyzing the pre-control parameters and ambient temperature in the profile. This extreme value can reflect the future thermal state change trend of the battery cell in advance, providing data support for subsequent judgment on whether to trigger further risk assessment.

[0029] Step S120 in the method provided in this application embodiment includes: Obtain the preset battery cell model; Input the preset cell model into the threshold calibration table and match it with the predefined pressure threshold and cycle number threshold. Using the preset cell model as a constraint, collect the dynamic historical profile of the first cell with zero abnormal pressure counts and a cycle count less than or equal to the cycle count threshold. Abnormal pressure counts indicate that the number of times the pressure monitoring value is greater than or equal to the pressure threshold is zero. Based on the preset cell model and the dynamic historical profile of the first cell, an initial set of cell temperature extreme value detection values ​​is collected, where the number of abnormal pressure tests is zero and the number of cycles is less than or equal to the cycle number threshold. Central trend analysis is then performed to obtain the first cell temperature extreme value detection value. Add the first cell dynamic historical image to the cell dynamic historical image set, and add the first cell temperature extreme value detection value to the cell temperature extreme value detection value set.

[0030] In this embodiment of the application, in order to ensure that the training data of the cell temperature extreme value prediction model can accurately reflect the temperature characteristics of the cell under normal operation, it is necessary to select high-quality training data through clear constraints, construct an accurate dynamic historical profile set of the cell and a set of cell temperature extreme value detection values, so as to improve the accuracy of the model in predicting the extreme value of the target cell temperature and provide reliable data support for subsequent thermal runaway risk assessment.

[0031] Specifically, the first step should be to read the factory specifications of the target cell, the cell model record stored in the battery management system (BMS), or directly collect the cell model information marked on the battery pack to obtain the preset cell model.

[0032] Furthermore, the obtained preset cell model is input into the threshold calibration table. Through the stored data in the table that is pre-associated with the preset cell model, the predefined pressure threshold and cycle number threshold corresponding to the cell model are accurately matched.

[0033] The method provided in this application embodiment includes the following steps for constructing the "threshold calibration table": The initial pressure threshold is set to zero, and the number of cycles is set to j, where j represents a positive integer, the initial value of j is equal to 1, and the maximum value of j is equal to the rated number of cycles. Using the preset cell model and preset cell dynamic profile as constraints, a set of first cell temperature extreme value detection values ​​is collected where the number of abnormal pressures is zero and the number of cycles is less than or equal to the number of cycles threshold. Central tendency analysis is then performed to obtain the first fitted value of the cell temperature extreme value. Using the preset cell model and preset cell dynamic profile as constraints, a set of second cell temperature extreme value detection values ​​is collected where the number of abnormal pressure cycles is zero and the number of cycles is equal to j+1 cycles. Central tendency analysis is then performed to obtain the second fitted value of the cell temperature extreme value. When the first temperature extreme value deviation between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is greater than or equal to the temperature extreme value deviation threshold, a pressure threshold is configured based on the cycle number threshold, and the cycle number threshold, the pressure threshold, and the preset cell model are associated and stored, and added to the threshold calibration table. When the deviation between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is less than the temperature extreme value deviation threshold, j is incremented by one, and the loop is executed.

[0034] In this embodiment of the application, in order to match exclusive and accurate pressure thresholds and cycle number thresholds for different models of battery cells, it is necessary to construct a threshold calibration table through cycle testing and deviation judgment to ensure that each preset battery cell model has a corresponding judgment standard that can accurately define the state of no pressure damage and no excessive aging, so as to provide a basis for subsequent screening and training data.

[0035] Specifically, the initial pressure threshold is first set to zero, the initial value of the cycle number threshold j is 1, and the maximum value of j is set to the rated cycle number of the battery cell.

[0036] The initial pressure threshold is set to zero to start testing from the most stringent pressure criteria and gradually find the critical value that can distinguish the effect of pressure on cell temperature. The cycle number threshold starts from 1 to track the effect of cycle number on temperature from the early stage of the cell's life cycle to ensure coverage of the entire life cycle of normal cell use. The maximum value of j is set to the rated cycle number to avoid exceeding the cell's designed lifespan and causing the data to lose its reference value.

[0037] After initial parameter configuration, data is collected under constraints based on preset cell model and preset cell dynamic profile. The preset cell dynamic profile includes pre-control parameters and ambient temperature data within a controllable range. This constraint ensures that the collected cell temperature extreme values ​​are only affected by the number of cycles, eliminating interference from other external environmental or parameter anomalies. Zero abnormal pressure cycles further ensure that the cell has not been damaged by pressure, and the collected temperature extreme values ​​can truly reflect the correlation between the number of cycles and temperature.

[0038] Furthermore, based on the above constraints, a set of first cell temperature extreme value detection values ​​is collected where the number of abnormal pressure tests is zero and the number of cycles is less than or equal to the number of cycles threshold (i.e., not exceeding the currently set j value, where j is initially 1 and does not exceed the rated number of cycles of the cell). Central tendency analysis (i.e., calculating the median) is then performed on this first set of first cell temperature extreme value detection values ​​to obtain the first fitted value of the cell temperature extreme value, so as to eliminate the random error of a single temperature detection and accurately reflect the stable temperature extreme value characteristics of the cell within the range of the number of cycles.

[0039] For example, for a battery cell with the preset model "ternary lithium-18650", the cycle number threshold j=700, collect 50 battery cells of this model, with zero abnormal pressure cycles and a cycle number between 1 and 700, collect 25 temperature extreme values ​​for each battery cell, and obtain 1250 temperature data. After sorting, take the median of 41℃, which is the first fitted value of the battery cell temperature extreme value.

[0040] Furthermore, based on the above constraints, a set of second cell temperature extreme value detection values ​​is collected where the number of abnormal pressure cycles is zero and the number of cycles is equal to the number of cycles j+1. Central tendency analysis (also calculating the median) is performed on this set of second cell temperature extreme value detection values ​​to obtain the second fitted value of the cell temperature extreme value.

[0041] For example, continuing with the "ternary lithium-18650" cell case above, the cycle number threshold j+1=701, 45 cells of this model with zero abnormal pressure cycles and exactly 701 cycles were collected. The temperature extreme values ​​of each cell were collected 25 times, resulting in 1125 temperature data points. After sorting, the median of 44℃ was taken, which is the second fitted value of the cell temperature extreme value.

[0042] Furthermore, the difference between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is calculated to obtain the first temperature extreme value deviation, and this deviation is compared with the predefined temperature extreme value deviation threshold.

[0043] When the deviation of the first temperature extreme value is greater than or equal to the temperature extreme value deviation threshold, it indicates that after the number of cycles increases from ≤j to j+1, the temperature extreme value of the cell has changed significantly. The current cycle number threshold j can effectively define the boundary of the cell not being over-aged. At this time, it is necessary to further configure the corresponding pressure threshold based on the cycle number threshold.

[0044] In the method provided in this application embodiment, "configuring a pressure threshold based on the cycle number threshold" includes: Loading pressure consistency deviation, wherein the pressure consistency deviation is a predefined pressure tolerance deviation; The characteristic pressure is obtained by summing the pressure consistency deviation with the initial pressure threshold. Using the preset cell model and preset cell dynamic profile as constraints, a third set of cell temperature extreme value detection values ​​is collected, where the number of abnormal pressures is not zero, the pressures of the abnormal pressures are all less than the characteristic pressures, and the number of cycles is less than or equal to the number of cycles threshold. Central tendency analysis is then performed to obtain the third fitted value of the cell temperature extreme value. When the deviation between the third fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is greater than or equal to the temperature extreme value deviation threshold, the initial pressure threshold is set as the pressure threshold. Otherwise, when the deviation between the third fitted value of the cell temperature extreme value and the second temperature extreme value of the first fitted value of the cell temperature extreme value is less than the temperature extreme value deviation threshold, the initial pressure threshold is updated using the characteristic pressure, and the loop is executed.

[0045] In this embodiment of the application, in order to determine the critical value that can accurately distinguish the effect of pressure on temperature for different types of battery cells, it is necessary to calculate the pressure consistency deviation and the initial pressure threshold together, collect data and compare the deviation, and gradually optimize and determine the final pressure threshold to ensure that the pressure threshold can cover a reasonable pressure tolerance range and accurately identify the pressure situation that has a significant impact on the temperature of the battery cell.

[0046] Specifically, the first step is to apply a pressure consistency deviation. This pressure consistency deviation is a predefined pressure tolerance based on the cell manufacturing process, material properties, and actual usage scenarios. For example, for a certain model of square lithium iron phosphate battery, considering slight pressure fluctuations during cell assembly and normal deformation during use, the pressure consistency deviation is set to 5 kPa. This deviation value must ensure that pressure fluctuations within a certain range are allowed without affecting the normal temperature performance of the cell, avoiding misinterpreting slight, harmless pressure changes as abnormal pressure.

[0047] Furthermore, after applying the pressure consistency deviation, it is summed with the initial pressure threshold to obtain the characteristic pressure. The initial pressure threshold is configured to zero at the beginning of the threshold calibration table construction, and the characteristic pressure at this time is the pressure consistency deviation; if the initial pressure threshold is subsequently updated, the characteristic pressure will be adjusted synchronously with the change of the initial pressure threshold.

[0048] Furthermore, a set of extreme temperature values ​​for the third battery cell is collected, constrained by a preset battery cell model and a preset battery cell dynamic profile. Similarly, the preset battery cell dynamic profile must include pre-control parameters and ambient temperature within a controllable range to eliminate interference from abnormal pre-control parameters and ambient temperature fluctuations on the battery cell temperature; the number of abnormal pressure tests is not zero and the abnormal pressure tests are all less than the characteristic pressure; the number of cycles is less than or equal to the cycle number threshold.

[0049] Meanwhile, a central tendency analysis (also to calculate the median) was performed on the collected set of third cell temperature extreme values ​​to obtain the third fitted value of the cell temperature extreme value.

[0050] Furthermore, by comparing the second temperature extreme value deviation between the third fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value, the relationship between this deviation and the temperature extreme value deviation threshold is determined.

[0051] Among them, the second temperature extreme value deviation is the difference between the third fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value, reflecting the degree of influence of slight pressure changes on the cell temperature.

[0052] Specifically, if the deviation of the second temperature extreme value is greater than or equal to the temperature extreme value deviation threshold (e.g., 3℃), it indicates that the pressure range corresponding to the current initial pressure threshold has had a significant impact on the cell temperature. Further expanding the pressure range may lead to abnormal temperature fluctuations. Therefore, the initial pressure threshold is set as the final pressure threshold.

[0053] Conversely, if the deviation of the second temperature extreme value is less than the temperature extreme value deviation threshold, it means that the pressure range corresponding to the current initial pressure threshold is still within the tolerance range that the cell can withstand. Slight pressure changes have not had a significant impact on the temperature. The initial pressure threshold needs to be updated using the characteristic pressure, and the steps of "calculating the new characteristic pressure - collecting new data - analyzing the deviation" are executed again until an initial pressure threshold that can make the deviation of the second temperature extreme value reach or exceed the temperature extreme value deviation threshold is found.

[0054] For example, for a battery cell with the preset model "ternary lithium-21700", the cycle number threshold has been determined to be 800 times, the initial pressure threshold is 0 kPa, and the pressure consistency deviation is 5 kPa. First, the characteristic pressure is calculated to be 0 + 5 = 5 kPa.

[0055] Furthermore, with the cell model, controllable pre-control parameters (rated voltage 3.7V, rated capacity 5.0Ah) and ambient temperature (25℃±2℃) as constraints, 40 cell samples were collected with abnormal pressure number not zero, pressure less than 5kPa and cycle number ≤800. For each sample, 30 temperature extreme values ​​were collected, and central tendency analysis was performed to obtain the third fitted value of the cell temperature extreme value of 43℃.

[0056] Furthermore, the previously obtained first fitted value for the cell temperature extreme value was 41℃, and the second temperature extreme value deviation was 2℃, which is less than the temperature extreme value deviation threshold of 3℃. Therefore, the initial pressure threshold was updated to 5kPa. The characteristic pressure was recalculated as 5+5=10kPa. Cell samples under the same conditions with pressures all less than 10kPa were collected. Central tendency analysis showed that the third fitted value for the cell temperature extreme value was 45℃, which deviated from the first fitted value for the cell temperature extreme value by 4℃, which is greater than the temperature extreme value deviation threshold. At this point, 5kPa was set as the final pressure threshold.

[0057] Ultimately, the pressure threshold determined through the above-mentioned step-by-step optimization method can not only avoid misjudging slight pressure fluctuations as abnormalities, but also accurately identify pressure conditions that have a significant impact on cell temperature. This ensures that when screening the dynamic historical profiles of cells, cell samples with pressure damage can be accurately excluded, providing reliable training data for the cell temperature extreme value prediction model.

[0058] Furthermore, after determining the final cycle number threshold and pressure threshold, these two thresholds need to be associated and bound with the corresponding preset cell model to form a complete correspondence of "cell model - cycle number threshold - pressure threshold".

[0059] At the same time, the result needs to be determined based on the first temperature extreme value deviation between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value.

[0060] Specifically, if the cycle number threshold and pressure threshold of the current model of battery cell have been configured and stored in the threshold calibration table because the first temperature extreme value deviation is greater than or equal to the temperature extreme value deviation threshold, then the threshold calibration process of the battery cell of that model is completed.

[0061] Conversely, if the previous deviation of the first temperature extreme value was less than the temperature extreme value deviation threshold, it means that the current cycle number threshold j has not yet reached the boundary that can cause a significant change in the cell temperature. The value of j needs to be increased by 1 (i.e., from the initial 1 to 2, then from 2 to 3, etc.), and the entire process of "collecting data with the new j as the cycle number threshold - calculating the first and second fitted values ​​- comparing the first temperature extreme value deviation" is re-executed until a cycle number threshold that can make the first temperature extreme value deviation meet the threshold requirement is found. After configuring the pressure threshold based on this cycle number threshold, the complete correlation is added to the threshold calibration table to ensure that each preset cell model can find its own judgment threshold in the threshold calibration table.

[0062] Furthermore, once the threshold calibration table has stored the pressure threshold and cycle number threshold corresponding to the preset cell model, the preset cell model can be input into the threshold calibration table. Based on the pre-established "cell model-threshold" association in the table, the predefined pressure threshold and cycle number threshold specific to that cell model can be automatically matched.

[0063] Furthermore, after matching the corresponding threshold, the battery cell samples that match the preset battery cell model are selected as constraints, while ensuring that the samples meet the conditions that the number of abnormal pressure tests is zero and the number of cycles is less than or equal to the number of cycles threshold.

[0064] Abnormal pressure is determined by the number of times the pressure monitoring value is greater than or equal to the matched pressure threshold. This excludes cells that have been damaged by pressure or are excessively aged. Dynamic historical data of these cells that meet the conditions are collected to form the first dynamic historical profile of the cell. This profile must fully cover key dynamic information such as the pre-control parameters and ambient temperature of the cell under normal operating conditions.

[0065] Furthermore, using the preset cell model and the dynamic historical profile of the first cell as dual constraints, the set of initial cell temperature extreme value detection values ​​is further collected.

[0066] During the data collection process, operating conditions must be maintained consistent with the dynamic historical profile of the first battery cell, ensuring that the number of abnormal pressure events is zero and the number of cycles does not exceed the pressure threshold. By collecting the extreme temperature values ​​of the battery cells under the same operating conditions multiple times, the random error of a single data entry is reduced. Central tendency analysis (also calculating the median) is performed on the initial set of detected extreme temperature values ​​of the collected battery cells to extract the extreme temperature values ​​of the first battery cell that can represent the stable temperature characteristics of this type of battery cell.

[0067] Furthermore, the collected dynamic historical images of the first battery cell are added one by one to the dynamic historical image set of the battery cell, and the corresponding extreme temperature detection values ​​of the first battery cell are added to the extreme temperature detection value set of the battery cell, ensuring that the entries of the two datasets correspond one-to-one, forming a well-structured and accurate model training sample pair.

[0068] Furthermore, machine learning is used to complete model training, specifically by using gradient boosting decision trees as the basic framework to build a cell temperature extreme value prediction model, so as to accurately capture the complex relationship between cell dynamic parameters and temperature extreme values.

[0069] Specifically, the core parameters of the model are set as follows: the learning rate is set to 0.05 to control the step size of parameter updates in each iteration, avoiding the model from getting trapped in local optima due to excessively rapid updates; the number of leaf nodes in a single tree is limited to 31 to prevent overfitting caused by excessive tree structure complexity; the maximum tree depth is set to 6 to further constrain model complexity and balance fitting ability and generalization ability; the minimum number of samples per leaf node is 20 to ensure that each leaf node has sufficient data support and improve prediction stability.

[0070] Meanwhile, the L1 regularization coefficient is configured to be 0.1 and the L2 regularization coefficient to be 0.2, in order to reduce the model's sensitivity to noisy data through regularization; the objective function is set to the regression type to adapt to the prediction needs of continuous values ​​such as temperature extremes.

[0071] During training, data preprocessing is performed first, which involves removing abnormal samples with temperatures ranging from -20℃ to 60℃ and negative cycle counts. Missing values ​​are filled with the mean of adjacent timestamp features to ensure the purity and integrity of the training dataset and to prevent abnormal or missing data from interfering with the model's learning patterns.

[0072] Furthermore, the dataset is divided into a training set and a validation set in a 7:3 ratio. The training set is used for learning and fitting model parameters, while the validation set is used to monitor performance changes during model training in real time, and to promptly identify overfitting or underfitting issues.

[0073] Furthermore, during model training, a 5-fold cross-validation strategy is adopted, with the root mean square error (RMSE) as the loss function. The parameters are iteratively optimized using the gradient descent algorithm. Training is stopped when the RMSE of the validation set no longer decreases for 5 consecutive rounds to avoid overfitting of the model and ensure that the model has good generalization ability and can accurately predict the extreme temperature values ​​of unknown battery cell data.

[0074] Furthermore, after completing the basic training, an independent test set (accounting for 20% of the total data) is used to evaluate the actual predictive performance of the model. The test set RMSE is required to be ≤1.5℃. If it is not met, the parameters are adjusted (e.g., the regularization coefficient is increased) and the training is repeated until the model meets the standard, so as to obtain the final cell temperature extreme value prediction model that can stably and accurately predict the extreme value of cell temperature.

[0075] Furthermore, after obtaining the cell temperature extreme value prediction model, the previously constructed target cell dynamic profile is input into the cell temperature extreme value prediction model. The model combines the pre-control parameters in the profile with the real-time ambient temperature, and processes them through the built-in algorithm. Finally, it outputs the first cell temperature extreme value that the target cell may reach in the future, providing key temperature basis for subsequent judgment on whether the cell has a risk of thermal runaway.

[0076] For example, for a certain type of lithium iron phosphate battery cell, its dynamic profile includes a rated capacity of 50Ah, an internal resistance reference value of 75mΩ, and a real-time ambient temperature of 28℃. After inputting the dynamic profile of the cell into a trained cell temperature extreme value prediction model, the model calculates that the first cell temperature extreme value is 42℃. This value can be directly used to compare with the thermal runaway temperature threshold to determine whether further static profile analysis of the cell is needed.

[0077] The above steps lay a data foundation for developing thermal runaway risk assessment strategies for subsequent scenarios. This avoids the lag of relying solely on real-time temperature monitoring and enables early identification of potential thermal runaway risks in battery cells through forward-looking temperature extreme value prediction, thus ensuring the accuracy and timeliness of battery thermal runaway early warning.

[0078] S130: When the extreme temperature of the first cell is less than the thermal runaway temperature threshold, load the abnormal pressure timing information and cycle number of the target cell to construct a static profile of the cell; In this embodiment of the application, in order to overcome the limitations of traditional methods that rely solely on temperature monitoring, and to improve the characterization of the cell state by supplementing the cell pressure history and usage loss information, it is necessary to load the abnormal pressure timing information and cycle number of the target cell and construct a static profile of the cell, so as to achieve a refined assessment of the long-term state of the cell and provide comprehensive data support for subsequent risk coefficient calculation.

[0079] Specifically, the pressure monitoring value and corresponding pressure monitoring timestamp of the target cell are first collected from the pressure sensor deployed in the battery pack. The pressure sensor needs to monitor the pressure changes of the cell in real time during use, ensuring that each pressure data point is accurately linked to a specific time.

[0080] Furthermore, the collected pressure monitoring values ​​are compared with predefined pressure thresholds. When the pressure monitoring value is greater than or equal to the pressure threshold, it indicates that the cell has experienced pressure exceeding the normal range, and the pressure monitoring value and the corresponding pressure monitoring timestamp should be added to the abnormal pressure timing information.

[0081] Conversely, if the pressure monitoring value is less than the pressure threshold, it will not be recorded to ensure that the abnormal pressure timing information only includes pressure data that has a potential impact on the cell status, thus avoiding invalid information from interfering with subsequent analysis.

[0082] Simultaneously, the cycle count of the target cell is extracted from the battery management system (BMS). The BMS records the cell's charge-discharge cycle status in real time, and the extracted cycle count must accurately reflect the cell's current level of wear and tear, providing a basis for determining whether the cell's performance has deteriorated due to long-term use, thereby increasing the risk of thermal runaway.

[0083] After acquiring the abnormal pressure timing information and cycle count, the two types of information are integrated to construct a static profile of the battery cell. This profile needs to clearly present the target battery cell's past abnormal pressure records (including the pressure value and time of each abnormal pressure) and the current cycle count, forming a static characterization of the battery cell's long-term state. This complements the previously presented dynamic profile of the battery cell reflecting its real-time state, and together they lay the foundation for subsequent retrieval of sample sets of battery cells of the same model and calculation of thermal runaway risk coefficients.

[0084] Step S130 in the method provided in this application embodiment includes: Pressure monitoring values ​​and pressure monitoring timestamps are collected from pressure sensors deployed in the target cells of the battery pack. When the pressure monitoring value is greater than or equal to the pressure threshold, the pressure monitoring value and the pressure monitoring timestamp are added to the abnormal pressure timing information; The number of cycles of the target cell is extracted from the battery BMS system.

[0085] In this embodiment of the application, in order to further obtain information on the long-term damage risk and usage loss of the target cell when the extreme temperature of the first cell does not reach the thermal runaway threshold, it is necessary to collect pressure data, screen abnormal pressure records and extract the number of cycles to provide key data for constructing a static profile of the cell.

[0086] Specifically, the pressure monitoring value and corresponding pressure monitoring timestamp of the target cell are first collected from the pressure sensors deployed in the battery pack. Each pressure sensor must be deployed in a one-to-one correspondence with a target cell to ensure that the collected pressure data accurately reflects the pressure condition of that cell during use, rather than the pressure status of other cells or the battery pack as a whole.

[0087] Meanwhile, the pressure monitoring timestamp needs to be accurate to the second to fully record the time node of each pressure change, providing a time dimension reference for tracing the occurrence period of abnormal pressure and analyzing the impact of pressure on the long-term performance of the battery cell.

[0088] Furthermore, after acquiring the pressure monitoring value and timestamp, the pressure monitoring value is compared with a predefined pressure threshold. If the pressure monitoring value is greater than or equal to the pressure threshold, it indicates that the pressure experienced by the cell has exceeded the safe range, which may lead to internal structural damage and increase the risk of thermal runaway. In this case, the current pressure monitoring value and the corresponding pressure monitoring timestamp need to be added to the abnormal pressure timing information to form an abnormal pressure record arranged in chronological order.

[0089] Conversely, if the pressure monitoring value is less than the pressure threshold, it is determined to be under normal pressure and is not recorded. This ensures that only key data that poses a potential risk to the cell status is retained in the abnormal pressure timing information, avoiding invalid data from occupying storage resources or interfering with subsequent analysis.

[0090] Simultaneously, the cycle count of the target cell is extracted from the battery management system (BMS). The BMS tracks and records the charge-discharge cycle process of each cell in real time, counting one cycle each time a complete charge-discharge cycle (from fully charged to the discharge cutoff voltage, and then recharged to the charging cutoff voltage) is completed. The extracted cycle count must be bound to the unique identifier of the target cell to ensure it is not confused with the cycle counts of other cells.

[0091] The number of cycles directly reflects the degree of wear and tear on the battery cell. The more cycles, the more obvious the degradation of the active materials inside the battery cell, and the higher the risk of thermal runaway. Therefore, it is a key indicator for assessing the aging status of the battery cell.

[0092] In practical applications, it is essential to ensure the synchronization of pressure data acquisition and cycle count extraction. For example, after a target battery cell's pressure monitoring value reaches the pressure threshold at 14:30 and is recorded in the abnormal pressure timing information, the cumulative cycle count of that battery cell at 14:30 must be extracted from the BMS system simultaneously. This establishes a correlation between the abnormal pressure record and the current cycle count, facilitating subsequent analysis of the impact of "abnormal pressure during a specific cycle stage" on the risk of thermal runaway in the battery cell.

[0093] In addition, the validity of the data must be verified during the data acquisition process. If a pressure sensor malfunctions, resulting in empty pressure monitoring values ​​or significant fluctuations, a sensor fault alarm must be triggered and a backup pressure sensor must be activated to re-acquire data.

[0094] Meanwhile, if there is a data delay or error when extracting the cycle count from the BMS system, the most recent valid cycle count record needs to be retrieved through the system data backup module, and the abnormal data status needs to be marked. The data will be updated after the system is restored to ensure that the collected pressure data and cycle count are reliable, laying the foundation for building an accurate static profile of the battery cell.

[0095] Finally, the collected abnormal pressure timing information and cycle count are integrated, and the abnormal pressure timing information and the extracted cycle count are associated and bound to form a dataset that can reflect the long-term pressure damage record and usage wear status of the target battery cell. Based on this dataset, a static profile of the battery cell is constructed to comprehensively present the long-term status information of the battery cell.

[0096] For example, for a target battery cell, the pressure monitoring value collected from the pressure sensor at 10:20 on May 10, 2025 is 9 kPa (greater than the pressure threshold of 8 kPa), and the pressure monitoring value at 14:35 on May 15, 2025 is 10 kPa (greater than the pressure threshold of 8 kPa). These two data are sorted by time to form abnormal pressure timing information.

[0097] Meanwhile, the battery cell's cycle count on May 15, 2025, was extracted from the BMS system and found to be 720. The static profile of the battery cell constructed by integrating the two data clearly shows that "the battery cell has two abnormal pressure records and the current cycle count is 720", providing a basis for subsequent risk assessment.

[0098] In the method provided in this application embodiment, when the abnormal pressure timing information is empty and the number of loops is less than or equal to the number of loops threshold, the monitoring program is returned.

[0099] In this embodiment of the application, when the abnormal pressure timing information is empty, it means that all pressure monitoring values ​​collected from the pressure sensor of the target cell are less than the predefined pressure threshold. That is, the cell has never experienced a pressure situation exceeding the safe range during use, thus eliminating the possibility of thermal runaway caused by pressure damage.

[0100] Meanwhile, if the number of cycles is less than or equal to the cycle number threshold, it indicates that the current wear and tear of the battery cell has not exceeded the safety boundary, the degradation of the internal active materials is within the normal range, and there is no risk of thermal runaway due to excessive aging.

[0101] Therefore, considering both conditions, it can be determined that the target cell is currently in a healthy state, and there is no need to further perform risk coefficient calculation and graded early warning process. Therefore, the monitoring program is returned to continuously collect real-time data such as cell temperature and pressure to dynamically track its status changes. Once abnormal pressure or cycle count exceeds the threshold occurs, the thermal runaway risk assessment process is restarted to ensure cell safety while avoiding unnecessary consumption of computing resources.

[0102] The method provided in this application embodiment further includes: when the extreme temperature of the first cell is greater than the thermal runaway temperature threshold, configuring the thermal runaway risk coefficient of the target cell to 1.

[0103] In this embodiment of the application, when the extreme temperature of the first cell is greater than the thermal runaway temperature threshold, it means that the temperature of the target cell has exceeded the critical range for safe operation, and the internal chemical reaction is very likely to enter a runaway state, and the risk of thermal runaway has reached the highest level.

[0104] At this point, the thermal runaway risk coefficient of the target cell is configured to 1. Through clear numerical judgment (risk coefficient 1 represents the highest level of risk), the emergency risk state of the cell is directly defined, avoiding delays in early warning due to complex sample retrieval and proportion statistics processes. This ensures that the highest level of early warning response can be triggered as soon as possible, providing a clear basis for subsequent rapid emergency measures such as shutdown and cooling, and minimizing the probability of thermal runaway accidents.

[0105] S140: When the abnormal pressure timing information is not empty, or / and the number of cycles is greater than the number of cycles threshold, retrieve a sample set of the same model of battery cells that satisfy the static profile and the dynamic profile of the battery cell; In this embodiment of the application, in order to avoid the one-sidedness of risk judgment caused by relying solely on the data of the target cell itself, it is necessary to retrieve a sample set of cells of the same model that meet the conditions of static cell profile and dynamic cell profile, so as to construct a comparable sample benchmark and provide data support for subsequent calculation of thermal runaway risk coefficient.

[0106] Specifically, the criteria for determining the trigger conditions for the retrieval are first clarified. Among them, "abnormal pressure timing information is not empty" means that there is at least one record in the historical data collected from the pressure sensor of the target cell where the pressure monitoring value is greater than or equal to the pressure threshold, indicating that the cell has experienced pressure beyond the safe range and may have internal structural damage; "the number of cycles is greater than the cycle number threshold" means that the current usage loss of the cell has exceeded the preset safety boundary, and the decay of internal active materials may lead to a decrease in thermal stability. As long as either of these two conditions is met, or both conditions are met at the same time, the sample set retrieval step must be initiated.

[0107] Furthermore, the dual-profile constraints required for the retrieval are determined. The static profile of the battery cell includes the abnormal pressure timing information and cycle count of the target cell, reflecting the damage and wear accumulated during long-term use. The dynamic profile covers the target cell's current pre-control parameters and real-time ambient temperature, reflecting the cell's current operating conditions and external environmental influences. These two profiles together constitute the core screening criteria for the retrieval, ensuring that the retrieved sample cells are highly similar to the target cell in both "long-term state" and "real-time operating conditions," thereby guaranteeing the reference value for subsequent risk assessment.

[0108] Furthermore, after clarifying the constraints, a search for a sample set of battery cells of the same model is conducted. Specifically, firstly, based on the preset battery cell model, samples that are completely consistent with the target battery cell model are initially selected from the battery cell sample database.

[0109] Furthermore, a secondary screening is performed based on the static profile of the battery cell. For example, if the target battery cell has two abnormal pressure records and the current cycle count is 750 (greater than the cycle count threshold of 700), then samples of the same model that also have abnormal pressure records (unlimited number of times) or cycle counts greater than 700 are selected.

[0110] Finally, the cells were screened three times based on their dynamic profiles. Samples with a pre-control parameter error of less than ±5% and a current ambient temperature difference of less than ±3℃ were matched with the target cells to exclude samples that were interfered with by excessive differences in control parameters or ambient temperature.

[0111] During the retrieval process, it is necessary to ensure the completeness and timeliness of the sample database. The sample database needs to continuously include data on the same model of battery cells under different usage stages and operating environments, including samples in normal operation and samples with thermal runaway failures, and regularly update the sample information, removing failed or duplicate sample data.

[0112] For example, for a target cell with the preset model "lithium iron phosphate - square 50Ah", its static cell profile shows that there is one abnormal pressure record with a pressure value of 9kPa (pressure threshold is 8kPa) and a cycle count of 820 (cycle count threshold is 800). The dynamic cell profile shows that the current pre-controlled parameters are rated voltage 3.2V, rated capacity 50Ah, and ambient temperature 26℃.

[0113] During the search, all cell samples of the "lithium iron phosphate - square 50Ah" model were first screened out. Then, samples with "abnormal pressure records or cycle count > 800" were selected. Finally, samples with "pre-control parameters 3.2V±5%, 50Ah±5%, ambient temperature 26℃±3℃" were matched to form a sample set of 120 cells of the same model, including 15 thermal runaway failure samples and 105 normal operation samples.

[0114] In addition, the validity of the retrieved sample set of battery cells of the same model needs to be verified. The verification includes the completeness, authenticity, and relevance of the sample data. If invalid samples are found, they need to be removed from the sample set in a timely manner to ensure the high quality and reliability of the sample set, and to lay an accurate data foundation for subsequent statistical analysis of the proportion of thermal runaway samples and calculation of the thermal runaway risk coefficient of the target battery cell.

[0115] S150: The proportion of thermal runaway samples in the same type of battery cell sample set is set as the thermal runaway risk coefficient of the target battery cell; In this embodiment of the application, in order to avoid relying on subjective experience to judge the risk level of thermal runaway of battery cells, it is necessary to statistically analyze the proportion of thermal runaway samples of the same type of battery cells and set it as the thermal runaway risk coefficient of the target battery cell, so as to achieve accurate quantification of the thermal runaway risk of the target battery cell and provide numerical basis for the subsequent implementation of the graded early warning strategy.

[0116] Specifically, the first step is to distinguish between thermal runaway samples and non-thermal runaway samples within the same cell sample set. Thermal runaway samples must meet the condition that they have experienced thermal runaway phenomena such as sudden temperature rise and smoke, and have complete fault records. Non-thermal runaway samples are normally operating or compliantly retired samples that have not shown any signs of thermal runaway. During classification, the historical operating logs and fault reports of each sample must be checked one by one to ensure accurate classification.

[0117] Furthermore, the thermal runaway risk coefficient is calculated using the formula: "Thermal runaway risk coefficient = Number of thermal runaway samples / Total number of valid samples in the sample set". For example, a sample set of a certain type of battery cell contains 120 valid samples, of which 9 are thermal runaway samples. The calculated thermal runaway risk coefficient of the battery cell is 0.075 (i.e., 9 / 120 = 7.5%). This value directly reflects the thermal runaway risk level of the target battery cell.

[0118] Simultaneously, basic verification of the calculation results is required. If the sample set is too small, the reference value of the results needs to be evaluated, and the search scope should be expanded to supplement the samples if necessary. If thermal runaway samples are concentrated under specific conditions (such as the number of cycles far exceeding the cycle threshold), the thermal runaway risk coefficient needs to be adjusted in combination with the actual state of the target cell to ensure that the quantification results can match the real risk situation of the target cell and provide reliable support for subsequent early warning.

[0119] S160: Based on the thermal runaway risk coefficient of the target battery cell, execute a graded early warning strategy with the corresponding coefficient.

[0120] In this embodiment of the application, in order to avoid over- or under-response caused by using a uniform early warning method for cells with different risk levels, it is necessary to implement a graded early warning strategy by loading a predefined risk threshold and combining it with the risk coefficient of the target cell, so as to achieve precise control over thermal runaway risk and balance safety assurance and usage efficiency.

[0121] Specifically, the system first loads predefined first and second risk coefficient thresholds. The second risk coefficient threshold is higher than the first risk coefficient threshold. These two thresholds are determined based on a large amount of thermal runaway fault data and safe operation experience of the same type of battery cells, thereby dividing the system into low, medium, and high risk ranges and providing clear boundaries for subsequent graded responses.

[0122] Furthermore, the thermal runaway risk coefficient of the target battery cell is compared with the first and second risk coefficient thresholds to match the corresponding early warning strategy.

[0123] Specifically, if the thermal runaway risk coefficient of the target battery cell is less than or equal to the first risk coefficient threshold, it means that the current thermal runaway risk of the battery cell is low and there is no need to suspend its use. Only a potential thermal runaway risk warning message needs to be sent to remind staff to pay attention to the subsequent changes in the status of the battery cell.

[0124] In addition, if the thermal runaway risk coefficient of the target cell is greater than the first risk coefficient threshold but less than the second risk coefficient threshold, it indicates that the cell has a certain potential for thermal runaway. A cell inspection prompt needs to be generated, and potential problems need to be investigated through professional inspection.

[0125] In addition, if the thermal runaway risk coefficient of the target cell is greater than or equal to the second risk coefficient threshold, it means that the thermal runaway risk of the cell has reached a high level. An operation prohibition warning should be generated immediately to forcibly stop the operation of the battery system where the cell is located. At the same time, an audible and visual alarm should be issued to notify personnel to evacuate, so as to minimize the risk of accident.

[0126] This step links the quantified target cell thermal runaway risk coefficient with a tiered early warning strategy, making the early warning measures more aligned with the actual risk situation of the cells. This ensures safety while avoiding excessive intervention in low-risk cells that could affect normal use, thus improving the overall flexibility and accuracy of battery management.

[0127] As attached Figure 2 As shown, step S160 in the method provided in this application embodiment includes: Load a predefined first risk coefficient threshold and a second risk coefficient threshold, wherein the second risk coefficient threshold is greater than the first risk coefficient threshold; When the thermal runaway risk coefficient of the target battery cell is less than or equal to the first risk coefficient threshold, a potential thermal runaway risk warning message is sent. When the thermal runaway risk coefficient of the target battery cell is greater than the first risk coefficient threshold and less than the second risk coefficient threshold, a battery cell maintenance prompt is generated. When the thermal runaway risk coefficient of the target battery cell is greater than or equal to the second risk coefficient threshold, a warning to prohibit operation is generated.

[0128] In this embodiment of the application, in order to take differentiated countermeasures according to the different degrees of thermal runaway risk of the target battery cell, and to avoid excessive intervention in low-risk battery cells or insufficient response to high-risk battery cells, which may lead to safety hazards or waste of resources, it is necessary to divide the risk level by loading a predefined risk coefficient threshold and match the corresponding graded early warning strategy in order to achieve precise control of the thermal runaway risk of the battery cell.

[0129] Specifically, firstly, predefined first risk coefficient threshold and second risk coefficient threshold are loaded.

[0130] The first and second risk coefficient thresholds are determined based on a large amount of thermal runaway fault data of the same type of battery cells, long-term operating experience, and safety standards. For example, based on the historical fault statistics of a certain type of ternary lithium battery, it was found that when the risk coefficient is ≤0.05 (5%), the probability of thermal runaway of the battery cell is extremely low; when the risk coefficient is between 0.05 and 0.15 (5%-15%), the battery cell has potential hidden dangers but does not affect emergency use for the time being; when the risk coefficient is ≥0.15 (15%), the risk of thermal runaway of the battery cell increases significantly. Therefore, the first risk coefficient threshold is set to 0.05 and the second risk coefficient threshold is set to 0.15, and the second risk coefficient threshold is ensured to be greater than the first risk coefficient threshold, so as to form a clear risk level division boundary.

[0131] Furthermore, the thermal runaway risk coefficient of the target battery cell is compared with the first and second risk coefficient thresholds, and then the corresponding early warning strategy is executed.

[0132] Specifically, when the thermal runaway risk coefficient of the target battery cell is less than or equal to the first risk coefficient threshold, it indicates that the battery cell is currently in a low-risk state and there is no need to suspend its use or disassemble it for maintenance. At this time, a potential thermal runaway risk warning message can be sent.

[0133] For example, the battery management system (BMS) can push text or pop-up notifications to the terminal devices of maintenance personnel, with the content including "The target cell has a low risk of thermal runaway (risk coefficient 0.03), and it is recommended to strengthen temperature and pressure data monitoring every 24 hours". This reminds staff to pay attention to the status of the cell while avoiding excessive intervention that may affect normal operations.

[0134] Conversely, when the thermal runaway risk coefficient of the target battery cell is greater than the first risk coefficient threshold but less than the second risk coefficient threshold, it indicates that the battery cell already has a certain potential for thermal runaway. If not addressed promptly, the risk may escalate. In this case, a battery cell maintenance prompt should be generated. The maintenance prompt must clearly specify the specific maintenance requirements, such as noting "The target battery cell has a medium risk of thermal runaway (risk coefficient 0.12), and requires completion of battery cell appearance inspection, internal resistance testing, and charge / discharge performance evaluation within 72 hours." The prompt should also automatically assign maintenance work orders to the corresponding maintenance team to ensure that potential hazards can be identified and addressed within a controllable timeframe, preventing further escalation of the risk.

[0135] Conversely, when the thermal runaway risk coefficient of the target cell is greater than or equal to the second risk coefficient threshold, it means that the thermal runaway risk of the cell has reached a high level and a safety accident may occur at any time. At this time, a warning to prohibit operation must be generated.

[0136] Among them, the warning of prohibited operation must have mandatory intervention, such as immediately cutting off the charging and discharging circuit of the battery module where the cell is located, suspending the operation of related equipment, and notifying on-site personnel to evacuate to a safe area through sound and light alarms, emergency text messages, etc., while coordinating with fire emergency devices to prepare, so as to minimize the casualties and property losses caused by thermal runaway accidents.

[0137] For example, for a certain type of lithium iron phosphate square cell, the first risk coefficient threshold is 0.04, and the second risk coefficient threshold is 0.12. If the thermal runaway risk coefficient of a target cell is 0.03 (≤0.04), a prompt message of "suggesting enhanced daily monitoring" is sent; if the risk coefficient is 0.08 (0.04 < 0.08 < 0.12), a work order of "complete maintenance within 48 hours" is generated; if the risk coefficient is 0.15 (≥0.12), a prohibition on operation warning is immediately triggered, the circuit is cut off, and an alarm is issued.

[0138] Simultaneously, when implementing a tiered early warning strategy, it is also necessary to record and trace the early warning information. For example, a complete early warning file should be formed, including the unique identifier of the target cell associated with each early warning, the basis for calculating the risk coefficient, the early warning trigger time, and the subsequent processing results. This will provide data support for adjusting the early warning strategy for the same type of cell, ensuring that the implementation of the tiered early warning strategy is scientific and adaptable, thereby continuously improving the accuracy and reliability of battery thermal runaway early warning.

[0139] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: This application proposes an intelligent hierarchical early warning method for battery thermal runaway. First, pre-control parameters and ambient temperature of the target battery cell are loaded. A dynamic profile of the battery cell is constructed through data synchronization, integration, and validity verification, accurately depicting the real-time operating status of the cell. Next, the dynamic profile is input into a temperature extreme value prediction model trained with data from cells with no pressure damage and low cycle loss, yielding a first extreme temperature value for the battery cell. If this first extreme temperature value is greater than the thermal runaway temperature threshold, the thermal runaway risk coefficient is directly configured to 1. If it is less, abnormal pressure timing information and cycle count are loaded to construct a static profile of the battery cell, presenting its long-term damage and loss status. When the battery cell has abnormal pressure records or the cycle count exceeds the threshold, a sample set of the same model of battery cells matching the dynamic and static profiles is retrieved, and the proportion of thermal runaway samples is statistically analyzed as the thermal runaway risk coefficient of the target battery cell. Finally, predefined first and second risk coefficient thresholds are loaded, and potential risk warnings, maintenance prompts, or prohibited operation warnings are sent according to the magnitude of the risk coefficient, respectively, to achieve differentiated management of thermal runaway risk.

[0140] The method provided in this application, through the technical solution of "real-time status analysis of cell dynamic profiling - temperature extreme value prediction and judgment - long-term status analysis of cell static profiling - sample set retrieval and risk quantification - graded early warning and differentiated management", solves the problems of traditional thermal runaway management, which overemphasizes the overall status of the battery pack while neglecting the detailed analysis of individual cells and only issues early warnings when the temperature is abnormal, resulting in serious lag. It realizes a full-dimensional assessment from the real-time status of the cells to long-term risks, providing reliable technical support for the safe operation and efficient maintenance of battery systems, effectively balancing safety assurance and usage efficiency, and reducing the probability of thermal runaway accidents.

[0141] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0142] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0143] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for intelligent graded early warning of battery thermal runaway, characterized in that, include: Load the pre-control parameters and ambient temperature of the target battery cell to construct a dynamic profile of the battery cell; The cell temperature extreme value prediction model is used to process the cell dynamic profile to obtain the first cell temperature extreme value. The cell temperature extreme value prediction model is obtained by training the cell dynamic historical profile set with zero abnormal pressure and a cycle number less than or equal to the cycle number threshold and the cell temperature extreme value detection set using machine learning. When the extreme temperature of the first cell is less than the thermal runaway temperature threshold, the abnormal pressure timing information and cycle number of the target cell are loaded to construct a static profile of the cell. When the abnormal pressure timing information is not empty, or / and the number of cycles is greater than the number of cycles threshold, retrieve a sample set of the same type of battery cell that satisfies the static profile and the dynamic profile of the battery cell; The proportion of thermal runaway samples in the same type of battery cell sample set is set as the thermal runaway risk coefficient of the target battery cell; Based on the thermal runaway risk coefficient of the target battery cell, a graded early warning strategy corresponding to the coefficient is implemented.

2. The method as described in claim 1, characterized in that, Also includes: When the extreme temperature of the first cell is greater than the thermal runaway temperature threshold, the thermal runaway risk coefficient of the target cell is configured to 1.

3. The method as described in claim 1, characterized in that, The cell temperature extreme value prediction model is obtained by training a set of dynamic historical images of cells with zero abnormal pressure cycles and a cycle count less than or equal to a cycle count threshold, and a set of cell temperature extreme value detections using machine learning. This includes: Obtain the preset battery cell model; Input the preset cell model into the threshold calibration table and match it with the predefined pressure threshold and cycle number threshold. Using the preset cell model as a constraint, collect the dynamic historical profile of the first cell with zero abnormal pressure counts and a cycle count less than or equal to the cycle count threshold. Abnormal pressure counts indicate that the number of times the pressure monitoring value is greater than or equal to the pressure threshold is zero. Based on the preset cell model and the dynamic historical profile of the first cell, an initial set of cell temperature extreme value detection values ​​is collected, where the number of abnormal pressure tests is zero and the number of cycles is less than or equal to the cycle number threshold. Central trend analysis is then performed to obtain the first cell temperature extreme value detection value. Add the first cell dynamic historical image to the cell dynamic historical image set, and add the first cell temperature extreme value detection value to the cell temperature extreme value detection value set.

4. The method as described in claim 3, characterized in that, The steps for constructing the threshold calibration table include: The initial pressure threshold is set to zero, and the number of cycles is set to j, where j represents a positive integer, the initial value of j is equal to 1, and the maximum value of j is equal to the rated number of cycles. Using the preset cell model and preset cell dynamic profile as constraints, a set of first cell temperature extreme value detection values ​​is collected where the number of abnormal pressures is zero and the number of cycles is less than or equal to the number of cycles threshold. Central tendency analysis is then performed to obtain the first fitted value of the cell temperature extreme value. Using the preset cell model and preset cell dynamic profile as constraints, a set of second cell temperature extreme value detection values ​​is collected where the number of abnormal pressure cycles is zero and the number of cycles is equal to j+1 cycles. Central tendency analysis is then performed to obtain the second fitted value of the cell temperature extreme value. When the first temperature extreme value deviation between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is greater than or equal to the temperature extreme value deviation threshold, a pressure threshold is configured based on the cycle number threshold, and the cycle number threshold, the pressure threshold, and the preset cell model are associated and stored, and added to the threshold calibration table. When the deviation between the second fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is less than the temperature extreme value deviation threshold, j is incremented by one, and the loop is executed.

5. The method as described in claim 4, characterized in that, Based on the aforementioned cycle count threshold, a pressure threshold is configured, including: Loading pressure consistency deviation, wherein the pressure consistency deviation is a predefined pressure tolerance deviation; The characteristic pressure is obtained by summing the pressure consistency deviation with the initial pressure threshold. Using the preset cell model and preset cell dynamic profile as constraints, a third set of cell temperature extreme value detection values ​​is collected, where the number of abnormal pressures is not zero, the pressures of the abnormal pressures are all less than the characteristic pressures, and the number of cycles is less than or equal to the number of cycles threshold. Central tendency analysis is then performed to obtain the third fitted value of the cell temperature extreme value. When the deviation between the third fitted value of the cell temperature extreme value and the first fitted value of the cell temperature extreme value is greater than or equal to the temperature extreme value deviation threshold, the initial pressure threshold is set as the pressure threshold. Otherwise, when the deviation between the third fitted value of the cell temperature extreme value and the second temperature extreme value of the first fitted value of the cell temperature extreme value is less than the temperature extreme value deviation threshold, the initial pressure threshold is updated using the characteristic pressure, and the loop is executed.

6. The method as described in claim 1, characterized in that, When the abnormal pressure timing information is empty and the number of loops is less than or equal to the number of loops threshold, return to the monitoring program.

7. The method as described in claim 1, characterized in that, Load the abnormal pressure timing information and cycle count of the target battery cell to construct a static profile of the battery cell, including: Pressure monitoring values ​​and pressure monitoring timestamps are collected from pressure sensors deployed in the target cells of the battery pack. When the pressure monitoring value is greater than or equal to the pressure threshold, the pressure monitoring value and the pressure monitoring timestamp are added to the abnormal pressure timing information; The number of cycles of the target cell is extracted from the battery BMS system.

8. The method as described in claim 1, characterized in that, Based on the thermal runaway risk coefficient of the target battery cell, a graded early warning strategy corresponding to the coefficient is implemented, including: Load a predefined first risk coefficient threshold and a second risk coefficient threshold, wherein the second risk coefficient threshold is greater than the first risk coefficient threshold; When the thermal runaway risk coefficient of the target battery cell is less than or equal to the first risk coefficient threshold, a potential thermal runaway risk warning message is sent. When the thermal runaway risk coefficient of the target battery cell is greater than the first risk coefficient threshold and less than the second risk coefficient threshold, a battery cell maintenance prompt is generated. When the thermal runaway risk coefficient of the target battery cell is greater than or equal to the second risk coefficient threshold, a warning to prohibit operation is generated.

Images