Battery thermal runaway early warning method and system based on swelling force trend perception

Abstract

The application belongs to the field of lithium ion battery safety monitoring, and provides a battery thermal runaway early warning method and system based on expansion force trend perception, which synchronously acquires battery expansion force, voltage, surface temperature multi-source time sequence signals, and carries out standardized pretreatment on the multi-source time sequence signals; the pretreated signals are divided into multiple data windows by adaptive dynamic sliding window, expansion force multi-dimensional trend characteristics of each data window are extracted by adopting second-order polynomial fitting, a comprehensive trend score is constructed based on the expansion force multi-dimensional trend characteristics; the median absolute deviation is calculated according to the median of the internal expansion force change rate sequence, and the real-time dynamic warning threshold is generated in combination with the signal baseline mean and the sensitivity coefficient; abnormality is determined by comparing the comprehensive trend score, the expansion force change rate and the dynamic warning threshold, and the voltage constraint is used, and two-step continuous confirmation mechanism is adopted to realize graded thermal runaway warning. The application guarantees the warning accuracy.

Description

Technical Field

[0001] This invention belongs to the field of lithium-ion battery safety monitoring, specifically relating to an early warning method and system for battery thermal runaway based on expansion force trend sensing. Background Technology

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

[0003] Lithium-ion batteries are widely used in new energy vehicles and electrochemical energy storage due to their high energy density and long cycle life. However, lithium batteries can undergo a series of side reactions under overcharging, high temperature, and fast charging abuse, resulting in internal short circuits, heat generation, positive feedback, and rapid fire and explosion.

[0004] Current mainstream battery management systems (BMS) rely on temperature and voltage as monitoring signals. In recent years, gas sensing and single expansion force threshold early warning solutions have emerged, but all of these solutions have significant shortcomings: The temperature monitoring scheme suffers from multiple layers of thermal conduction lag in the surface temperature. During the stages of gas generation and electrode expansion inside the battery, the surface temperature does not rise significantly. Only after severe internal heat release does the temperature rise sharply, and by the time an alarm is triggered, thermal runaway is imminent, resulting in an extremely short warning window.

[0005] The voltage monitoring scheme only detects voltage anomalies after lithium dendrites pierce the separator and form an internal short circuit. It cannot capture precursory characteristics such as gas production and expansion in the early stages of thermal runaway and has no early warning capability.

[0006] Existing mainstream mechanical early warning schemes are affected by battery capacity, charging rate, ambient temperature, and module preload. Fixed thresholds cannot adapt to multiple operating conditions, resulting in frequent false alarms or missed alarms under high temperature / fast charging conditions. They cannot distinguish between reversible expansion due to the "breathing effect" of normal charging and discharging and irreversible continuous expansion due to faults. High-rate charging can easily trigger false alarms. They use mean and variance statistical thresholds, which are sensitive to sensor electromagnetic noise, mechanical vibration, and signal glitches. Outliers directly cause threshold shifts. All of them use fixed-length sliding windows, which result in slow response under stable operating conditions and statistical distortion under drastic fluctuation conditions.

[0007] Existing mechanical early warning algorithms only identify single-point values ​​of the instantaneous rate of change of expansion force, failing to explore the nonlinear trend characteristics of the continuous increase of expansion force; they also fail to distinguish between the rate of change of expansion and the evolution law of acceleration, resulting in a significant reduction in the early warning lead time for high-rate, high-temperature, and pouch battery scenarios, leading to poor engineering practicality. Summary of the Invention

[0008] To address the aforementioned problems, this invention proposes an early warning method and system for battery thermal runaway based on expansion force trend perception. This invention can promptly capture early signs of internal gas generation and electrode expansion in the early stages of thermal runaway, reducing false alarm rates and ensuring the accuracy of early warnings.

[0009] According to some embodiments, the present invention adopts the following technical solution: An early warning method for battery thermal runaway based on expansion force trend sensing includes the following steps: Simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals; The preprocessed signal is divided into multiple data windows by adaptive dynamic sliding window. Second-order polynomial fitting is used to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, a comprehensive trend score is constructed by fusion. Based on the median of the internal expansion force change rate sequence, the median absolute deviation is calculated, and a real-time dynamic early warning threshold is generated by combining the signal baseline mean and sensitivity coefficient. Anomaly determination is made by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and a two-step confirmation mechanism is adopted to realize graded thermal runaway early warning.

[0010] As an alternative implementation, the process of simultaneously acquiring multi-source time-series signals of battery expansion force, voltage, and surface temperature includes: using a tension / compression sensor arranged between the clamp and the battery to acquire the preload applied by the clamp; The surface temperature of the battery is obtained using a temperature sensor. Use voltage acquisition equipment to obtain the battery terminal voltage.

[0011] As an alternative implementation, the process of standardizing and preprocessing multi-source time-series signals includes: time synchronization calibration, missing value completion, outlier removal, and filtering, smoothing, and noise reduction.

[0012] As an alternative implementation, the process of adaptively and dynamically dividing the preprocessed signal into multiple data windows includes: calculating the coefficient of variation of the expansion force data within the window, wherein the coefficient of variation is the standard deviation of the window data / the mean of the window data; when the mean of the window approaches zero, the coefficient of variation is set to zero; and automatically adjusting the window length according to the degree of fluctuation of the coefficient of variation, wherein the smaller the coefficient of variation, the smaller the window length; and configuring independent adaptive sliding windows for the rate of change of expansion force and the trend acceleration of expansion force respectively.

[0013] As an alternative implementation method, a second-order polynomial is used to extract the multidimensional trend features of expansion force in each data window. Based on the multidimensional trend features of expansion force, the process of integrating and constructing a comprehensive trend score includes: traversing the expansion force sequence of first to fourth-order polynomials, and judging by the three indicators of goodness of fit, root mean square error and random distribution of residuals. Among them, the first-order underfit and the third / fourth-order overfit noise determine the second-order polynomial as the optimal model. The coefficients of the second-order polynomial are solved by the least squares method. The first derivative of the polynomial is obtained to get the trend slope. The trend acceleration is obtained by the time-series difference of the slope. The goodness of fit is calculated simultaneously. Weights are assigned to the trend slope, goodness of fit, and trend acceleration. A comprehensive trend score is constructed by weighted fusion. The slope is normalized to the range of 0 to 1. Only when the slope is greater than a predetermined value is it determined to be a valid fault upward trend.

[0014] As an alternative implementation method, the process of calculating the median absolute deviation based on the median of the internal expansion force change rate sequence and generating a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient includes: calculating the median of the expansion force change rate sequence within the window, calculating the median absolute deviation, introducing a standard conversion factor to convert it into a robust standard deviation estimate; and generating a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient.

[0015] As a further defined implementation, the threshold is refreshed synchronously for each set of data updated in the window, automatically adapting to changes in capacity, temperature, and multiplier.

[0016] As an alternative implementation method, the process of determining anomalies by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints includes: setting a graded threshold for the comprehensive trend score; based on the relationship between the comprehensive trend score and the graded threshold, the comparison result between the rate of change of expansion force and the dynamic early warning threshold, and the voltage overcharge auxiliary constraint, when the comprehensive trend score meets the graded threshold, the rate of change of expansion force exceeds the dynamic early warning threshold, and the voltage overcharge auxiliary constraint is met, it is determined to be an anomaly.

[0017] As an alternative implementation method, the process of implementing graded thermal runaway early warning using a continuous two-step confirmation mechanism includes: if the conditions are met in a single sampling, only a temporary early warning is marked and the number of early warnings is incremented by one; if the conditions are not met in the next moment, the number of early warnings is reset to zero; only when the judgment conditions are met in two consecutive sampling cycles is an effective thermal runaway early warning signal generated.

[0018] An early warning system for battery thermal runaway based on expansion force trend sensing, comprising: The signal acquisition and preprocessing module is configured to simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals. The comprehensive trend score calculation module is configured to adaptively and dynamically divide the preprocessed signal into multiple data windows by sliding windows, and use second-order polynomial fitting to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, the comprehensive trend score is constructed by fusion. The warning threshold dynamic generation module is configured to calculate the median absolute deviation based on the median of the internal expansion force change rate sequence, and generate a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient. The anomaly detection module is configured to perform anomaly detection by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and to implement graded thermal runaway early warning by adopting a continuous two-step confirmation mechanism.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes expansion force to directly respond to gas generation and electrode deformation inside the battery, eliminating defects such as temperature heat conduction and voltage short circuit hysteresis. It provides early warning under standard operating conditions, allowing sufficient safe handling time for the BMS to perform current reduction, power cut-off, active cooling, and battery isolation, which is superior to existing electrical and single mechanical warning schemes.

[0020] The MAD median statistics of this invention are not affected by sensor glitch, electromagnetic interference, or vibration outliers; coupled with a continuous two-step confirmation mechanism, the probability of false alarms is greatly reduced; and the classification accuracy under all operating conditions is high.

[0021] This invention relies on relative trend characteristics rather than absolute expansion force values ​​to automatically adapt the adaptive window to battery capacity, ambient temperature, charging rate, and packaging form; it is also compatible with two mainstream types of power batteries: square hard-shell and soft-pack.

[0022] This invention requires only a pressure sensor, thermocouple, and basic voltage acquisition hardware, without the need for expensive gas sensors or online impedance detection equipment; the polynomial fitting and MAD statistical calculations are simple and can be directly embedded into existing BMS controllers without replacing the hardware.

[0023] This invention can expand multi-point distributed pressure sensing arrays, temperature compensation algorithms, and voltage / gas multi-source signal fusion strategies to further improve early warning performance in extreme scenarios such as high-rate, high-temperature, and pouch batteries.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 The diagram shows the overall experimental principle and structure of one embodiment, where (a) is the overall experimental device diagram, (b) is the experimental battery structure setup, and (c) is the temperature measurement point setup. Figure 2 The curves showing the changes in voltage, temperature, expansion force, and CO2 concentration of a battery after overcharging under standard operating conditions are shown in one embodiment. Figure 3 The results of the experiment on a 32Ah square hard shell at 25°C and 1C in one embodiment are as follows: (a) is the warning time and classification accuracy, (b) is the expansion force time series, (c) is the abnormal trend score, and (d) is the expansion force change rate. Figure 4 The results of an experiment on a 30Ah square hard shell at 25°C and 1C in one embodiment are as follows: (a) is the warning time and classification accuracy, (b) is the expansion force time series, (c) is the abnormal trend score, and (d) is the expansion force change rate. Figure 5 The results of the experiment on a 32Ah square hard shell at 30°C and 3C in one embodiment are as follows: (a) is the warning time and classification accuracy, (b) is the expansion force time series, (c) is the abnormal trend score, and (d) is the expansion force change rate. Figure 6 The results of the experiment on a 32Ah square hard shell at 70°C and 3C in one embodiment are as follows: (a) is the warning time and classification accuracy, (b) is the expansion force time series, (c) is the abnormal trend score, and (d) is the expansion force change rate. Figure 7 The results of the experiment in one embodiment are 32Ah soft pack, 30℃, 3C, where (a) is the warning time and classification accuracy, (b) is the expansion force time series, (c) is the abnormal trend score, and (d) is the expansion force change rate. Figure 8 This is a flowchart illustrating one embodiment. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0030] Where there is no conflict, the embodiments and features described in this application may be combined with each other.

[0031] Example 1 An early warning method for battery thermal runaway based on expansion force trend sensing, such as Figure 8 As shown, it includes the following steps: Simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals; The preprocessed signal is divided into multiple data windows by adaptive dynamic sliding window. Second-order polynomial fitting is used to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, a comprehensive trend score is constructed by fusion. Based on the median of the internal expansion force change rate sequence, the median absolute deviation is calculated, and a real-time dynamic early warning threshold is generated by combining the signal baseline mean and sensitivity coefficient. Anomaly determination is made by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and a two-step confirmation mechanism is adopted to realize graded thermal runaway early warning.

[0032] In this embodiment, multi-source time-series signals of battery expansion force, voltage, and surface temperature are acquired simultaneously. During the standardization preprocessing of the multi-source time-series signals, a hardware monitoring platform for simultaneous acquisition of multiple physical quantities is built, such as... Figure 1 As shown: It mainly includes an explosion-proof test chamber, an adjustable pre-tightening battery clamp, a high-precision charging and discharging device, a constant temperature control system, a multi-channel synchronous data acquisition instrument, and a high-definition industrial camera; The sensor setup includes a G709 high-precision tensile and compressive sensor with a range of 0~20000N and a sampling frequency of 1Hz, positioned between the clamp and the battery. A mica heat insulation board is laid between the clamp and the battery to prevent heat conduction. The clamp has an adjustable initial preload of 0~20kN. After applying the preload, the clamp is left to stand for ≥1h until the expansion force fluctuation stabilizes within ±0.5%FS before data acquisition. Temperature sensing: Two K-type thermocouples are attached to the center of the large surface of the battery to collect the surface temperature; Voltage signal: The charging and discharging equipment synchronously acquires the battery terminal voltage; The test battery used in this embodiment is a lithium iron phosphate (LFP) battery, which includes two types of packaging (square hard case and soft case) and two types of capacity (30Ah and 32Ah). Test conditions covered: charging rate 1C / 3C, ambient temperature 25℃ / 30℃ / 70℃.

[0033] Next, we perform standardization preprocessing on the raw multi-source time series data.

[0034] The original signal contains time misalignment, missing values, electromagnetic glitches, and abrupt changes. A four-step preprocessing step is performed uniformly: (1) Time synchronization calibration: Based on the high-precision clock of the charging and discharging equipment, linear resampling is used to unify the sampling interval of all signals to 1s, eliminating the clock deviation of multiple devices; (2) Missing value completion: A small number of missing points are filled by linear interpolation to ensure temporal continuity; ; in t 1 and t 2 represents the nearest valid data moment before and after the missing point. x ( t 1) and x ( t 2) is the corresponding signal value.

[0035] (3) Outlier removal: The sliding window mean-standard deviation criterion is used to identify abrupt outliers, and the window mean is used to replace outliers to suppress single-point spikes; (4) Savitzky-Golay (SG) filtering smooths and reduces noise, retains the early slight upward trend of expansion force, and filters out high-frequency electromagnetic noise.

[0036] The process of adaptive dynamic sliding window partitioning: (1) Calculate the coefficient of variation (CV) of the expansion force data within the window = standard deviation of the window data. / Window data mean When the window mean approaches 0, set CV=0 directly to avoid calculation errors. ; (2) Automatically adjust the window length according to the degree of CV fluctuation: CV < 0.05: The signal is extremely stable, so the minimum window is used to improve the algorithm's response speed; 0.05≤CV<0.15: Slight fluctuations, slightly increase the window size; 0.15≤CV<0.30: Moderate fluctuation, moderate window; 0.30≤CV<0.50: Large fluctuations, requiring a significant increase in the window size; CV ≥ 0.50: Significant fluctuations; use the maximum window to ensure statistical stability. (3) Independent window processing: Independent adaptive sliding windows are configured for the rate of change of expansion force and the acceleration of expansion force trend, so as to avoid feature interference caused by sharing windows.

[0037] Perform second-order polynomial trend fitting to extract multidimensional trend features: (1) Optimal fitting order selection: The expansion force sequence is fitted by polynomials of orders 1 to 4. The goodness of fit R², root mean square error RMSE, and random distribution of residuals are used to determine the following: order 1 is underfitting, order 3 / 4 is overfitting noise, and order 2 is determined to be the optimal model. (2) Solve the coefficients of the second-order polynomial using the least squares method, and obtain the trend slope k (characterizing the instantaneous increase rate of expansion force) by taking the first derivative of the polynomial; obtain the trend acceleration a (characterizing the degree of accelerated deterioration of expansion force) by the time difference of the slope; and calculate the goodness of fit R² simultaneously. (3) Weighted fusion to construct a comprehensive trend score S (value 0~1): weight allocation: trend slope 50%, goodness of fit 30%, trend acceleration 20%; the slope is normalized to the 0~1 range, and only slope > 0.05 is judged as a valid fault upward trend; (4) Three-level trend score warning thresholds: S≥0.35 initial abnormality, S≥0.4 moderate abnormality, S≥0.5 strong abnormality.

[0038] Robust dynamic threshold calculation based on the absolute deviation of the median of MAD: A dynamic threshold is constructed for the rate of change of expansion force dF / dt within the window to avoid the defect that the mean and variance are sensitive to outliers: (1) Calculate the median Med of the expansion force change rate sequence within the window; (2) Calculate the median absolute deviation (MAD); ; x i For any data point, MAD is the expression for any data point. x i Calculate the absolute deviation of the median.

[0039] (3) Introduce standard conversion factors to convert to robust standard deviation estimates; (4) Combine the signal baseline mean (MED(x)) and sensitivity coefficient (k=1.8) to generate a real-time dynamic warning threshold Th; the threshold is refreshed synchronously for each set of data updated in the window, and automatically adapts to changes in capacity, temperature and multiplier.

[0040] ; Two-step early warning confirmation judgment mechanism: (1) Joint judgment conditions: The comprehensive trend score S reaches the grading threshold + the rate of change of expansion force exceeds the MAD dynamic threshold Th + voltage overcharge auxiliary constraint; Thermal runaway is induced by overcharging and abuse; SEI decomposition, lithium plating, gas production and expansion will not occur in the normal charging range; only when the battery enters the overcharge range will the internal side reaction be initiated, and only when the expansion force is abnormal will there be a risk of thermal runaway.

[0041] If the battery is not overcharged, even if there are slight fluctuations in expansion force (normal breathing effect, assembly stress drift, sensor noise), there is no risk of thermal runaway. Voltage constraint is used to filter out misjudgments caused by normal charging and discharging.

[0042] In this embodiment, the voltage-assisted constraint determination rule is: the real-time battery terminal voltage U ≥ the full charge cutoff voltage.

[0043] During normal charge and discharge (voltage < full charge cutoff voltage), the battery only undergoes reversible breathing expansion, without internal gas generation or irreversible lithium plating side reactions. At this time, even if the expansion force trend fraction or rate of change occasionally exceeds the threshold, the voltage constraint is not met, and the warning count is not triggered directly, completely avoiding a large number of false alarms caused by reversible expansion during normal charge and discharge, and solving the pain point of "false alarms during normal charging at high rates" in the existing fixed threshold scheme.

[0044] In this embodiment, an early warning is triggered when the following conditions are met: 1) The comprehensive trend score S reaches the graded early warning threshold (S≥0.35 initial anomaly / 0.4 moderate / 0.5 strong anomaly). 2) The current window expansion force change rate dF / dt > the real-time dynamic warning threshold Th calculated by MAD; 3) The overcharge auxiliary constraint condition is met.

[0045] (2) Timing-based false alarm prevention logic: If the condition is met in a single sampling, only a temporary warning is marked and the counter is incremented by 1; if the condition is not met in the next moment, the counter is cleared; only if the judgment condition is met in two consecutive sampling cycles, a valid thermal runaway warning signal is output. It filters false warnings caused by instantaneous sensor jitter, electromagnetic interference, and single-point glitches, without increasing the amount of complex calculations, and is suitable for low-computing operation of embedded BMS.

[0046] Unified model performance verification standard: Using the temperature rise rate dT / dt ≥ 1℃ / s as the label for the actual occurrence time of thermal runaway, two core evaluation indicators are quantified: (1) Early warning time: The time difference between the first effective warning and the actual moment of thermal runaway; (2) Classification accuracy: The model’s accuracy in identifying the three states: normal, warning, and thermal runaway.

[0047] For the temperature sequence at sampling time ti, the rate of temperature rise is defined as the ratio of the temperature change between two adjacent sampling points to the time interval, and its calculation formula is as follows: ; In the formula, for The rate of temperature rise at any given time, expressed in °C / s; This represents the current surface temperature of the battery. The temperature at the previous sampling time; This represents the time interval between adjacent sampling points.

[0048] To verify the effectiveness of the method provided by this invention, such as... Figure 2 - Figure 7 As shown, the following five specific implementation methods are provided: (1) Standard operating conditions: 32Ah square hard-case LFP battery, 25℃, 1C overcharge The preprocessed signal is stable, with CV < 0.05 and an adaptive minimum window. The second-order fitting yields a continuously rising trend score, exceeding the strong early warning threshold of 0.5. The rate of change of expansion force is consistently higher than the MAD dynamic threshold. An early warning is output when the condition is met for two consecutive frames. Compared with the time when the thermal runaway temperature rise rate reaches 1℃ / s, the early warning is given 681s (11.35min) in advance, with a classification accuracy of 98.95%. This is 92s, 144s, and 192s earlier than the traditional three-level temperature alarms of 52 / 56 / 60℃, respectively.

[0049] (2) Capacity difference operating conditions: 30Ah square hard-case LFP battery, 25℃, 1C overcharge The algorithm relies on relative trend characteristics, not on the absolute amplitude of expansion force, and the adaptive window automatically adapts to the capacity baseline; it provides early warning in 438 seconds with an accuracy of 99.16%, proving its strong versatility for batteries of different capacities.

[0050] (3) High-rate operating conditions: 32Ah square hard shell, 30℃, 3C fast charging overcharge 3C fast charging accelerates internal side reactions, shortening the incubation period for thermal runaway; this solution still captures the accelerating upward trend of expansion force, providing an early warning of 160 seconds with an accuracy rate of 99.04%.

[0051] (4) High temperature and high rate coupling conditions: 32Ah square hard shell, 70℃, 3C overcharge High temperatures accelerate SEI membrane decomposition and gas production, resulting in severe signal baseline drift; the robust MAD threshold effectively suppresses baseline interference, providing an early warning of 142 seconds with an accuracy of 98.84%.

[0052] (5) Operating conditions for soft-pack batteries: 32Ah soft-pack battery, 30℃, 3C overcharge The soft-shell design is highly flexible, and the internal expansion stress is released through localized bulging, resulting in low amplitude of single-point expansion force. It still provides stable early warning based on the continuous upward trend, with an advance warning time of 122 seconds and an accuracy rate of 98.11%.

[0053] Comparative Example Comparative Example 1: Traditional fixed temperature threshold alarms have a warning lead time of less than 100 seconds, resulting in severe lag. Comparative Example 2: Fixed expansion force threshold algorithm, frequent false alarms during normal charging and discharging at high rates above 2C, and a large number of missed alarms after battery aging baseline drift; Comparative Example 3: Mean-variance dynamic threshold, noise and vibration cause threshold shift, false alarm rate 3.27%; Comparative Example 4: First-order linear expansion force fitting cannot identify nonlinear accelerated gas production trends, resulting in widespread underreporting of extreme high temperature and high-rate operating conditions. Comparative Example 5: Existing relaxation voltage early warning methods have a maximum early warning window of only 240s, which is far lower than the 681s standard operating condition advance of this invention.

[0054] Example 2 An early warning system for battery thermal runaway based on expansion force trend sensing, comprising: The signal acquisition and preprocessing module is configured to simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals. The comprehensive trend score calculation module is configured to adaptively and dynamically divide the preprocessed signal into multiple data windows by sliding windows, and use second-order polynomial fitting to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, the comprehensive trend score is constructed by fusion. The warning threshold dynamic generation module is configured to calculate the median absolute deviation based on the median of the internal expansion force change rate sequence, and generate a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient. The anomaly detection module is configured to perform anomaly detection by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and to implement graded thermal runaway early warning by adopting a continuous two-step confirmation mechanism.

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made by those skilled in the art without creative effort within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for early warning of battery thermal runaway based on expansion force trend perception, characterized in that, Includes the following steps: Simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals; The preprocessed signal is divided into multiple data windows by adaptive dynamic sliding window. Second-order polynomial fitting is used to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, a comprehensive trend score is constructed by fusion. Based on the median of the internal expansion force change rate sequence, the median absolute deviation is calculated, and a real-time dynamic early warning threshold is generated by combining the signal baseline mean and sensitivity coefficient. Anomaly determination is made by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and a two-step confirmation mechanism is adopted to realize graded thermal runaway early warning.

2. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of synchronously acquiring multi-source time-series signals of battery expansion force, voltage, and surface temperature includes: using tensile and compressive force sensors placed between the clamp and the battery to acquire the preload applied by the clamp; The surface temperature of the battery is obtained using a temperature sensor. Use voltage acquisition equipment to obtain the battery terminal voltage.

3. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The standardization preprocessing of multi-source time series signals includes: time synchronization calibration, missing value completion, outlier removal, and filtering for smoothing and noise reduction.

4. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of adaptively and dynamically dividing the preprocessed signal into multiple data windows includes: calculating the coefficient of variation of the expansion force data within the window, wherein the coefficient of variation is the standard deviation of the window data / the mean of the window data; when the mean of the window approaches zero, the coefficient of variation is set to zero; and automatically adjusting the window length according to the degree of fluctuation of the coefficient of variation, wherein the smaller the coefficient of variation, the smaller the window length; and configuring independent adaptive sliding windows for the rate of change of expansion force and the trend acceleration of expansion force.

5. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of extracting the multidimensional trend features of expansion force in each data window by fitting a second-order polynomial includes: traversing the expansion force sequence by fitting polynomials from the first to the fourth order, and judging by the three indicators of goodness of fit, root mean square error and random distribution of residuals. Among them, the first order underfitting and the third / fourth order overfitting noise determine the second-order polynomial as the optimal model. The coefficients of the second-order polynomial are solved by the least squares method. The first derivative of the polynomial is obtained to get the trend slope. The trend acceleration is obtained by the time-series difference of the slope. The goodness of fit is calculated simultaneously. Weights are assigned to the trend slope, goodness of fit, and trend acceleration. A comprehensive trend score is constructed by weighted fusion. The slope is normalized to the range of 0 to 1. Only when the slope is greater than a predetermined value is it determined to be a valid fault upward trend.

6. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of calculating the median absolute deviation based on the median of the internal expansion force change rate sequence, and generating a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient includes: calculating the median of the expansion force change rate sequence within the window, calculating the median absolute deviation, introducing a standard conversion factor to convert it into a robust standard deviation estimate; and generating a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient.

7. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 6, characterized in that, Each time the window updates a set of data, the threshold is refreshed synchronously to automatically adapt to changes in capacity, temperature, and scaling factor.

8. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of determining anomalies by combining the comprehensive trend score, the rate of change of expansion force, the dynamic early warning threshold, and voltage constraints includes: setting a graded threshold for the comprehensive trend score; based on the relationship between the comprehensive trend score and the graded threshold, the comparison between the rate of change of expansion force and the dynamic early warning threshold, and the voltage overcharge auxiliary constraint, when the comprehensive trend score meets the graded threshold, the rate of change of expansion force exceeds the dynamic early warning threshold, and the voltage overcharge auxiliary constraint is met, it is determined to be an anomaly.

9. The method for early warning of battery thermal runaway based on expansion force trend perception as described in claim 1, characterized in that, The process of implementing graded thermal runaway early warning using a continuous two-step confirmation mechanism includes: if the conditions are met in a single sampling, only a temporary early warning is marked and the warning count is incremented by one; if the conditions are not met in the next moment, the warning count is reset to zero; only when the judgment conditions are met in two consecutive sampling cycles is a valid thermal runaway early warning signal generated.

10. A battery thermal runaway early warning system based on expansion force trend sensing, characterized in that, include: The signal acquisition and preprocessing module is configured to simultaneously acquire multi-source time-series signals of battery expansion force, voltage, and surface temperature, and perform standardized preprocessing on the multi-source time-series signals. The comprehensive trend score calculation module is configured to adaptively and dynamically divide the preprocessed signal into multiple data windows by sliding windows, and use second-order polynomial fitting to extract the multidimensional trend features of expansion force of each data window. Based on the multidimensional trend features of expansion force, the comprehensive trend score is constructed by fusion. The warning threshold dynamic generation module is configured to calculate the median absolute deviation based on the median of the internal expansion force change rate sequence, and generate a real-time dynamic warning threshold by combining the signal baseline mean and sensitivity coefficient. The anomaly detection module is configured to perform anomaly detection by combining the comprehensive trend score, the rate of change of expansion force and the dynamic early warning threshold with voltage constraints, and to implement graded thermal runaway early warning by adopting a continuous two-step confirmation mechanism.

Images