A power battery thermal runaway early warning method based on multi-source voltage characteristics

Abstract

The application discloses a kind of power battery thermal runaway early warning methods based on multi-source voltage characteristics, it is related to battery thermal runaway technical field, comprising the following steps: S100, multi-source voltage signal acquisition;S200, multi-source voltage signal preprocessing;S300, multi-source voltage feature extraction;S400, thermal runaway risk discrimination;S500, early warning signal output;S600, early warning model iteration optimization.The application is combined by the core collection structure of multi-channel synchronous acquisition, time sequence synchronous check and anti-interference transmission, realizes the accurate, distortionless acquisition of retired lithium iron phosphate battery monomer and module voltage signal, provides high-quality basic data support for subsequent early warning whole process, realizes the accurate elimination of voltage signal interference and baseline deviation elimination, while retaining the effective features of thermal runaway precursor, further solve the problem that current preprocessing means is not enough in pertinence, signal purity is low, leading to large feature extraction deviation, build the core foundation of accurate early warning.

Description

Technical Field

[0001] This invention relates to the field of battery thermal runaway technology, and more specifically, to a method for early warning of thermal runaway of automotive power batteries based on multi-source voltage characteristics. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the number of power batteries in use continues to grow, and the number of retired power batteries is also increasing year by year. Retired lithium iron phosphate batteries, because they still have a certain amount of remaining capacity, are widely used in the field of second-hand vehicle installation, realizing resource recycling and conforming to the concept of green development.

[0003] However, after retired lithium iron phosphate batteries are installed in vehicles, the long-term charge-discharge cycle process easily leads to a coupled state of uneven capacity decay and accelerated cycle aging. This special state results in complex and variable characteristics in the battery voltage signal, constituting a core technical challenge in the thermal runaway early warning process. Under this coupled state, the capacity decay of different individual cells varies significantly, and the voltage baseline shifts noticeably. At the same time, vibration and electromagnetic interference in the vehicle environment further exacerbate the distortion of the voltage signal, making it difficult to detect the weak voltage anomalies in the early stages of thermal runaway. This signal characteristic makes existing general-purpose early warning models designed for new power batteries unsuitable. Their fixed early warning thresholds are difficult to dynamically adjust with the battery aging state, and as the number of battery cycles increases and the degree of decay intensifies, they are prone to discrimination failure. In view of this, we propose a thermal runaway early warning method for automotive power batteries based on multi-source voltage characteristics. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art, adapt to practical needs, and provide a method for early warning of thermal runaway of vehicle power batteries based on multi-source voltage characteristics, aiming to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for early warning of thermal runaway of vehicle power batteries based on multi-source voltage characteristics, comprising the following steps:

[0006] S100, Multi-source voltage signal acquisition: Accurately acquires the voltage signal of retired lithium iron phosphate batteries in the target coupling state after being installed in vehicles through the vehicle-mounted acquisition module, adjusts the acquisition frequency as needed, and transmits it to the processing module;

[0007] S200, Multi-source voltage signal preprocessing: The acquired voltage signal is denoised, normalized and smoothed to remove various interferences and unify the numerical range, so as to ensure the accuracy of subsequent feature extraction.

[0008] S300, Multi-source voltage feature extraction: Extract multi-source voltage features of the adaptive coupling state and lithium iron phosphate battery characteristics from the preprocessed voltage signal, providing a core basis for thermal runaway risk assessment; multi-source voltage features include single-cell voltage-related features and module voltage-related features;

[0009] S400 Thermal runaway risk assessment: Construct a dedicated thermal runaway early warning model, adopt a dynamic threshold adjustment mechanism, and accurately determine the thermal runaway risk level through the dual logic of initial judgment and secondary judgment;

[0010] S500, Early Warning Signal Output: Generates corresponding early warning signals based on the final thermal runaway risk level and outputs them through multiple channels to ensure that relevant personnel can obtain them in a timely manner and take preventive measures.

[0011] S600, Early Warning Model Iterative Optimization: Collect relevant operational and case data, optimize feature selection and model discrimination logic, and continuously improve the model's adaptability to target coupling states and early warning accuracy.

[0012] Preferably, step S100 specifically includes the following sub-steps:

[0013] S101. Connect the vehicle-mounted acquisition module to the single cell and battery module of the retired lithium iron phosphate battery that are in a state of uneven capacity decay and accelerated cyclic aging after being installed in the vehicle. The connection adopts an anti-interference connector to reduce vibration and electromagnetic interference in the vehicle environment.

[0014] S102. Through the multi-channel synchronous acquisition function of the vehicle-mounted acquisition module, the individual unit voltage signal and the module voltage signal are acquired in real time, and the acquisition timing is synchronously verified by the acquisition timing synchronization verification algorithm.

[0015] The formula for the acquisition timing synchronization verification algorithm is as follows:

[0016] ;

[0017] in, This represents the maximum timing deviation across all acquisition channels; For the first The acquisition time of the voltage signal of each individual battery cell; For the first The acquisition time of the voltage signal of each battery module; This is the serial number of the battery module; This indicates taking the maximum value of the time difference between the acquisition time of all individual units and the module; This refers to the total number of individual battery cells within the power battery pack. This refers to the total number of battery modules within the power battery pack.

[0018] S103. Adjust the sampling frequency according to the charging and discharging state of the power battery. Increase the sampling frequency during charging and discharging and decrease the sampling frequency when the battery is stationary, balancing sampling accuracy and energy consumption.

[0019] S104. The collected voltage signal is transmitted to the data processing module through an encrypted transmission protocol. A data verification mechanism is set during the transmission process to ensure that the signal is not lost or distorted.

[0020] Preferably, step S200 specifically includes the following sub-steps:

[0021] S201. Wavelet transform filtering algorithm is used to denoise the voltage signal, specifically eliminating high-frequency noise and low-frequency drift interference.

[0022] The wavelet transform filtering algorithm formula is as follows:

[0023] ;

[0024] in, These are the wavelet transform coefficients of the original voltage signal; This is a function of the acquired raw voltage signal; As the independent variable; The original voltage signal was acquired. The time differential symbol; For scale parameters; For position parameters; wavelet basis functions conjugate complex number

[0025] S202. Normalize the denoised voltage signal using a normalization algorithm to unify the numerical range of the voltage signal.

[0026] The normalization algorithm formula is as follows:

[0027] ;

[0028] in, This is the normalized voltage signal; The signal is the voltage signal after wavelet transform filtering and noise reduction. This is the minimum value of the voltage signal within the historical acquisition period; This represents the maximum value of the voltage signal within the historical acquisition period;

[0029] S203. The normalized voltage signal is smoothed using a moving average voltage signal algorithm.

[0030] The formula for the moving average algorithm of the voltage signal is:

[0031] ;

[0032] in, The signal is a smoothed voltage signal; The length of the sliding window; For the normalized voltage signal at time... The summation symbol represents the summation over time. At that time of The normalized voltage signal values ​​are summed. This indicates that the summation result is averaged to achieve signal smoothing.

[0033] Preferably, the wavelet transform filtering algorithm further includes a voltage signal reconstruction algorithm;

[0034] The formula for the voltage signal reconstruction algorithm is as follows:

[0035] ;

[0036] in, This is the clean voltage signal obtained after noise reduction; These are the wavelet transform coefficients after screening and correction; These are wavelet basis functions; , These are the scale parameter and the location parameter, respectively; the double summation symbol indicates that the wavelet coefficients of all effective scales and locations are superimposed to achieve signal reconstruction.

[0037] Preferably, step S300 specifically includes the following sub-steps:

[0038] S301, Individual-level feature extraction: For the preprocessed individual voltage signals, analyze and extract the individual voltage plateau difference and individual voltage curve distortion features on an individual basis.

[0039] S302, Module-level feature extraction: Extract the module voltage range fluctuation rate and inter-module voltage coupling deviation from the preprocessed module voltage signal;

[0040] S303. Feature Verification and Screening: The multi-source voltage features (single cell voltage plateau difference, single cell voltage plateau offset feature, single cell voltage mutation feature, module voltage range fluctuation rate, and inter-module voltage coupling deviation) extracted in steps S301 and S302 are verified for correlation. Redundant features are eliminated to ensure that each feature can independently reflect different state dimensions of the battery and avoid information overlap between features that leads to reduced subsequent discrimination efficiency.

[0041] Preferably, step S400 specifically includes the following sub-steps:

[0042] S401. Combining the cycle aging pattern of retired lithium iron phosphate batteries, a correlation model is constructed between multi-source voltage characteristics and battery capacity decay level and cumulative cycle number to form a dedicated thermal runaway early warning model.

[0043] S402. Based on the capacity decay level and cumulative cycle count of retired lithium iron phosphate batteries, dynamically adjust the range of warning threshold values ​​for each multi-source voltage characteristic.

[0044] S403. The extracted multi-source voltage features are input into the early warning model, and combined with the dynamically adjusted threshold, a preliminary judgment is made to determine the thermal runaway risk level of the power battery. The risk level is divided into three categories: no risk, mild warning, and severe warning.

[0045] Among them, the mild warning threshold and the severe warning threshold adopt the mild warning threshold dynamic algorithm and the severe warning threshold dynamic algorithm, which are optimized synchronously with the capacity decay level and the number of cycles;

[0046] S404. A second verification is performed on the preliminary judgment results to confirm that the preliminary risk level is the final thermal runaway risk level.

[0047] Preferably, the dynamic algorithm formula for the mild warning threshold is:

[0048] ;

[0049] in, For capacity decay level The cumulative number of loops is The threshold for a mild early warning at that time; This serves as the initial baseline threshold for a mild warning. This is the adjustment coefficient for the mild warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the mild warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles.

[0050] The dynamic algorithm formula for the severe warning threshold is:

[0051] ;

[0052] in, For capacity decay level The cumulative number of loops is The threshold for severe early warning at that time; The initial baseline threshold for severe warning; This is the adjustment coefficient for the severe warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the severe warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles.

[0053] Preferably, the secondary discrimination logic of step S404 is as follows: if the preliminary discrimination result is a mild warning, verify whether the trend of multi-source voltage characteristic change continues to deviate from the normal range; if it is a severe warning, verify whether the rate of change of multi-source voltage characteristic reaches the preset critical state. If any verification condition is met, the corresponding preliminary risk level is confirmed as the final thermal runaway risk level.

[0054] Preferably, step S500 specifically includes the following sub-steps:

[0055] S501. Generate an early warning signal based on the final thermal runaway risk level. The signal contains key information such as the risk level, abnormal unit / module number, and abnormal voltage characteristic type.

[0056] S502. Issue early warning signals through multiple channels and trigger emergency contact mechanisms to remind relevant personnel to take prevention and control measures.

[0057] Preferably, step S600 specifically includes the following sub-steps:

[0058] S601. Collect voltage signals, operating data, and thermal runaway case data of power batteries under different charge and discharge conditions, different cycle numbers, and different capacity decay levels. Classify and organize the data, remove invalid and abnormal data, and establish a complete dataset.

[0059] S602. Based on the processed dataset, analyze the correlation between multi-source voltage characteristics and thermal runaway risk, select the optimal feature combination to adapt to battery state changes, and eliminate invalid features; at the same time, optimize the dynamic threshold adjustment mechanism and dual discrimination logic to improve the model's response speed and discrimination accuracy to battery state changes.

[0060] S603. Supplement the voltage characteristic interference removal rules in complex automotive environments and verify the performance of the optimized model.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] 1. This invention achieves accurate and distortion-free acquisition of voltage signals from retired lithium iron phosphate battery cells and modules by designing a core acquisition structure that combines multi-channel synchronous acquisition, timing synchronous verification, and anti-interference transmission. This solves the core problem of inaccurate original voltage data caused by vehicle interference and timing deviation, which affects the reliability of early warning, and provides high-quality basic data support for the entire subsequent early warning process.

[0063] 2. This invention also achieves precise removal of voltage signal interference and elimination of baseline deviation by designing a preprocessing structure that links wavelet transform filtering, voltage signal reconstruction and normalization, and moving average, while fully preserving the effective features of thermal runaway precursors. This further solves the problems of insufficient targeting and low signal purity of current preprocessing methods, which lead to large feature extraction deviations, and strengthens the core foundation for accurate early warning.

[0064] 3. This invention also designs a multi-source voltage feature extraction logic that adapts to the battery coupling state and the characteristics of lithium iron phosphate batteries, and combines it with a warning structure that combines dynamic threshold adjustment and dual discrimination to achieve accurate identification and graded discrimination of thermal runaway risk. This further solves the problems of poor adaptability of current warning models, easy failure of fixed thresholds, high false alarm and false alarm rates, and inability to capture weak anomalies in the early stage of thermal runaway, and significantly improves the accuracy and reliability of warning.

[0065] 4. This invention also designs a denoising system that combines wavelet transform filtering with voltage signal reconstruction, and uses a combination of normalization and moving average algorithms to accurately remove high-frequency noise, low-frequency drift, and instantaneous fluctuations in the voltage signal. At the same time, it fully preserves the effective voltage features related to thermal runaway and eliminates the baseline deviation of the individual cell voltage. This further solves the problem that current preprocessing methods are not targeted enough and cannot adapt to the complex characteristics of the voltage signals of retired lithium iron phosphate batteries, resulting in low signal purity and large feature extraction deviations. This lays a solid foundation for accurate extraction of voltage features. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0067] Figure 2 This is a flowchart of the data acquisition process of the present invention;

[0068] Figure 3 This is a flowchart of the preprocessing process of the present invention;

[0069] Figure 4 This is a flowchart illustrating the risk assessment process of the present invention. Detailed Implementation

[0070] Example: Figures 1 to 4 As shown, the present invention relates to a method for early warning of thermal runaway of vehicle power batteries based on multi-source voltage characteristics, comprising the following steps:

[0071] S100, Multi-source voltage signal acquisition: Accurately acquires the voltage signal of retired lithium iron phosphate batteries in the target coupling state after being installed in vehicles through the vehicle-mounted acquisition module, adjusts the acquisition frequency as needed, and transmits it to the processing module;

[0072] S101 Hardware Connection Deployment: Connect the multi-channel acquisition probe of the vehicle acquisition module to the individual cells and battery modules in the power battery pack one by one to ensure that each individual cell and module has an independent signal acquisition channel; use anti-interference connectors during the connection process to reduce the impact of vehicle environment vibration and electromagnetic interference on voltage signal transmission, especially for individual cells with high capacity decay, to reduce the degree of interference of external interference on their weak voltage signals.

[0073] S102. Synchronous Signal Acquisition: The multi-channel synchronous acquisition function of the on-board acquisition module is used to acquire the individual cell voltage signal and the module voltage signal of each battery module in real time. To ensure that the acquisition timing of all voltage signals is completely consistent and to avoid errors in subsequent calculations of voltage differences, rates of change, and other characteristics due to timing deviations, the acquisition timing must be synchronously verified using an acquisition timing synchronization verification algorithm.

[0074] The formula for the acquisition timing synchronization verification algorithm is:

[0075] ;

[0076] in, This represents the maximum timing deviation across all acquisition channels; For the first The acquisition time of the voltage signal of each individual battery cell; For the first The acquisition time of the voltage signal of each battery module; This is a serial number for the battery module, used to distinguish different battery modules; This represents the maximum value of the time difference between the acquisition time of all individual units and the module; the absolute value ensures that the timing deviation is non-negative. This refers to the total number of individual battery cells within the power battery pack. This refers to the total number of battery modules within the power battery pack; when If the data is less than or equal to the timing synchronization threshold, the acquisition timing is determined to be completely consistent; otherwise, the acquisition module is triggered to resynchronize the acquisition.

[0077] The specific values ​​for the timing synchronization threshold are as follows:

[0078] S103. Dynamic adjustment of sampling frequency: The sampling frequency is adjusted according to the current operating status of the power battery.

[0079] When the power battery is in the charging and discharging state, the sampling frequency is increased to a high-frequency sampling mode to accurately capture the instantaneous changes in the voltage signal;

[0080] When the power battery is in a static state, the acquisition frequency is adjusted to a low-frequency acquisition mode to reduce system energy consumption and avoid the impact of redundant data on processing efficiency.

[0081] S104. Encrypted signal transmission: The collected individual voltage signals and module voltage signals are transmitted to the data processing module through an encrypted transmission protocol. During the transmission process, a data verification mechanism is used to ensure that the received voltage signals are completely consistent with the original collected signals, with no signal loss or distortion.

[0082] S200, Multi-source voltage signal preprocessing: The raw voltage signals acquired are processed in a targeted manner to remove interference components from the signals, unify signal processing standards, eliminate false anomalies in voltage signals caused by uneven capacity decay and accelerated cycle aging of retired lithium iron phosphate batteries, highlight the effective voltage characteristics related to thermal runaway risk, and clear data obstacles for subsequent feature extraction.

[0083] S201. Denoising Processing: Wavelet transform filtering algorithm is used to denoise the voltage signal transmitted to the data processing module, accurately identifying and removing high-frequency noise components and low-frequency drift components in the voltage signal. High-frequency noise mainly comes from vehicle electromagnetic interference and vibration of the acquisition equipment, while low-frequency drift mainly comes from changes in internal resistance caused by accelerated battery aging and voltage baseline shift caused by uneven capacity decay. While removing interference, the effective voltage characteristics are completely preserved to avoid loss of effective signal.

[0084] The formula for the wavelet transform filtering algorithm is:

[0085] ;

[0086] in, These are the wavelet transform coefficients of the original voltage signal, used to characterize the signal's features at different scales and locations; The function identifier for the acquired raw voltage signal, used to represent the change in voltage signal; The independent variable represents the time dimension of signal acquisition, that is, the time coordinate of the signal, which characterizes the value of the signal at different times. The acquired raw voltage signal, i.e., the voltage signal at time... The specific value is determined by the function. With independent variable To be determined jointly; The time differential symbol is used to represent a small increment in the time dimension. In integration operations, it is used to divide the integration interval to achieve the cumulative calculation of continuous time signals. The scale parameter is used to correspond to the frequency range of the signal. The larger the scale, the lower the corresponding signal frequency, and the smaller the scale, the higher the corresponding signal frequency. The position parameter is used to correspond to the time dimension of the signal, representing the position of the signal on the time axis; wavelet basis functions The conjugate complex number of the wavelet is used, and the wavelet basis function adopts the db4 wavelet for signal decomposition and reconstruction.

[0087] The voltage signal is decomposed at different scales by a voltage signal reconstruction algorithm, and the wavelet coefficients corresponding to high-frequency noise and low-frequency drift are processed to finally reconstruct the denoised voltage signal.

[0088] The formula for voltage signal reconstruction algorithm is:

[0089] ;

[0090] in, This is the clean voltage signal obtained after noise reduction; These are the wavelet transform coefficients after screening and correction; These are wavelet basis functions; These are the scale parameter and the location parameter, respectively; the double summation symbol indicates that the wavelet coefficients of all effective scales and locations are superimposed to achieve signal reconstruction.

[0091] After this denoising process, the signal-to-noise ratio of the voltage signal is improved, various interference components are eliminated, and a clean signal basis is provided for subsequent feature extraction.

[0092] S202, Normalization Processing: The denoised voltage signal is normalized using a normalization algorithm. Through linear transformation, all individual unit voltage signals and module voltage signals are mapped uniformly to eliminate voltage baseline differences caused by different capacity attenuation levels in different units. This ensures that the voltage features extracted subsequently are comparable and avoids interference from baseline differences on feature calculation results.

[0093] The normalization algorithm formula is:

[0094] ;

[0095] in, This is the normalized voltage signal; This is the voltage signal after wavelet transform filtering and noise reduction; This is the minimum value of the voltage signal within the historical acquisition period; This represents the maximum value of the voltage signal within the historical acquisition period;

[0096] This normalization process solves the problem of inconsistent voltage benchmarks among different cells in retired lithium iron phosphate batteries due to uneven capacity decay, enabling direct comparison of voltage characteristics of cells with different decay levels and aging rates, laying the foundation for feature analysis across cells and modules.

[0097] S203. Smoothing: The normalized voltage signal is smoothed using a moving average algorithm to eliminate instantaneous fluctuations in the signal. For local voltage fluctuations caused by instantaneous voltage jumps and uneven capacity decay during the cycling process of retired lithium iron phosphate batteries, the smoothing process preserves the overall trend of the signal and removes instantaneous interference with no reference value to ensure the accuracy of subsequent feature extraction.

[0098] The formula for the moving average algorithm of voltage signals is:

[0099] ;

[0100] in, The signal is a smoothed voltage signal; The sliding window length can be dynamically adjusted according to the sampling frequency; For the normalized voltage signal at time... The summation symbol represents the summation over time. At that time of The normalized voltage signal values ​​are summed. This indicates that the summation result is averaged to achieve signal smoothing;

[0101] After this smoothing process, the standard deviation of voltage signal fluctuation is reduced, effectively filtering out random fluctuations caused by a single acquisition, while fully preserving the long-term trend of voltage signal changes, ensuring that the voltage features extracted subsequently can accurately reflect the real state changes of the battery.

[0102] S300. From the preprocessed voltage signal, extract multi-source voltage features that accurately reflect the state changes of retired lithium iron phosphate batteries and are suitable for identifying their thermal runaway risk, construct a feature set, and provide core basis for thermal runaway risk judgment. Considering the characteristics of uneven capacity decay and accelerated cycle aging of this type of battery, the features selected in this step must be able to effectively distinguish between normal battery state changes and thermal runaway precursor signals, avoiding judgment bias caused by feature redundancy or feature loss. Specifically, multi-source voltage features that adapt to coupling state and lithium iron phosphate battery characteristics include single-cell voltage plateau difference, module voltage range fluctuation rate, single-cell voltage curve distortion characteristics, and inter-module voltage coupling deviation.

[0103] S301, Individual-level feature extraction: For the preprocessed individual cell voltage signal, analyze and extract the individual cell voltage plateau difference and individual cell voltage curve distortion features on an individual cell basis; among which, the individual cell voltage curve distortion features include voltage plateau offset features and voltage jump features, adapting to the inherent characteristics of the smooth voltage plateau of lithium iron phosphate batteries, capturing the abnormal offset of the voltage plateau caused by uneven capacity decay, and the abnormal jump of individual cell voltage in the early stage of thermal runaway.

[0104] The formula for calculating the voltage plateau difference of individual units is as follows:

[0105] The voltage plateau characteristics of lithium iron phosphate batteries can be characterized by the range where the voltage change rate during charging and discharging is less than a threshold. By setting a voltage change rate threshold, the voltage plateau range of a single cell is determined. First, the average plateau voltage within this range is calculated. The average plateau voltage is:

[0106] ;

[0107] in, For the first The average voltage plateau of each individual cell For the first The start time of the voltage plateau of each individual cell. For the first The end time of the voltage plateau of each individual cell. The duration of the voltage plateau. For the first The voltage signal of each individual cell after smoothing; the integral symbol is the first... The voltage integral of each individual cell within the voltage plateau range is used to calculate the average voltage of that range.

[0108] Based on the average voltage platform value of the individual units mentioned above, the voltage platform difference of the individual units is calculated, which is the difference between the average voltage platform value of the individual unit and the average voltage platform value of all individual units in the module. The formula is as follows:

[0109] ;

[0110] in, For the first The voltage plateau difference of each individual cell, with the absolute value ensuring that the difference is non-negative; For the first The average voltage plateau of each individual cell For the first The average voltage plateau of all cells within the module containing a given cell is calculated using the following formula: ,(in This refers to the number of individual cells within the module. This feature accurately reflects the difference in capacity decay between individual cells; the more severe the capacity decay, the more pronounced the difference. The larger the value;

[0111] Calculation of single-cell voltage plateau offset characteristics: Using the average standard voltage plateau value of a brand-new lithium iron phosphate battery as a benchmark, calculate the offset between the current average plateau voltage of a single cell and the benchmark value, as shown in the following formula:

[0112] ;

[0113] in, For the first The voltage plateau offset of a single cell; a positive value indicates that the current plateau voltage is higher than the standard value, and a negative value indicates that the current plateau voltage is lower than the standard value. For the first The average voltage plateau of each individual cell has the same meaning as described above; This represents the average standard voltage plateau of a brand-new lithium iron phosphate battery. After cycle aging, the average voltage plateau of retired lithium iron phosphate batteries decreases linearly with capacity decay; this characteristic directly characterizes the degree of battery aging. When the temperature continues to drop and the rate of drop exceeds a preset threshold, it indicates that the battery is at risk of early thermal runaway.

[0114] Calculation of single-cell voltage mutation characteristics: This is characterized by calculating the maximum mutation value of single-cell voltage per unit time, with a time window set. The specific formula is as follows:

[0115] ;

[0116] in, For the first The maximum voltage fluctuation of a single cell within a unit time window. This means taking the maximum value of the expression within the parentheses within the set time window. This is the start time of the time window. The time window length, ) is the first Each individual cell at time The smoothed voltage value, For the first Each individual cell at time The smoothed voltage value; the absolute value ensures that the sudden change value is non-negative. In the early stages of thermal runaway, anomalies such as micro-short circuits and lithium plating occur inside the individual cell, causing rapid voltage changes. This feature can effectively capture such sudden risks. When a preset threshold is reached, a preliminary risk assessment is triggered;

[0117] S302, Module-level feature extraction: For the pre-processed module voltage signal, extract the module voltage range fluctuation rate and the voltage coupling deviation between modules; the module voltage range fluctuation rate reflects the degree of consistency change of the individual cell voltage within the module, and the voltage coupling deviation between modules reflects the voltage coordination state between multiple modules. By capturing abnormal voltage changes at the module level through these features, the overall aging degree and health status of the battery can be reflected.

[0118] Module voltage range fluctuation rate:

[0119] First, calculate the voltage range of each individual cell within the module at any given time:

[0120] ;

[0121] Calculate the statistical period again Within the range volatility, statistical period Based on the battery charging and discharging conditions, the range fluctuation rate is the average of the absolute values ​​of the deviations between the range at each moment within the period and the average range within the period. The specific formula is as follows:

[0122] ;

[0123] in, This refers to the voltage range fluctuation rate of the module. The statistical period is set according to the battery charging and discharging conditions. For statistical periods Inner 1 to The voltage range of the modules at each time point is summed. For the module at time The individual unit voltage difference For statistical period The average range of the internal module voltage, Used to normalize the deviation and prevent the absolute value of the range from being affected by the mean range;

[0124] in, For statistical period The average voltage range within the module directly reflects the voltage consistency of individual cells within the module. As retired lithium iron phosphate batteries experience accelerated cycle aging, the differences in cell degradation within the module further increase, leading to… The increase is significant and can serve as an important early warning indicator for module-level thermal runaway risk;

[0125] The formula for calculating the voltage coupling deviation between modules is:

[0126] ;

[0127] in, To represent modules Voltage coupling deviation, To represent modules The average voltage, This is the average voltage of all modules within the battery pack.

[0128] in, ,in For modules The number of individual cells in the cell, Indicates the selection of a module All individual cells inside, To represent modules Inner Smoothed voltage signals of individual cells after preprocessing ,in, This refers to the total number of battery modules within the battery pack. To indicate the contents of the battery pack, 1 to The average voltage of all modules is summed. To indicate the first The average voltage of each module.

[0129] When the voltage coordination between modules is normal, the average voltage of each module tends to be consistent. The values ​​are small and fluctuate smoothly; however, when the overall aging of the battery pack accelerates or some modules show signs of impending thermal runaway, the average voltage of the corresponding modules will deviate abnormally, leading to... An abnormal increase can serve as an important indicator of module-level thermal runaway risk, enabling comprehensive monitoring of the overall state of the battery pack.

[0130] S303. Feature Verification and Screening: The multi-source voltage features (single-cell voltage plateau difference, single-cell voltage plateau offset feature, single-cell voltage mutation feature, module voltage range fluctuation rate, and inter-module voltage coupling deviation) extracted in steps S301 and S302 are verified for correlation. Redundant features are eliminated to ensure that each feature can independently reflect different states of the battery, avoiding information overlap between features that would reduce subsequent discrimination efficiency. At the same time, considering the uneven capacity decay of retired lithium iron phosphate batteries, the combination of single-cell analysis and inter-module comparative analysis ensures that the extracted features can accurately cover the voltage state of each single cell and each module, avoiding incomplete feature representation due to differences in single-cell decay and inter-module states, and ensuring the comprehensiveness and effectiveness of the feature set.

[0131] The correlation test uses the Pearson correlation coefficient to characterize the degree of linear correlation between any two features. The specific formula is as follows:

[0132] ;

[0133] in, Represents any two voltage characteristics and The Pearson correlation coefficient is used to characterize the degree of linear association between two features. Representation of features and characteristics The covariance is used to quantify the strength of the linear correlation between two features; Representation of features standard deviation Representation of features The standard deviation is used to normalize the covariance. and These represent any two extracted multi-source voltage features.

[0134] S400: Construct a dedicated thermal runaway early warning model, adopt dynamic threshold adjustment and dual discrimination logic to accurately identify the thermal runaway risk level of retired lithium iron phosphate batteries, distinguish between voltage fluctuations caused by normal battery aging and voltage distortions that are precursors to thermal runaway, and solve the technical problems of high false alarm rate and high missed alarm rate of this type of battery.

[0135] S401. Early warning model construction: Combining the cycle aging law of retired lithium iron phosphate batteries, construct a correlation model between multi-source voltage characteristics and battery capacity decay level and cumulative cycle number to form a dedicated thermal runaway early warning model; during the model construction process, fully consider the dynamic changes of uneven battery capacity decay and accelerated cycle aging, avoid using general early warning models, and ensure the model's adaptability to the state changes of this type of battery.

[0136] The early warning model employs a multi-feature weighted fusion risk scoring model, with the core formula being:

[0137] ;

[0138] in, A comprehensive score for the risk of thermal runaway. For the first The weight coefficients of each core feature For the first Standardized values ​​for each core feature.

[0139] S402, Dynamic Threshold Setting: Based on the capacity decay level and cumulative cycle count of retired lithium iron phosphate batteries, dynamically adjust the range of warning threshold values ​​for each multi-source voltage characteristic; as the number of battery cycles increases and the degree of capacity decay escalates, synchronously adjust the warning threshold of the corresponding characteristic to adapt to the dynamic changes in battery status and avoid the judgment failure problem that occurs during battery aging when the fixed threshold is used.

[0140] S403. Preliminary Risk Assessment: The extracted multi-source voltage features are input into the early warning model and combined with the dynamically adjusted threshold for preliminary assessment. The thermal runaway risk level of the power battery is preliminarily determined. The risk level is divided into three categories: no risk, mild warning, and severe warning.

[0141] Among them, the mild warning threshold and the severe warning threshold adopt the mild warning threshold dynamic algorithm and the severe warning threshold dynamic algorithm, which are optimized synchronously with the capacity decay level and the number of cycles;

[0142] The dynamic algorithm formula for the mild warning threshold is:

[0143] ;

[0144] in, For capacity decay level The cumulative number of loops is The threshold for a mild early warning at that time; This serves as the initial baseline threshold for a mild warning. This is the adjustment coefficient for the mild warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the mild warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles.

[0145] The dynamic algorithm formula for the severe warning threshold is:

[0146] ;

[0147] in, For capacity decay level The cumulative number of loops is The threshold for severe early warning at that time; The initial baseline threshold for severe warning; This is the adjustment coefficient for the severe warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the severe warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles.

[0148] The preliminary judgment rules are as follows:

[0149] When the overall score ( When the threshold for mild warning is reached, it is determined to be risk-free. At this time, the battery is in normal condition and there is no risk of thermal runaway.

[0150] when ( When the threshold for severe warning is reached, it is judged as a mild warning. At this time, the battery has a potential risk of thermal runaway and needs to be continuously monitored.

[0151] when When the battery shows obvious signs of impending thermal runaway, a severe warning is issued, requiring immediate triggering of the warning output.

[0152] S404. Secondary Risk Verification: A secondary verification is performed on the preliminary judgment results. If the preliminary judgment result is a mild warning, verify whether the changing trend of the multi-source voltage characteristics continues to deviate from the normal range. If the preliminary judgment result is a severe warning, verify whether the changing rate of the multi-source voltage characteristics reaches the preset critical state. If either verification condition is met, the preliminary risk level is confirmed as the final thermal runaway risk level. If not, the preliminary judgment is re-performed to eliminate misjudgments caused by transient interference.

[0153] S500 generates differentiated early warning signals based on the final determined thermal runaway risk level and outputs them through multiple channels to ensure that relevant personnel can obtain early warning information in a timely manner, take corresponding prevention and control measures, and reduce the probability of thermal runaway accidents.

[0154] S501, Early Warning Signal Generation: Generate a corresponding early warning signal based on the final thermal runaway risk level. The signal content includes key information such as risk level, abnormal unit / module number, and abnormal voltage characteristic type, which facilitates personnel to quickly locate the risk location and type.

[0155] S502, Multi-channel early warning output: Early warning signals are transmitted through multiple channels, including vehicle terminal display, audible and visual alarms, and remote terminal push notifications. Among them, mild early warnings correspond to text reminders on the vehicle terminal, low-intensity audible and visual alarms, and remote terminal information push notifications, reminding drivers and maintenance personnel to pay attention to the battery status. Severe early warnings correspond to continuous flashing display on the vehicle terminal, high-intensity audible and visual alarms, and real-time early warning push notifications on the remote terminal, while triggering an emergency contact mechanism to notify maintenance personnel to take timely preventive measures such as shutdown and cooling.

[0156] S600 continuously optimizes the selection of multi-source voltage characteristics and the judgment logic of early warning models by collecting actual operating data and thermal runaway case data of power batteries, thereby improving the model's adaptability to changes in the state of retired lithium iron phosphate batteries and ensuring the accuracy and robustness of early warnings.

[0157] S601. Data collection and organization: Collect voltage signals and operating data of power batteries under different charge and discharge conditions, different cycle numbers, and different capacity decay levels, as well as voltage change data in thermal runaway cases. Classify and organize the collected data, remove invalid and abnormal data, and establish a complete dataset to provide reliable data support for subsequent feature and logic optimization and model performance verification.

[0158] S602, Feature and Logic Optimization: Based on the processed dataset, analyze the correlation between multi-source voltage features and thermal runaway risk, select the optimal feature combination to adapt to battery state changes, and eliminate invalid features; at the same time, optimize the dynamic threshold adjustment mechanism and dual discrimination logic to improve the model's response speed and discrimination accuracy to battery state changes.

[0159] S603. Model Performance Verification: Based on the collected dataset, the performance of the optimized early warning model is verified. The model's early warning accuracy, false alarm rate, and missed alarm rate are tested under different operating conditions. Further adjustments and optimizations are made to address the problems that arise during the verification until the model performance meets the actual vehicle application requirements. At the same time, new data is collected regularly to iteratively optimize the model.

[0160] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A method for early warning of thermal runaway in automotive power batteries based on multi-source voltage characteristics, characterized in that, Includes the following steps: S100, Multi-source voltage signal acquisition: Accurately acquires the voltage signal of retired lithium iron phosphate batteries in the target coupling state after being installed in vehicles through the vehicle-mounted acquisition module, adjusts the acquisition frequency as needed, and transmits it to the processing module; S200, Multi-source voltage signal preprocessing: The acquired voltage signal is denoised, normalized and smoothed to remove various interferences and unify the numerical range, so as to ensure the accuracy of subsequent feature extraction. S300, Multi-source voltage feature extraction: Extract multi-source voltage features of the adaptive coupling state and lithium iron phosphate battery characteristics from the preprocessed voltage signal, providing a core basis for thermal runaway risk assessment; multi-source voltage features include single-cell voltage-related features and module voltage-related features; S400 Thermal runaway risk assessment: Construct a dedicated thermal runaway early warning model, adopt a dynamic threshold adjustment mechanism, and accurately determine the thermal runaway risk level through the dual logic of initial judgment and secondary judgment; S500, Early Warning Signal Output: Generates corresponding early warning signals based on the final thermal runaway risk level and outputs them through multiple channels to ensure that relevant personnel can obtain them in a timely manner and take preventive measures. S600, Early Warning Model Iterative Optimization: Collect relevant operational and case data, optimize feature selection and model discrimination logic, and continuously improve the model's adaptability to target coupling states and early warning accuracy.

2. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, Step S100 specifically includes the following sub-steps: S101. Connect the vehicle-mounted acquisition module to the single cell and battery module of the retired lithium iron phosphate battery that are in a state of uneven capacity decay and accelerated cyclic aging after being installed in the vehicle. The connection adopts an anti-interference connector to reduce vibration and electromagnetic interference in the vehicle environment. S102. Through the multi-channel synchronous acquisition function of the vehicle-mounted acquisition module, the individual unit voltage signal and the module voltage signal are acquired in real time, and the acquisition timing is synchronously verified by the acquisition timing synchronization verification algorithm. The formula for the acquisition timing synchronization verification algorithm is as follows: ； in, This represents the maximum timing deviation across all acquisition channels; For the first The acquisition time of the voltage signal of each individual battery cell; For the first The acquisition time of the voltage signal of each battery module; This is the serial number of the battery module; This indicates taking the maximum value of the time difference between the acquisition time of all individual units and the module; This refers to the total number of individual battery cells within the power battery pack. This refers to the total number of battery modules within the power battery pack. S103. Adjust the sampling frequency according to the charging and discharging state of the power battery. Increase the sampling frequency during charging and discharging and decrease the sampling frequency when the battery is stationary, balancing sampling accuracy and energy consumption. S104. The collected voltage signal is transmitted to the data processing module through an encrypted transmission protocol. A data verification mechanism is set during the transmission process to ensure that the signal is not lost or distorted.

3. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, Step S200 specifically includes the following sub-steps: S201. Wavelet transform filtering algorithm is used to denoise the voltage signal, specifically eliminating high-frequency noise and low-frequency drift interference. The wavelet transform filtering algorithm formula is as follows: ； in, These are the wavelet transform coefficients of the original voltage signal; This is a function of the acquired raw voltage signal; As the independent variable; The original voltage signal was acquired. The time differential symbol; For scale parameters; For position parameters; wavelet basis functions The conjugate of complex numbers; S202. Normalize the denoised voltage signal using a normalization algorithm to unify the numerical range of the voltage signal. The normalization algorithm formula is as follows: ； in, This is the normalized voltage signal; The signal is the voltage signal after wavelet transform filtering and noise reduction. This is the minimum value of the voltage signal within the historical acquisition period; This represents the maximum value of the voltage signal within the historical acquisition period; S203. The normalized voltage signal is smoothed using a moving average voltage signal algorithm. The formula for the moving average algorithm of the voltage signal is: ； in, The signal is a smoothed voltage signal; The length of the sliding window; For the normalized voltage signal at time... The summation symbol represents the summation over time. At that time of The normalized voltage signal values ​​are summed. This indicates that the summation result is averaged to achieve signal smoothing.

4. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 3, characterized in that, The wavelet transform filtering algorithm also includes a voltage signal reconstruction algorithm; The formula for the voltage signal reconstruction algorithm is as follows: ； in, This is the clean voltage signal obtained after noise reduction; These are the wavelet transform coefficients after screening and correction; These are wavelet basis functions; , These are the scale parameter and the location parameter, respectively; the double summation symbol indicates that the wavelet coefficients of all effective scales and locations are superimposed to achieve signal reconstruction.

5. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, Step S300 specifically includes the following sub-steps: S301, Individual-level feature extraction: For the preprocessed individual voltage signals, analyze and extract the individual voltage plateau difference and individual voltage curve distortion features on an individual basis. S302, Module-level feature extraction: Extract the module voltage range fluctuation rate and inter-module voltage coupling deviation from the preprocessed module voltage signal; S303, Feature Verification and Screening: The multi-source voltage features extracted in steps S301 and S302 are verified for correlation, redundant features are eliminated, and each feature is ensured to independently reflect different state dimensions of the battery, avoiding information overlap between features that would reduce subsequent discrimination efficiency.

6. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, Step S400 specifically includes the following sub-steps: S401. Combining the cycle aging pattern of retired lithium iron phosphate batteries, a correlation model is constructed between multi-source voltage characteristics and battery capacity decay level and cumulative cycle number to form a dedicated thermal runaway early warning model. S402. Based on the capacity decay level and cumulative cycle count of retired lithium iron phosphate batteries, dynamically adjust the range of warning threshold values ​​for each multi-source voltage characteristic. S403. The extracted multi-source voltage features are input into the early warning model, and combined with the dynamically adjusted threshold, a preliminary judgment is made to determine the thermal runaway risk level of the power battery. The risk level is divided into three categories: no risk, mild warning, and severe warning. Among them, the mild warning threshold and the severe warning threshold adopt the mild warning threshold dynamic algorithm and the severe warning threshold dynamic algorithm, which are optimized synchronously with the capacity decay level and the number of cycles; S404. A second verification is performed on the preliminary judgment results to confirm that the preliminary risk level is the final thermal runaway risk level.

7. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 6, characterized in that, The formula for the dynamic algorithm of the mild warning threshold is: ； in, For capacity decay level The cumulative number of loops is The threshold for a mild early warning at that time; This serves as the initial baseline threshold for a mild warning. This is the adjustment coefficient for the mild warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the mild warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles. The formula for the dynamic algorithm of the severe warning threshold is: ； in, For capacity decay level The cumulative number of loops is The threshold for severe early warning at that time; The initial baseline threshold for severe warning; This is the adjustment coefficient for the severe warning threshold corresponding to the capacity decay level; This is an indicator of capacity degradation level; This is the adjustment coefficient for the severe warning threshold corresponding to the number of cycles; This represents the cumulative number of battery cycles.

8. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 7, characterized in that, The secondary discrimination logic of step S404 is as follows: if the preliminary discrimination result is a mild warning, verify whether the trend of multi-source voltage characteristic change continues to deviate from the normal range; if it is a severe warning, verify whether the rate of change of multi-source voltage characteristic reaches the preset critical state. If either verification condition is met, the corresponding preliminary risk level is confirmed as the final thermal runaway risk level.

9. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, Step S500 specifically includes the following sub-steps: S501. Generate an early warning signal based on the final thermal runaway risk level. The signal contains key information such as the risk level, abnormal unit / module number, and abnormal voltage characteristic type. S502. Issue early warning signals through multiple channels and trigger emergency contact mechanisms to remind relevant personnel to take prevention and control measures.

10. The method for early warning of thermal runaway of vehicle power battery based on multi-source voltage characteristics according to claim 1, characterized in that, S600 specifically includes the following sub-steps: S601. Collect voltage signals, operating data, and thermal runaway case data of power batteries under different charge and discharge conditions, different cycle numbers, and different capacity decay levels. Classify and organize the data, remove invalid and abnormal data, and establish a complete dataset. S602. Based on the processed dataset, analyze the correlation between multi-source voltage characteristics and thermal runaway risk, select the optimal feature combination to adapt to battery state changes, and eliminate invalid features; at the same time, optimize the dynamic threshold adjustment mechanism and dual discrimination logic to improve the model's response speed and discrimination accuracy to battery state changes. S603. Supplement the voltage characteristic interference removal rules in complex automotive environments and verify the performance of the optimized model.

Images