A method and system for early warning of battery thermal runaway based on multi-dimensional feature fusion

Abstract

This invention relates to the field of battery safety management technology, specifically to a battery thermal runaway early warning method and system based on multi-dimensional feature fusion. The method includes: collecting thermal feature data, gas feature data, and electrochemical feature data of the battery; calculating a temperature gradient vector and isothermal consistency index using the thermal feature data, and capturing abnormal changes in battery hotspot regions using a local dynamic upsampling method; calculating the phase difference between internal and external temperature fields using the thermal and gas feature data, and determining the heat source attribute by combining it with a heat contribution ratio model; calculating the phase difference of the cross-modal early response and constructing a phase difference feature matrix; combining the temperature gradient vector, heat source attribute, and phase difference feature matrix to form a fused feature matrix, which is then input into a dynamic weighted fusion discrimination model to calculate the thermal runaway risk probability. This invention, based on a multi-dimensional feature fusion method, calculates the thermal runaway risk probability in real time, achieving early and accurate early warning of battery thermal runaway.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery safety management, and in particular to a battery thermal runaway early warning method and system based on multi-dimensional feature fusion. BACKGROUND

[0002] With the wide application of power batteries in new energy vehicles, energy storage systems and high-power devices, the number of single batteries and system integration are continuously increasing, and the threat of thermal runaway accidents to personnel and equipment safety is increasingly prominent. The existing thermal runaway early warning method mainly relies on single modal monitoring such as temperature or voltage, and fixed threshold judgment. Under complex operating conditions, a single parameter is easily disturbed by external environmental changes or local abnormalities, resulting in insufficient early warning accuracy. Therefore, multi-modal fusion analysis based on thermal, gas and electrochemical characteristics has become a key technology direction to improve battery safety monitoring.

[0003] During the operation of the battery pack, there are still key problems that are difficult to effectively monitor in some key scenarios: first, when the overall temperature distribution on the surface of the battery tends to be uniform, the temperature rise signal of the local early hot spot will be covered, forming an isothermal shield, so that the monitoring method based on temperature uniformity index misses the risk point; second, when external heat sources, such as adjacent module heat dissipation, abnormal backflow of the cooling system or environmental heat radiation, act on the local area of the battery, an abnormal signal similar to internal spontaneous heat will be generated in the temperature distribution, forming an external heat source false hot spot, which leads to system misjudgment and triggers unnecessary protection actions; third, the response time of different physical and chemical processes in the early stage of thermal runaway is different, which will cause the time sequence correlation characteristics to be ignored in risk assessment. SUMMARY

[0004] The purpose of the present application is to provide a battery thermal runaway early warning method and system based on multi-dimensional feature fusion to solve the key problems that are difficult to effectively monitor in some key scenarios as proposed in the background.

[0005] To achieve the above purpose, the technical solution of the present application is: a battery thermal runaway early warning method based on multi-dimensional feature fusion, comprising:

[0006] S1, based on an internal and external field synchronous sampling mechanism, collecting thermal feature data, gas feature data and electrochemical feature data of the battery;

[0007] S2, calculating a temperature gradient vector and an isothermal consistency index using the thermal feature data, and when the isothermal consistency index is lower than a preset isothermal threshold and there is an anomaly in the non-thermal feature mode, using a local dynamic oversampling method to capture abnormal changes in the battery hot spot area;

[0008] S3, calculating the phase difference of the internal and external field temperature difference using the thermal feature data and the gas feature data, and determining the heat source attribute to distinguish the external heat source and the self-heating of the battery;

[0009] S4, calculating the phase difference of the cross-modal early response based on the thermal feature data, the gas feature data, and the electrochemical feature data, constructing a phase difference feature matrix, combining the temperature gradient vector, the heat source attribute, and the phase difference feature matrix to form a fusion feature matrix, and inputting the fusion feature matrix into a dynamic weight fusion discrimination model to calculate the thermal runaway risk probability.

[0010] Preferably, in S1, the internal and external field synchronous sampling mechanism is a multi-modal sensor synchronous trigger collection method with unified global time base and time drift compensation between the internal and external of the battery pack, which is used for consistent synchronous sampling of multi-modal data in time domain and spatial domain;

[0011] The thermal feature data includes battery surface temperature, local temperature gradient, and cooling outlet temperature difference; the gas feature data includes combustible gas concentration, gas release rate, and gas composition ratio; and the electrochemical feature data includes terminal voltage, working current, and internal resistance change rate.

[0012] Preferably, in S2, the isothermal consistency index is used to represent the uniformity of the temperature field of each monitoring position of the battery, and the specific calculation method is as follows:

[0013] Based on the thermal feature data, a temperature gradient vector is calculated using a three-dimensional space temperature field interpolation algorithm, and an isothermal consistency index is calculated according to the length distribution of the temperature gradient vector.

[0014] Preferably, in S2, the preset isothermal threshold is calculated by the thermal feature data collected under the rated working condition of the battery, and the upper limit of the isothermal consistency index is determined by combining the critical value of the isothermal consistency index to obtain the preset isothermal threshold;

[0015] The non-thermal feature modal is two types of non-temperature field signal modal, i.e., gas feature data and electrochemical feature data; the gas feature data is used to reflect the change of the battery gas release state; and the electrochemical feature data is used to reflect the change of the internal electrochemical reaction and the electrical conductivity of the battery.

[0016] Preferably, in S2, the local dynamic oversampling method is to determine the oversampling area based on the change rate of the temperature gradient vector and the abnormal modal trigger signal, and temporarily improve the sensor sampling frequency and spatial sampling density in the oversampling area, which is used to capture the subtle dynamic characteristics of local temperature change in the early stage of thermal runaway;

[0017] The specific steps of using the local dynamic oversampling method to capture abnormal changes of the battery hot spot area are as follows:

[0018] The rate of change of the temperature gradient vector and the hotspot suspected area are calculated; the sampling frequency is dynamically adjusted and the spatial sampling points are increased in the hotspot suspected area; the data collected by the increased spatial sampling points are the oversampling data, the difference in the sliding time window of the oversampling data is calculated, the short-time temperature rise rate is extracted and compared with the benchmark temperature rise rate; when the short-time temperature rise rate is continuously higher than the benchmark temperature rise rate for more than a preset duration, it is determined that the hotspot suspected area has abnormal temperature change.

[0019] Preferably, the internal-external field temperature difference phase difference refers to the phase difference value of the battery shell surface temperature rise curve and the battery monomer internal temperature rise curve on the time axis, which represents the time sequence offset relationship between the external temperature rise and the internal temperature rise in the battery heat conduction process.

[0020] The internal-external field temperature difference phase difference is calculated by using the thermal characteristic data and the gas characteristic data, and specifically as follows:

[0021] The temperature sampling sequences arranged on the battery shell surface and the battery monomer internal are respectively acquired, and time sequence alignment is performed under the same time reference; the aligned temperature sampling sequences are denoised and normalized to obtain the external field temperature change curve and the internal field temperature change curve; the maximum correlation lag time of the external field temperature change curve and the internal field temperature change curve is calculated by using the cross-correlation function, and the maximum correlation lag time is normalized as a phase difference value; the change rate of the gas characteristic data in the same time window is extracted, and the phase difference value is corrected in combination with the gas abnormal occurrence time point to obtain the final internal-external field temperature difference phase difference.

[0022] Preferably, in S3, the heat contribution ratio model is a multi-source energy attribution calculation model established based on the thermal characteristic data and the gas characteristic data of the battery, which is used to calculate the respective heat proportion of the external heat source and the battery internal self-heating, and determine the heat source attribute.

[0023] The heat source attribute determined in combination with the heat contribution ratio model is specifically as follows:

[0024] The total heat change amount of the battery external field and the internal field in the target time window is calculated according to the thermal characteristic data; the chemical exothermic amount in the target time window is determined according to the gas characteristic data; the chemical exothermic amount is attributed to the battery internal self-heating component, and the heat change difference between the external field and the internal field is attributed to the external heat source component; the ratio of the internal self-heating component to the total heat change amount is calculated as the internal heat contribution rate; when the internal heat contribution rate is higher than the set proportion threshold, it is determined that the heat source attribute is internal self-heating, otherwise it is determined as an external heat source.

[0025] Preferably, in the S4, the cross-modal early response phase difference refers to a multi-modal response time difference quantitative index calculated by a time sequence alignment method and a phase extraction method when the changes of the thermal feature data, the gas feature data and the electrochemical feature data of the battery, and is used to represent the sequence relationship of different physical and chemical processes in the thermal runaway of the battery.

[0026] The specific method for calculating the cross-modal early response phase difference and constructing the phase difference feature matrix is as follows:

[0027] The thermal feature data, the gas feature data and the electrochemical feature data of the battery are preprocessed and time sequence synchronized under a unified time reference; the feature peak points of each modal signal are extracted; the response time difference between any two modalities is calculated and converted into a phase difference value; and the phase difference values between any two modalities are filled and constructed into a phase difference feature matrix according to the modal combination order.

[0028] Preferably, in the S4, the fusion feature matrix is formed by combining the temperature gradient vector, the heat source attribute and the phase difference feature matrix according to a preset feature arrangement order;

[0029] The dynamic weight fusion discriminant model is constructed by a base feature weight self-adaptive distribution algorithm and a nonlinear probability mapping mechanism, and is used to calculate the thermal runaway risk probability when the battery is in thermal runaway, and the specific method is as follows:

[0030] The weight coefficients of each feature in the fusion feature matrix are dynamically calculated, the fusion feature matrix is distributed and calculated according to the weight coefficients to obtain a weighted fusion vector, and the weighted fusion vector is input into a multi-layer nonlinear mapping structure for probability mapping to obtain a thermal runaway risk probability value.

[0031] On the other hand, the application provides a battery thermal runaway early warning system based on multi-dimensional feature fusion, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize the battery thermal runaway early warning method based on multi-dimensional feature fusion.

[0032] Compared with the prior art, the above technical scheme of the application has the following beneficial technical effects:

[0033] 1、In the application, based on the synchronous acquisition of the multi-modal sensor and the three-dimensional temperature field gradient analysis, the isothermal shielding area caused by the local temperature gradient being covered can be identified in the case that the overall temperature field is relatively uniform, so that the potential abnormal heating point can be found in advance;

[0034] 2. In this invention, by analyzing the heat contribution ratio and calculating the phase difference of the cross-modal response, it is possible to effectively distinguish between internal self-heating of the battery and false hot spots caused by external heat sources. Furthermore, by utilizing the sequential relationship of multiple signals, the accuracy and reliability of thermal runaway risk assessment can be improved. Attached Figure Description

[0035] Figure 1 This is a flowchart of one embodiment of the present invention. Detailed Implementation

[0036] Example 1, as Figure 1 As shown, the present invention proposes a battery thermal runaway early warning method based on multi-dimensional feature fusion, and its specific implementation steps are as follows:

[0037] S1. Based on the synchronous sampling mechanism of internal and external fields, thermal characteristic data, gas characteristic data and electrochemical characteristic data of the battery are collected;

[0038] S2. Calculate the temperature gradient vector and isothermal consistency index using thermal feature data. When the isothermal consistency index is lower than the preset isothermal threshold and there are abnormalities in the non-thermal feature modes, use the local dynamic upsampling method to capture abnormal changes in the battery hot spot area.

[0039] S3. Calculate the phase difference of the internal and external field temperature difference using thermal characteristic data and gas characteristic data, and determine the heat source attribute by combining the heat contribution ratio model, so as to distinguish between external heat sources and internal self-heating of the battery.

[0040] S4. Based on thermal characteristic data, gas characteristic data and electrochemical characteristic data, calculate the phase difference of the early response across modes and construct the phase difference feature matrix. Combine the temperature gradient vector, heat source attributes and phase difference feature matrix to form a fusion feature matrix, and input it into the dynamic weight fusion discrimination model to calculate the probability of thermal runaway risk.

[0041] In this embodiment S1, the internal and external field synchronous sampling mechanism is a multi-modal sensor synchronous triggering acquisition method with unified global time base and time drift compensation between the inside and outside of the battery pack, which is used to perform consistent synchronous sampling of multi-modal data in the time domain and spatial domain.

[0042] The thermal characteristic data includes battery surface temperature, local temperature gradient, and cooling outlet temperature difference; the gas characteristic data includes combustible gas concentration, release rate, and gas composition ratio; the electrochemical characteristic data includes terminal voltage, operating current, and internal resistance change rate.

[0043] In this embodiment, the design purpose of the internal and external field synchronous sampling mechanism is to ensure that the battery pack external environment monitoring sensor and the internal embedded sensor are strictly aligned in the collection time, avoiding the cross-modal data mismatch caused by time deviation; the internal and external field synchronous sampling mechanism realizes the unified triggering of all sensors by setting a unified global time base in the system main control unit, combining a temperature compensation type low drift crystal oscillator and a synchronous pulse trigger signal generation module; for the possible sampling clock drift, the system adopts a bidirectional time difference measurement and drift compensation algorithm for dynamic correction, thereby ensuring the time synchronization of the internal and external field data.

[0044] In the sensor layout, the thermal characteristic sensor is preferably distributed in the surface hot spot area of the battery module, the cooling liquid inlet and outlet, so as to capture the temperature gradient and the change of the cooling system heat exchange efficiency. The gas sensor is deployed in the gas flow path in the battery shell cavity and the near-field area outside the shell, supporting real-time detection of flammable gas concentration change and gas release component analysis; the electrochemical characteristic acquisition module is connected with the battery management system bus to obtain high-precision terminal voltage, instantaneous working current and internal resistance change rate based on multi-frequency alternating current impedance measurement; the collected raw data is attached with a synchronous time stamp and collection position information after preliminary filtering and quantization processing at the collection end, and is transmitted to the central processing unit through a low delay bus.

[0045] In this embodiment S2, the isotherm consistency index is used to represent the uniformity of the temperature field of each monitoring position of the battery, and the specific calculation method is as follows:

[0046] Based on the thermal characteristic data, the temperature gradient vector is calculated by using the three-dimensional space temperature field interpolation algorithm, and the isotherm consistency index is calculated according to the length distribution of the temperature gradient vector.

[0047] In this embodiment, the calculation of the temperature gradient vector aims to reflect the spatial distribution characteristics of the temperature change between different monitoring points of the battery pack; through the thermal characteristic data collected by the internal and external field synchronization, a three-dimensional space coordinate system of the battery pack is established by using the three-dimensional space temperature field interpolation algorithm, and according to the physical installation position of each sensor and the measured temperature value, a continuous three-dimensional temperature field is generated by using the multivariate spline interpolation or radial basis function interpolation method; in the three-dimensional temperature field, the temperature gradient vector is composed of the first order partial derivative of the temperature of each coordinate point, the vector direction represents the fastest path of temperature rise, and the vector length represents the strength of the local temperature change; wherein, the three-dimensional space temperature field interpolation algorithm here refers to a calculation method for reconstructing the continuous temperature distribution of the battery in three-dimensional space based on discrete temperature sampling points, which is used to obtain a complete spatial temperature field in the case of limited number of sensors.

[0048] In the embodiment, the isothermal consistency index is used to quantify the uniformity of the entire temperature field, and the calculation is based on the statistical distribution of the temperature gradient vector length; specifically, the entire battery pack monitoring area is divided into several spatial units, the mean and variance of the temperature gradient vector length in each unit are calculated, and the inverse value of the variance after normalization is evaluated in the global range as the isothermal consistency index; the value of the isothermal consistency index ranges from 0 to 1, and the closer the value is to 1, the more uniform the temperature field distribution, and there is no significant hot spot area inside; on the contrary, there is a strong local temperature difference phenomenon.

[0049] In the embodiment, to improve the anti-interference, a temperature sensor calibration compensation and outlier rejection mechanism is introduced in the process of calculating the isothermal consistency index to avoid the amplification effect of single-point measurement error on the isothermal consistency index; the isothermal consistency index is not only used for single-time state judgment, but also can be combined with time series analysis to monitor its change trend, so as to identify the abnormal evolution process of the temperature field caused by the decrease of external cooling efficiency or internal local heating in advance.

[0050] In the embodiment S2, the preset isothermal threshold is calculated by the thermal characteristic data of the battery under the rated working condition, and the upper limit of the isothermal consistency index is calculated, and the preset isothermal threshold is determined by weighting the critical value of the isothermal consistency index;

[0051] The non-thermal characteristic modalities are two types of non-temperature field signal modalities of gas characteristic data and electrochemical characteristic data; the gas characteristic data is used to reflect the change of the battery gas emission state; the electrochemical characteristic data is used to reflect the change of the internal electrochemical reaction and the electric conduction characteristic of the battery.

[0052] In the embodiment, the preset isothermal threshold is set based on the thermal characteristic data collected under the rated working condition, and the thermal characteristic data is sampled in layers according to the environmental temperature interval, the state of charge interval and the cooling working condition, and the temperature sequence in the stable period is obtained by using the internal and external field synchronous sampling mechanism; after the sensor zero point / range calibration and drift correction, the outlier rejection and denoising processing of each layer sample, the three-dimensional temperature field is reconstructed, the group distribution of the isothermal consistency index is calculated, the upper limit of each layer distribution is determined by combining the median absolute deviation statistical method, the isothermal consistency critical value from the historical thermal runaway event is introduced, the weighted synthesis is performed according to the scene weight, and the preset isothermal threshold is formed by adding the safety margin and the hysteresis interval; the preset isothermal threshold supports online updating, and the sliding time window and scene recognition trigger recalculation are used, and the freeze / thaw strategy is used when the working condition is switched to avoid frequent jitter; when the sensor fails or the data is incomplete, the factory calibration threshold is used as a fallback and the event is recorded.

[0053] In the embodiment, the non-thermal feature modalities are two types of gas feature data and electrochemical feature data: the gas feature data includes combustible gas concentration, outgassing rate, and gas composition ratio, and is collected by using double-channel constant-flow sampling, combined with cross-sensitivity compensation and temperature and humidity compensation, and screened by using response time and stability thresholds to filter effective segments; the electrochemical feature data includes terminal voltage, working current, and internal resistance change rate, and is collected by using direct-current internal resistance under stable load or specified pulse conditions, extracting alternating current impedance elements within a limited frequency band, and combining OCV-SOC mapping and load normalization to generate comparable feature segments; the two types of non-thermal feature modalities, i.e., the gas feature data and the electrochemical feature data, are aligned with the thermal feature data under a unified time base.

[0054] In the embodiment S2, the local dynamic oversampling method is used to determine the oversampling region based on the change rate of the temperature gradient vector and the abnormal modal trigger signal, temporarily increase the sensor sampling frequency and spatial sampling density in the oversampling region, and capture the fine dynamic characteristics of local temperature changes in the early stage of thermal runaway;

[0055] The specific steps of using the local dynamic oversampling method to capture abnormal changes in the battery hotspot region are as follows:

[0056] The change rate of the temperature gradient vector and the hotspot suspected region are calculated; the sampling frequency is dynamically adjusted and the spatial sampling points are increased in the hotspot suspected region; the data collected by the increased spatial sampling points are oversampling data, the difference in the sliding time window of the oversampling data is calculated, the short-time temperature rise rate is extracted and compared with the reference temperature rise rate; when the short-time temperature rise rate is continuously higher than the reference temperature rise rate for more than a preset duration, it is confirmed that the hotspot suspected region has abnormal temperature changes.

[0057] In the embodiment, the change rate of the temperature gradient vector is a mathematical process for quantifying the temperature field change speed in the time sequence, which is used to identify whether the temperature rise trend of the hotspot region has accelerated signs in the battery thermal runaway early warning scene; the change rate of the temperature gradient vector is calculated by first calculating the temperature gradient vector at each time section, and then comparing the gradient change amplitude between adjacent time sections, specifically: the temperature gradient vector is calculated based on the three-dimensional space temperature field interpolation result at each sampling time; the difference between the lengths of the gradient vectors of adjacent time points is divided by the sampling period to obtain the change rate of the temperature gradient vector.

[0058] In this embodiment, the local dynamic oversampling method is used to temporarily collect high-density data in a specific area when an abnormal trend of thermal feature data is detected, so as to improve the capture accuracy of early thermal runaway signs. The local dynamic oversampling method first calculates the change rate of the temperature gradient vector in the continuous sampling period based on the thermal feature data stream under the unified time base, identifies the spatial range of the hot spot suspected area through the peak value distribution of the change rate and the local clustering method, and combines the trigger signal of the non-thermal feature mode to eliminate false hot spots caused by environmental disturbance or measurement noise, so as to ensure the effectiveness of the positioning of the oversampling area. After determining the oversampling area, the sampling frequency of the temperature sensor or sensor array in the relevant area is temporarily increased, for example, the sampling frequency is increased to twice the original sampling frequency, and additional measurement points are introduced in space. The additional measurement points can be enabled by the adjacent sensor to perform multi-channel measurement or simulated by a virtual measurement point interpolation algorithm, thereby increasing the spatial sampling density.

[0059] The collected oversampling data is subjected to sliding difference operation in a unified time window, a short-time temperature rise rate curve is extracted, and the curve is compared with a pre-calibrated reference temperature rise rate point by point. A continuous time length determination strategy is used to distinguish between incidental disturbance and continuous abnormal temperature rise. If it is determined that the temperature rise is continuous, the area is marked as a high-risk hot spot area. The continuous time length determination strategy is a time threshold value determination strategy for determining whether the short-time temperature rise rate is higher than the reference temperature rise rate.

[0060] In this embodiment S3, the internal and external field temperature difference phase difference refers to the phase difference value of the battery shell surface temperature rise curve and the battery monomer internal temperature rise curve on the time axis, which represents the time sequence shift relationship between the external temperature rise and the internal temperature rise in the battery heat conduction process.

[0061] The internal and external field temperature difference phase difference is calculated using thermal feature data and gas feature data as follows:

[0062] Temperature sampling sequences arranged on the surface of the battery shell and inside the battery monomer are obtained respectively, and time sequence alignment is performed under the same time reference. The aligned temperature sampling sequences are denoised and normalized to obtain the external field temperature change curve and the internal field temperature change curve. The maximum correlation lag time of the external field temperature change curve and the internal field temperature change curve is calculated using the cross-correlation function, and the maximum correlation lag time is normalized to a phase difference value. The change rate of the gas feature data in the same time window is extracted, and the phase difference value is corrected in combination with the gas anomaly occurrence time point to obtain the final internal and external field temperature difference phase difference.

[0063] In the embodiment, the internal-external field temperature difference phase difference is used to quantitatively characterize the time delay characteristic between internal heat generation and external temperature rise in the heat conduction process of the battery, thereby providing a basis for judging the heat source attribute; the physical meaning of the internal-external field temperature difference phase difference lies in that when the heat mainly comes from the external heat source, the temperature rise on the surface of the shell usually occurs earlier than the internal temperature rise; and when the heat mainly comes from the internal self-heating of the battery, the internal temperature rise occurs earlier than the external temperature rise, and the positive and negative and absolute value of the phase difference can directly reflect the difference of the heat conduction path.

[0064] In the embodiment, the cross-correlation function is a mathematical tool for measuring the degree of similarity of two signals at different time delays, which is used to find the best alignment position of the two signals on the time axis, that is, the maximum correlation lag time, thereby quantifying their timing relationship; in the battery thermal runaway early warning scene, the cross-correlation function can be used to compare the external field temperature rise curve and the internal field temperature rise curve to find the time offset of their temperature rise changes, that is, the internal-external field temperature difference phase difference.

[0065] In the embodiment, the calculation process of the internal-external field temperature difference phase difference includes the following detailed steps: according to a unified global time base, the external field temperature sampling sequence and the internal field temperature sampling sequence are obtained from the temperature sensors arranged on the surface of the battery shell and the internal battery monomer respectively, and the timing synchronization and sampling frequency alignment are performed; the wavelet threshold denoising method is used for denoising the sampling sequence to suppress measurement noise and transient disturbance, and then the interval normalization method is used to eliminate the absolute temperature deviation between different measuring points to obtain the external field temperature change curve and the internal field temperature change curve; in the signal processing stage, the cross-correlation function is used to analyze the correlation coefficients of the two temperature change curves at different time delays, to determine the lag time corresponding to the maximum correlation coefficient, and to normalize the lag time and the sampling period to obtain the phase difference value.

[0066] In the embodiment, in order to avoid the phase difference misjudgment caused by short-time gas leakage or abnormal gas release, the gas characteristic data is introduced as a correction factor: the gas concentration change rate is extracted in the same time window, and the gas abnormality occurrence time point is located, the time point is taken as a weight correction parameter to weight and adjust the original phase difference value, thereby obtaining the final internal-external field temperature difference phase difference which can reflect the real timing relationship of the heat source.

[0067] In the embodiment S3, the heat contribution ratio model is a multi-source energy attribution calculation model established based on the thermal characteristic data and the gas characteristic data of the battery, which is used to calculate the heat proportion of the external heat source and the internal self-heating of the battery, and to determine the heat source attribute;

[0068] The heat source attribute determined by combining the heat contribution ratio model is as follows:

[0069] According to the thermal characteristic data, the total heat change amount of the battery external field and the internal field in the target time window is calculated; according to the gas characteristic data, the chemical exothermic amount in the target time window is determined; the chemical exothermic amount is attributed to the internal spontaneous heat component, and the difference between the external field and the internal field heat change amount is attributed to the external heat source component; the ratio of the internal spontaneous heat component to the total heat change amount is calculated as the internal heat contribution rate; when the internal heat contribution rate is higher than the set proportion threshold, it is determined that the heat source attribute is internal spontaneous heat, otherwise it is determined to be an external heat source.

[0070] In the embodiment, the heat contribution ratio model is used to distinguish the relative contribution ratio of internal spontaneous heat and external heat source input to the overall heat change in the early stage of battery thermal runaway, so as to provide targeted response basis for early warning strategy; the heat contribution ratio model uses thermal characteristic data and gas characteristic data to establish a multi-source energy attribution calculation framework, and realizes the quantitative determination of the heat source attribute through the energy conservation relationship and the constraint condition of chemical reaction exothermic amount; in the specific implementation process, first, under the unified time reference, a target time window is selected, and the temperature data of the battery shell surface and the battery monomer inside are integrated and converted to obtain the total heat change amount of the external field and the internal field respectively; according to the gas characteristic data combined with the known chemical reaction heat parameters of the battery, the chemical exothermic amount in the target time window is calculated, and it is directly attributed to the internal spontaneous heat component.

[0071] In the embodiment, the target time window refers to a time range of a preset length extending forward or backward from the initial time point of the risk sign of the battery thermal runaway after the risk sign is detected, which is used to count the thermal characteristic data and the gas characteristic data of the battery in the time range, so as to perform heat attribution calculation; the starting point of the target time window is determined by a certain trigger condition, for example, the isothermal consistency index is lower than the threshold value, the temperature gradient vector change rate exceeds the set value, and the gas concentration is abnormal; the length of the target time window can be fixed or adaptive; the target time window must maintain the time reference uniformity between the multi-modal data; the target time window is a time interval of a preset length extending forward or backward from the starting time point of the trigger event when the battery thermal runaway risk sign is detected, or a time interval adaptively adjusted according to the data change rate, which is used to synchronously count the thermal characteristic data and the gas characteristic data in the time period.

[0072] In the embodiment, in order to distinguish the external heat source action, the difference between the total heat change amount of the external field and the internal field is regarded as the heat component introduced by the external heat source; at the same time, the external heat source component is compensated and corrected according to the environmental temperature change and the heat dissipation condition, so as to avoid overestimation or underestimation caused by environmental fluctuations; the ratio of the internal spontaneous heat component to the total heat change amount is calculated to obtain the internal heat contribution rate; when the contribution rate is higher than the preset proportion threshold, it is determined that the heat source attribute is internal spontaneous heat, otherwise it is determined to be an external heat source; wherein the preset proportion threshold is determined based on experimental calibration or historical data statistics.

[0073] In the present embodiment S4, the cross-modal early response phase difference refers to a multi-modal response time difference quantitative index calculated by a time alignment method and a phase extraction method when the changes of the thermal feature data, the gas feature data and the electrochemical feature data of the battery, and is used to represent the sequence relationship of different physical and chemical processes in the thermal runaway of the battery;

[0074] The specific method of calculating the cross-modal early response phase difference and constructing the phase difference feature matrix is as follows:

[0075] The thermal feature data, the gas feature data and the electrochemical feature data of the battery are preprocessed and time-synchronized under a unified time reference; the feature peak points of each modal signal are extracted; the response time difference between any two modes is calculated and converted into a phase difference value; and the phase difference values between any two modes are filled and constructed into a phase difference feature matrix according to the modal combination order.

[0076] In the present embodiment, the introduction of the cross-modal early response phase difference aims to reveal the sequence relationship and interaction rules between the thermal features, the gas features and the electrochemical features in the thermal runaway process of the battery through time domain quantitative means; in the specific calculation process of the cross-modal early response phase difference, first, the three types of feature data need to be preprocessed under a unified time reference, including denoising, amplitude normalization and baseline drift correction; the time synchronization method of dynamic time warping is used to ensure that the key change nodes of each mode have comparability at the same reference time; in the peak feature point extraction link, the feature peak identification criteria are set according to the physical properties of different modal signals, for example, the thermal feature data can take the time when the first temperature rise rate reaches a set threshold, the gas feature data can take the time when the combustible gas concentration rate peak value appears, and the electrochemical feature data can take the time when the voltage or resistance rate significantly deviates from the steady state; after obtaining the peak points of each mode, the time difference between any two modal peak values is calculated and converted into a phase difference value; these phase difference values are sequentially filled to form a phase difference feature matrix according to the modal combination order, for example, the order of thermal-gas, thermal-electrochemical and gas-electrochemical.

[0077] In the present embodiment S4, the fusion feature matrix is formed by combining the temperature gradient vector, the heat source attribute and the phase difference feature matrix according to a preset feature arrangement order;

[0078] The dynamic weight fusion discriminant model is constructed by a base feature weight adaptive allocation algorithm and a nonlinear probability mapping mechanism, and is used to calculate the thermal runaway risk probability when the battery is in thermal runaway, and the specific method is as follows:

[0079] The weight coefficients of each feature in the fusion feature matrix are dynamically calculated, the weighted fusion vector is calculated by distributing the fusion feature matrix according to the weight coefficients, and the weighted fusion vector is input into a multi-layer nonlinear mapping structure for probability mapping to obtain the thermal runaway risk probability value.

[0080] In the embodiment, the construction of the fusion feature matrix aims to orderly integrate the temperature gradient vector, the heat source attribute and the phase difference feature matrix three types of core features to form a multi-dimensional input set covering spatial distribution features, energy source features and cross-modal time sequence features; the core of the dynamic weight fusion discriminant model is to realize real-time adaptive adjustment of feature weights instead of using fixed weights, so that the feature importance changes caused by the battery operating state, environmental conditions and sensor distribution differences can be coped with; the feature weight adaptive allocation algorithm can be based on the joint statistical results of historical samples and real-time data, and the sensitivity of each feature to the thermal runaway discrimination result is evaluated by using gradient update, entropy weight method or attention mechanism, and the weight coefficients are dynamically allocated accordingly.

[0081] In the embodiment, in the nonlinear probability mapping mechanism part, a multi-layer neural network structure is used to map the weighted fusion vector to the risk probability value in the [0, 1] interval; the nonlinear probability mapping mechanism not only considers the linear superposition relationship between the features, but also captures the high-order interaction between the features through the nonlinear activation function; for example, the joint abnormality of the temperature gradient change and the phase difference feature is often a precursor of thermal runaway, and the nonlinear mapping mechanism can amplify the risk response under this multi-feature combination.

[0082] In embodiment two, the present application proposes a battery thermal runaway early warning system based on multi-dimensional feature fusion, which is applied to the battery thermal runaway early warning method based on multi-dimensional feature fusion proposed in embodiment one, and includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize the battery thermal runaway early warning method based on multi-dimensional feature fusion in embodiment one.

[0083] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge range of those skilled in the art without departing from the purpose of the present application.

Claims

1. A battery thermal runaway early warning method based on multi-dimensional feature fusion, characterized in that, The method comprises the following steps: S1, based on the internal and external field synchronous sampling mechanism, collecting thermal characteristic data, gas characteristic data and electrochemical characteristic data of the battery; S2, calculating the temperature gradient vector and the isothermal consistency index using the thermal characteristic data, and using the local dynamic oversampling method to capture abnormal changes in the hot spot area of the battery when the isothermal consistency index is lower than the preset isothermal threshold and the non-thermal characteristic mode exists abnormally; S3, calculating the internal and external field temperature difference phase difference using the thermal characteristic data and the gas characteristic data, and determining the heat source attribute combined with the heat contribution ratio model to distinguish the external heat source and the battery internal spontaneous heat; The internal and external field temperature difference phase difference refers to the phase difference value of the battery shell surface temperature rise curve and the battery monomer internal temperature rise curve on the time axis, which represents the time sequence shift relationship between the external temperature rise and the internal temperature rise in the battery heat conduction process; The internal and external field temperature difference phase difference is calculated using the thermal characteristic data and the gas characteristic data, specifically as follows: the temperature sampling sequences arranged on the battery shell surface and the battery monomer internal are obtained respectively, and time sequence alignment is performed under the same time reference; the aligned temperature sampling sequences are denoised and normalized to obtain the external field temperature change curve and the internal field temperature change curve; the maximum correlation lag time of the external field temperature change curve and the internal field temperature change curve is calculated using the cross-correlation function, and the maximum correlation lag time is normalized to a phase difference value; the change rate of the gas characteristic data in the same time window is extracted, and the phase difference value is corrected combined with the gas anomaly occurrence time point to obtain the final internal and external field temperature difference phase difference; In S3, the heat contribution ratio model is a multi-source energy attribution calculation model established based on the thermal characteristic data and the gas characteristic data of the battery, which is used to calculate the heat proportion of the external heat source and the battery internal spontaneous heat, and to determine the heat source attribute; The heat source attribute is determined combined with the heat contribution ratio model, specifically as follows: the total heat change amount of the battery external field and internal field in the target time window is calculated according to the thermal characteristic data; the chemical exothermic amount in the target time window is determined according to the gas characteristic data; the chemical exothermic amount is attributed to the battery internal spontaneous heat component, and the heat change difference value of the external field and the internal field is attributed to the external heat source component; the ratio of the internal spontaneous heat component to the total heat change amount is calculated as the internal heat contribution rate; when the internal heat contribution rate is higher than the set proportion threshold, the heat source attribute is determined as internal spontaneous heat, otherwise it is determined as external heat source; S4, based on the thermal characteristic data, the gas characteristic data and the electrochemical characteristic data, calculating the cross-modal early response phase difference and constructing the phase difference feature matrix, and combining the temperature gradient vector, the heat source attribute and the phase difference feature matrix to form a fusion feature matrix, which is input into a dynamic weight fusion discriminant model to calculate the thermal runaway risk probability. 2.The battery thermal runaway early warning method based on multi-dimensional feature fusion of claim 1, characterized in that: In S1, the internal and external field synchronous sampling mechanism is a multi-modal sensor synchronous trigger collection method with unified global time base and time drift compensation between the internal and external of the battery pack, which is used for consistent synchronous sampling of multi-modal data in time domain and spatial domain; The thermal feature data includes battery surface temperature, local temperature gradient and cooling outlet temperature difference; the gas feature data includes combustible gas concentration, gas release rate and gas composition ratio; and the electrochemical feature data includes terminal voltage, working current and internal resistance change rate.

3. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 2, characterized in that: In S2, the isothermal consistency index is used to represent the uniformity of the temperature field of each monitoring position of the battery, and the specific calculation method is as follows: Based on the thermal feature data, the temperature gradient vector is calculated by using the three-dimensional space temperature field interpolation algorithm, and the isothermal consistency index is calculated according to the length distribution of the temperature gradient vector.

4. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 3, characterized in that: In S2, the preset isothermal threshold is calculated by using the thermal feature data of the battery under the rated working condition, and the upper limit of the isothermal consistency index is calculated, and the preset isothermal threshold is determined by combining the critical value of the isothermal consistency index and weighted. The non-thermal feature mode is two types of non-temperature field signal modes of gas feature data and electrochemical feature data; the gas feature data is used to reflect the change of the battery gas release state; and the electrochemical feature data is used to reflect the change of the internal electrochemical reaction and the electrical conductivity of the battery.

5. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 4, characterized in that: In S2, the local dynamic oversampling method is used to determine the oversampling region based on the change rate of the temperature gradient vector and the abnormal mode trigger signal, and temporarily improve the sensor sampling frequency and spatial sampling density in the oversampling region, which is used to capture the subtle dynamic characteristics of local temperature change in the early stage of thermal runaway. The specific steps of using the local dynamic oversampling method to capture abnormal changes of the battery hotspot area are as follows: Calculate the change rate of the temperature gradient vector and the hotspot suspected area; dynamically adjust the sampling frequency and increase the spatial sampling points in the hotspot suspected area; the data collected by the increased spatial sampling points is the oversampling data, calculate the difference in the sliding time window of the oversampling data, extract the short-time temperature rise rate and compare it with the reference temperature rise rate; when the short-time temperature rise rate is continuously higher than the reference temperature rise rate for more than a preset duration, it is confirmed that there is temperature abnormal change in the hotspot suspected area.

6. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 5, characterized in that: In S4, the cross-modal early response phase difference refers to the multi-modal response time difference quantitative index calculated by the time sequence alignment method and the phase extraction method when the thermal feature data, the gas feature data and the electrochemical feature data of the battery change, which is used to represent the sequence relationship of different physical and chemical processes in the battery thermal runaway; The specific method of calculating the cross-modal early response phase difference and constructing the phase difference feature matrix is as follows: Pretreat and time sequence synchronization process the thermal feature data, the gas feature data and the electrochemical feature data of the battery under the unified time reference; extract the feature peak points of each modal signal respectively; calculate the response time difference between any two modalities and convert it into phase difference value; fill and construct the phase difference feature matrix according to the modal combination order.

7. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 6, characterized in that: In S4, the fusion feature matrix is formed by combining the temperature gradient vector, the heat source attribute and the phase difference feature matrix according to the preset feature arrangement order; The dynamic weight fusion discriminant model is constructed by using the base feature weight adaptive allocation algorithm and the nonlinear probability mapping mechanism, which is used to calculate the thermal runaway risk probability when the battery thermal runaway occurs, and the specific method is as follows: The weight coefficients of each feature in the fusion feature matrix are dynamically calculated, a weighted fusion vector is calculated according to the weight coefficients, and the weighted fusion vector is input into a multi-layer nonlinear mapping structure for probability mapping to obtain a thermal runaway risk probability value. 8.A battery thermal runaway early warning system based on multi-dimensional feature fusion, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the battery thermal runaway early warning method based on multi-dimensional feature fusion as claimed in any one of claims 1-7.

Images