Battery thermal runaway prediction method and device based on temperature stratification, equipment and medium

Abstract

The present application relates to the technical field of battery safety control, and particularly relates to a battery thermal runaway prediction method and device based on temperature stratification, equipment and medium, wherein the method comprises: obtaining temperature data of each specified position in the battery pack; determining whether temperature stratification occurs in the battery pack based on the temperature data; if yes, determining whether there is a single out-of-group stratification temperature collection unit, marking an abnormal point to obtain first feature data, and determining whether there is a group stratification type stratification area, filtering data to obtain second feature data; calculating the proportion of the battery pack that does not appear temperature stratification under extreme temperature to obtain third feature data; and predicting the battery thermal runaway probability based on the three types of feature data. The method can accurately identify early temperature abnormalities, improve the accuracy and reliability of thermal runaway prediction, and ensure battery safety.

Description

Technical Field

[0001] This invention relates to the field of battery safety control technology, and in particular to a method, apparatus, device, and medium for predicting battery thermal runaway based on temperature stratification. Background Technology

[0002] With the widespread application of lithium-ion batteries in electric vehicles and energy storage systems, their safety and reliability have become paramount to ensuring stable system operation. Batteries inevitably generate heat during operation. However, uneven heat dissipation or internal faults in individual cells (such as abnormally increased internal resistance or micro-short circuits) can easily lead to significant localized temperature increases, creating an abnormal temperature gradient distribution within the battery pack. This significant temperature difference between localized areas and surrounding areas constitutes the so-called "temperature stratification" phenomenon. This phenomenon is one of the key early signs of potential thermal runaway risks within the battery. This subtle temperature stratification often occurs before an overall temperature rise or significant voltage anomaly, serving as an early warning signal of an impending serious thermal runaway accident.

[0003] Existing technologies struggle to identify subtle temperature stratification within batteries in a timely manner, often relying on overall temperature increases or voltage fluctuations. This lack of sensitivity means that group stratification can easily mask individual cell anomalies, leading to inaccurate fault location and low accuracy in predicting thermal runaway. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and medium for predicting battery thermal runaway based on temperature stratification, aiming to solve the technical problem of low accuracy in predicting early-stage thermal runaway of batteries in existing technologies.

[0005] To achieve the aforementioned objective, the first aspect of this invention proposes a battery thermal runaway prediction method based on temperature stratification, the method comprising:

[0006] Temperature data at designated locations within the battery pack are acquired through multiple temperature acquisition units, wherein the temperature data is data within a battery thermal runaway prediction cycle;

[0007] Based on the temperature data, analyze whether temperature stratification has occurred within the battery pack. Temperature stratification refers to the temperature difference between a local area and the surrounding area within the battery pack reaching a preset condition.

[0008] If temperature stratification has occurred, based on the data corresponding to the occurrence of temperature stratification, it is determined whether there is a single outlier temperature acquisition unit. If so, the temperature acquisition unit is marked as an anomaly, and first feature data is obtained. Here, the single outlier refers to a temperature acquired by a single temperature acquisition unit that deviates from the temperatures acquired by other nearby temperature acquisition units, reaching a first deviation threshold.

[0009] Determine whether there is a group-stratified region. If so, filter out the temperature stratification data of that part to obtain the second feature data. The group-stratified region refers to a temperature region formed by multiple adjacent temperature acquisition units acquiring similar temperatures, and the temperature of this temperature region deviates from the temperature of other regions and reaches a second deviation threshold.

[0010] The percentage of battery packs that did not exhibit temperature stratification under extreme temperatures was statistically analyzed to obtain the third characteristic data, where extreme temperatures refer to temperatures exceeding the normal operating temperature range of the battery.

[0011] Based on the first feature data, the second feature data, and the third feature data, the probability of thermal runaway of the battery is predicted.

[0012] Further, before predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data, the following steps are included:

[0013] Based on the first feature data, the second feature data, and the third feature data, multiple temperature anomaly detection indicators are constructed, including:

[0014] Year-on-year temperature difference at the end of the charging cycle: The change in temperature difference between each temperature acquisition unit / area of ​​the battery pack at the end of the current charging cycle and the same period of the previous month.

[0015] High-temperature temperature rise rate: The rate at which the temperature rises, as measured by the temperature acquisition unit, during the charging process in an environment exceeding the first temperature.

[0016] Low-temperature charging recovery rate: the rate at which the temperature collected by the temperature acquisition unit recovers to the normal operating temperature during the charging process in an environment below the second temperature; wherein, the first temperature is greater than the second temperature;

[0017] If multiple temperature anomaly detection indicators exceed preset over-limit thresholds, then proceed to the step of predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data.

[0018] Further, predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data includes:

[0019] Based on the first feature data, calculate the temperature deviation index of the temperature collected by the abnormal temperature acquisition unit.

[0020] Based on the second feature data, assess the degree of interference of population stratification on the overall temperature distribution;

[0021] Based on the third feature data, the temperature consistency of the battery pack under extreme temperature conditions is analyzed.

[0022] Based on the temperature deviation index, interference level, and temperature consistency, a thermal runaway risk assessment model is constructed to predict the probability of thermal runaway of the battery.

[0023] Furthermore, by comprehensively considering the temperature deviation index, interference level, and temperature consistency, a thermal runaway risk assessment model is constructed to predict the probability of thermal runaway of the battery, including:

[0024] Based on the temperature deviation index, calculate the Z-score value, probability value, and slope of the temperature collected by each abnormal temperature acquisition unit.

[0025] Based on the degree of interference, determine the proportion of remaining valid data after data filtering in the group-stratified stratified region;

[0026] Based on the temperature consistency, the deviation between the percentage of unstratified areas under extreme temperatures and historical data for the same period is obtained.

[0027] The Z-score value, probability value, slope of the Z-score, percentage of valid data, and deviation value are input into a preset risk assessment model. A comprehensive thermal runaway risk score is generated through weighted calculation, and the probability of thermal runaway of the battery is predicted based on the score.

[0028] Furthermore, the risk assessment model is as follows:

[0029]

[0030] In the formula, Z is the Z-score value, p is the probability value, R is the proportion of valid data, k is the slope of the Z-score, and ΔP is the bias value. , , , , The weighting coefficients are and satisfy the following conditions: + + + + = 1, calibrated using historical thermal runaway data.

[0031] Furthermore, the risk assessment model is a deep learning neural network model, which assesses the risk of thermal runaway using the following dynamic weight recursive formula:

[0032]

[0033] in, is the normalized feature vector; Z is the Z-score value, p is the probability value, R is the proportion of valid data, k is the slope of the Z-score, and ΔP is the bias value;

[0034] These are dynamic weights that are updated during training iterations. For adaptive learning rate, satisfy L is the loss function. , The model predicts probabilities, and α and γ are hyperparameters that balance positive and negative samples;

[0035] b(t) is the bias term in the neural network model that is dynamically updated during the training iteration process;

[0036] t represents the number of iterations for model training;

[0037] This is the Sigmoid activation function.

[0038] Furthermore, the step of analyzing whether temperature stratification has occurred within the battery pack based on the temperature data includes:

[0039] Obtain temperature sub-data collected by each temperature acquisition unit in the same frame of the battery pack from the temperature data, and calculate the temperature difference sequence of temperatures collected by adjacent temperature acquisition units.

[0040] Perform wavelet transform on the temperature difference sequence to obtain wavelet coefficients;

[0041] Based on the distribution of the modulus maxima of the wavelet coefficients, determine the location and amplitude of the temperature abrupt change;

[0042] When there is at least one temperature abrupt change location, and the corresponding temperature abrupt change amplitude is greater than a preset threshold, and in conjunction with the spatial structure of the battery pack, it is determined that temperature stratification has occurred within the battery pack.

[0043] Furthermore, determining whether temperature stratification exists within the battery pack based on the temperature data includes:

[0044] Acquire temperature sub-data for each location in the battery pack within the same frame of the temperature data, and calculate the temperature difference between adjacent temperature acquisition points;

[0045] Calculate the standard deviation of the temperature difference based on the temperature difference between the adjacent temperature collection points;

[0046] The standard deviation of the temperature difference is compared with a preset threshold. If the standard deviation of the temperature difference is greater than the preset threshold, it is determined that temperature stratification has occurred in the battery pack.

[0047] A second aspect of the present invention provides a battery thermal runaway prediction device based on temperature stratification, comprising:

[0048] The acquisition unit is used to acquire temperature data at various designated locations within the battery pack through multiple temperature acquisition units, wherein the temperature data is data within a battery thermal runaway prediction cycle;

[0049] The judgment unit is used to analyze whether temperature stratification has occurred in the battery pack based on the temperature data. Temperature stratification refers to the temperature difference between a local area and the surrounding area in the battery pack reaching a preset condition.

[0050] The first acquisition unit is used to determine whether there is a single outlier temperature acquisition unit if temperature stratification has occurred. If there is, the temperature acquisition unit is marked as an outlier and the first feature data is obtained. The single outlier refers to the temperature acquired by a single temperature acquisition unit deviating from the temperature acquired by other nearby temperature acquisition units and reaching a first deviation threshold.

[0051] The second acquisition unit is used to determine whether there is a group-layered stratification region if temperature stratification has occurred. If it exists, the temperature stratification data of this part is filtered out to obtain the second feature data. The group-layered stratification refers to the temperature of multiple adjacent temperature acquisition units that are similar to form a temperature region, and the temperature of this temperature region deviates from the temperature of other regions and reaches the second deviation threshold.

[0052] The third acquisition unit is used to count the percentage of the battery pack that does not exhibit temperature stratification under extreme temperatures, and obtain the third feature data, wherein the extreme temperature refers to the temperature that exceeds the normal operating temperature range of the battery.

[0053] The prediction unit is used to predict the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data.

[0054] A third aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the battery thermal runaway prediction method based on temperature stratification as described in any of the preceding claims.

[0055] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the battery thermal runaway prediction method based on temperature stratification as described in any of the preceding claims.

[0056] Beneficial effects:

[0057] This invention discloses a battery thermal runaway prediction method, apparatus, device, and medium based on temperature stratification. It can identify subtle temperature stratification phenomena within the battery, known as "temperature stratification," a key early sign of potential thermal runaway risk. Compared to existing technologies that rely on overall temperature rise or voltage fluctuations, this method is more sensitive to identifying early temperature anomalies. It can capture early warning signals of impending severe thermal runaway accidents before significant overall temperature rise or voltage anomalies occur, providing more time to prevent such accidents. When temperature stratification exists, it can accurately determine whether there are outlier battery cells with stratification and mark them as anomalies. Simultaneously, it can identify and filter data related to group-based stratification regions. This avoids the problem of group stratification masking individual cell anomalies and accurately locates faulty cells, solving the problem of inaccurate fault location in traditional methods and facilitating targeted maintenance. Unlike traditional threshold-based judgment methods, this method reduces environmental noise interference and lowers false alarm and false negative rates through analysis and processing of different types of temperature stratification, achieving accurate identification of abnormal cells and improving the reliability of thermal runaway prediction. By combining the first feature data (outlier stratified anomaly data of individual cells), the second feature data (data after filtering out stratified regions of the population), and the third feature data (the proportion of cells that do not exhibit temperature stratification under extreme temperatures), the probability of battery thermal runaway is predicted. This method considers multiple dimensions, including individual cell anomalies, elimination of population interference, and performance under extreme temperatures, making the prediction results more comprehensive and reliable. This provides a strong guarantee for the safe and stable operation of batteries in electric vehicles and energy storage systems. Attached Figure Description

[0058] Figure 1 A schematic flowchart illustrating a battery thermal runaway prediction method based on temperature stratification according to an embodiment of the invention;

[0059] Figure 2 A schematic diagram of temperature stratification in a battery pack according to an embodiment of the invention;

[0060] Figure 3 This is a schematic diagram of a battery thermal runaway prediction device based on temperature stratification according to an embodiment of the invention.

[0061] Figure 4 This is a schematic diagram of the structure of a computer device according to an embodiment of the invention.

[0062] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0064] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of features, integers, steps, operations, elements, modules, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, modules, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any modules and all combinations of one or more associated listed items.

[0065] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0066] Reference Figure 1 This invention provides a battery thermal runaway prediction method based on temperature stratification, comprising the following steps S1-S6:

[0067] S1: Temperature data at designated locations within the battery pack are acquired through multiple temperature acquisition units; wherein, the temperature data is the data within a battery thermal runaway prediction cycle, and the designated locations include individual battery cells or a local area.

[0068] The aforementioned temperature data refers to the temperature values ​​of individual battery cells or localized areas within the battery pack, collected by temperature acquisition units such as temperature sensors or thermal imagers. A battery thermal runaway prediction cycle refers to the time difference between the last predicted battery thermal runaway time and the current predicted battery thermal runaway time. This cycle is usually manually set, such as one week or one month, and can be set according to the battery usage scenario and safety requirements. During the operation and charging / discharging of an electric vehicle, temperature sensors deployed at different locations within the power battery pack are used to collect the temperature values ​​of each individual battery cell and localized areas composed of several battery cells in real time, for example, collecting data every 10 seconds to form a temperature data sequence. For instance, an electric vehicle power battery pack has 100 battery cells, each equipped with a temperature sensor. Furthermore, every 10 battery cells are divided into localized areas, and regional temperature acquisition points are also set up. This allows for the acquisition of temperature data for 100 individual cells and 10 regional temperature data. This temperature data provides fundamental data support for subsequent determination of whether temperature stratification has occurred within the battery pack and for thermal runaway prediction, ensuring a reliable data source for subsequent analysis.

[0069] S2: Analyze whether temperature stratification has occurred within the battery pack based on the temperature data. Temperature stratification refers to the temperature difference between a local area and the surrounding area within the battery pack reaching a preset condition.

[0070] Temperature stratification occurs when a pre-defined temperature difference condition is met, indicating an abnormal gap or uneven temperature distribution within the battery pack. This manifests as a significant temperature difference between certain areas and their surroundings. Analyzing the collected temperature data involves methods such as calculating the standard deviation of temperature differences between adjacent data points. When the standard deviation exceeds a pre-defined threshold, temperature stratification is considered to have occurred. For example, for the temperature data of the aforementioned electric vehicle battery pack, the temperature difference between adjacent battery cells is calculated, and then the standard deviation of these temperature differences is calculated. If the pre-defined threshold is 2℃, and the calculated standard deviation is 3℃, exceeding the threshold, then temperature stratification is determined to have occurred within the battery pack. Identifying abnormal temperature distribution within the battery pack provides a prerequisite for further analysis of temperature stratification types and for predicting thermal runaway.

[0071] S3: If temperature stratification has occurred, then based on the data corresponding to the occurrence of temperature stratification, determine whether there is a single outlier temperature acquisition unit. If so, mark the temperature acquisition unit as an anomaly and obtain the first feature data. The single outlier refers to the temperature acquired by a single temperature acquisition unit deviating from the temperature acquired by other nearby temperature acquisition units and reaching a first deviation threshold.

[0072] The aforementioned first feature data consists of relevant data after anomalies are identified, including the temperature of the abnormal temperature acquisition unit and information about the corresponding battery cell. For cases with temperature stratification, the difference between the temperature collected by each temperature acquisition unit and the average temperature collected by neighboring temperature acquisition units is analyzed. When the difference reaches a first deviation threshold (e.g., 3°C), the temperature acquisition unit is considered an outlier and is marked as an anomaly. The temperature data of these anomalies is collected as the first feature data. It is important to note that within a battery thermal runaway prediction cycle, the first feature data is a collection of temperature data corresponding to different anomalies. For example, in an electric vehicle battery pack with temperature stratification, each temperature acquisition unit is examined, and the difference between the temperature collected by that unit and the average temperature of the three surrounding temperature acquisition units is calculated. If a temperature acquisition unit collects a temperature of 50°C, and the average temperature collected by the three surrounding units is 45°C, the difference of 5°C is greater than the deviation threshold of 3°C. Therefore, this temperature acquisition unit is considered an outlier and is marked as an anomaly, with its temperature data included as part of the first feature data. The temperature acquisition units that accurately locate temperature anomalies are often the individual cells corresponding to these anomaly points, which are potential risk points for thermal runaway. This provides key abnormal feature data for subsequent thermal runaway prediction.

[0073] S4: Determine whether there is a group-layered stratified region. If so, filter out the temperature stratified data of this part to obtain the second feature data. The group-layered type refers to the temperature collected by multiple adjacent temperature acquisition units being similar to form a temperature region, and the temperature of this temperature region deviates from the temperature of other regions and reaches the second deviation threshold.

[0074] The second feature data mentioned above is the effective temperature data remaining after filtering out the temperature data of the group-stratified stratified regions. Observing the temperature distribution collected by each temperature acquisition unit, if multiple adjacent temperature acquisition units collect temperatures that are consistent in distribution and degree of temperature difference, a group-stratified stratified region is formed. The temperature stratification data of these regions is filtered out, and the remaining temperature data is used as the second feature data. It is important to note that within a battery thermal runaway prediction cycle, the second feature data is a collection of data generated at different temperature stratifications. For example, in the aforementioned electric vehicle power battery pack, there are three regions where multiple temperature acquisition units within each region collect temperatures that are basically consistent, but there are significant differences in temperatures collected by temperature acquisition units in other regions, and the temperature difference patterns are similar. This belongs to a group-stratified stratified region. The temperature stratification data of these three regions is filtered out, and the temperature data of the remaining battery cells and regions corresponding to the temperature acquisition units are used as part of the second feature data. Excluding group-stratified stratified regions, which may be affected by temperature stratification interference caused by non-thermal runaway risk factors (such as the influence of environmental uniformity), makes the second feature data more reflective of the actual temperature situation related to thermal runaway risk.

[0075] In a specific embodiment, refer to Figure 2 This is a battery pack (88 individual cells) equipped with 28 temperature acquisition units to collect temperature data at designated locations. During a single charge, a temperature stratification diagram is displayed. Specifically, the vertical axis represents temperature, and the horizontal axis represents frames. The first frame shows the data at the start of charging. As charging time increases (the number of frames increases), the temperature gradually rises. For example, around frame 166, the temperature collected by all 28 acquisition units rises to approximately 40 degrees Celsius. At this point, the liquid cooling system begins to activate. During the activation of the liquid cooling system, the temperature gradient is visible... Figure 2 The battery pack exhibits temperature stratification (frames 171-231). When the battery pack reaches a preset temperature, the liquid cooling system stops operating. As the battery pack temperature rises again, the liquid cooling system re-enters, and temperature stratification occurs again (frames 261-281). These two temperature stratification events occur from... Figure 2 This can be seen as a group phenomenon, which can be considered as group stratification. Furthermore, in these two temperature stratifications, we observed a significant outlier at the bottom line, which represents a single, outlier temperature acquisition unit.

[0076] S5: Calculate the percentage of battery packs that do not exhibit temperature stratification under extreme temperatures to obtain the third characteristic data.

[0077] The aforementioned extreme temperatures refer to temperatures exceeding the normal operating temperature range of the battery, such as excessively high (above 40°C) or excessively low (below 5°C). The third characteristic data is the percentage of battery packs that did not exhibit temperature stratification under extreme temperatures. Within a battery thermal runaway prediction cycle, the ratio of the number of times temperature stratification did not occur in the battery pack during charging under extreme temperature conditions (the overall temperature of the battery pack) to the sum of the number of times temperature stratification occurred and did not occur during charging under extreme temperature conditions is the percentage of battery packs that did not exhibit temperature stratification, which serves as the third characteristic data. Specifically, when analyzing whether temperature stratification occurred within the battery pack based on the aforementioned temperature data, if temperature stratification did occur, the number of times each temperature stratification occurred under extreme temperature conditions is counted (the number of times temperature stratification occurred in the battery pack during charging under extreme temperature conditions). Then, the total number of times the battery was charged under extreme temperature conditions (the sum of the number of times temperature stratification occurred and did not occur) is counted. Finally, the difference between the two (the number of times temperature stratification did not occur in the battery pack during charging under extreme temperature conditions) is calculated, thus yielding the calculated ratio. For example, if the battery thermal runaway prediction cycle is set to 1 week (7 days), and the charging process is monitored once a day under extreme temperature conditions (high temperature conditions in summer, with the overall battery pack temperature reaching 50℃), a total of 7 charging processes are monitored within 1 week. Among them, temperature stratification of the battery pack is detected in 3 charging processes and no temperature stratification is detected in 4 charging processes. Then the percentage of no temperature stratification under extreme temperature conditions = number of times no temperature stratification is detected / (number of times it occurs + number of times it does not occur) = 4 / (3 + 4) ≈ 57.1%, and this ratio is used as the third feature data. The lower the percentage of no temperature stratification under extreme temperature conditions, the more unstable the temperature distribution of the battery pack is during charging under extreme temperature conditions, and the higher the risk of thermal runaway. By replacing "cell / region number statistics" with "number of times within the cycle", the "spatial distribution data" is avoided from being mistakenly used as a "proportion indicator in the time dimension", making the third feature data more reflective of the long-term stability of the battery pack under extreme temperature charging scenarios and improving the accuracy of thermal runaway prediction.

[0078] S6: Based on the first feature data, the second feature data, and the third feature data, predict the probability of thermal runaway of the battery.

[0079] The first, second, and third feature data are input into a thermal runaway risk assessment model, which calculates the probability of thermal runaway of the battery. For example, inputting the first feature data (temperature collected by the abnormal temperature acquisition unit, etc.), the second feature data (filtered effective temperature data), and the third feature data (70% proportion) into the deep learning neural network model shown below, the model calculates and outputs a thermal runaway probability of 0.35. By comprehensively and accurately predicting the possibility of battery thermal runaway based on multiple feature data, a scientific basis for battery safety management can be provided, facilitating the early implementation of preventive measures.

[0080] This embodiment of the temperature-stratified battery thermal runaway prediction method acquires different types of feature data step by step, analyzes the battery pack temperature from multiple dimensions such as individual temperature acquisition unit anomalies, elimination of group interference, and extreme temperature stability, and then predicts the probability of thermal runaway based on this feature data. Compared with existing thermal runaway prediction methods that only consider a few temperature factors (such as focusing only on excessively high individual cell temperatures), this method considers more comprehensive factors, covering different types of temperature stratification and performance under extreme temperatures. It can more accurately predict the probability of battery thermal runaway, effectively improve the scientific nature and foresight of battery safety management, and reduce the risk of thermal runaway accidents.

[0081] In one implementation, before predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data, the following steps are included:

[0082] Based on the first feature data, the second feature data, and the third feature data, multiple temperature anomaly detection indicators are constructed, including but not limited to:

[0083] Year-on-year temperature difference at the end of the charging cycle: The temperature difference of each temperature acquisition unit / area of ​​the battery pack at the end of the current charging cycle compared with the temperature difference in the same period of the previous month. The calculation formula is "(current temperature difference - temperature difference in the same period of the previous month) / historical temperature difference in the same period × 100%".

[0084] High-temperature temperature rise rate: Under high-temperature environment (such as the overall temperature of the battery pack ≥ 40℃), the rate of temperature rise collected by the temperature acquisition unit during the charging process is monitored. The calculation formula is "(temperature at time t2 - temperature at time t1) / (t2 - t1)" (unit: ℃ / min).

[0085] Low-temperature charging recovery rate: In a low-temperature environment (such as the overall temperature of the battery pack ≤ 5℃), the rate at which the temperature collected by the temperature acquisition unit rises back to the normal operating temperature (such as 25℃) after charging is completed. The calculation formula is "(25℃ - charging start temperature) / recovery time" (unit: ℃ / min).

[0086] If all three indicators exceed the preset over-limit thresholds (e.g., the year-on-year temperature difference at the end of the charging month ≥ 20%, the high temperature rise rate ≥ 0.5℃ / min, and the low temperature recovery rate ≤ 0.2℃ / min), then proceed to step S6 above.

[0087] The aforementioned year-on-year temperature difference index for the last month of charging refers to the temperature difference (or percentage change) obtained by comparing the difference between the highest and lowest temperatures during all charging processes in the current month (this month's temperature difference value) with the difference between the highest and lowest temperatures during all charging processes in the previous month (last month's temperature difference value). At the end of each month, the highest and lowest temperatures of the battery pack during each charging process are statistically analyzed. The temperature difference for each charging process is calculated, and the maximum value among all temperature differences for the current month (or the overall difference between the highest and lowest temperatures during all charging processes in the current month is directly calculated) is taken as the temperature difference value for the current month. At the same time, the temperature difference value for the previous month is retrieved, and the year-on-year change is calculated using "(this month's temperature difference value - last month's temperature difference value) / last month's temperature difference value × 100%". If the percentage change exceeds a preset threshold (e.g., 20%), the index is considered to have exceeded the limit abnormally. For example, during all charging processes in March 2024 (this month), the highest temperature of the battery pack was 50℃ and the lowest temperature was 10℃, with a temperature difference of 50-10=40℃ for this month; the temperature difference in February 2024 (last month) was 32℃. Therefore, the year-on-year change in temperature difference at the end of the charging month is (40-32) / 32×100%=25%, exceeding the 20% threshold, indicating an abnormal exceedance of this indicator.

[0088] The aforementioned high-temperature temperature rise rate refers to the rate of temperature increase during each charge under all high-temperature conditions (e.g., battery pack temperature ≥ 40℃) within a battery thermal runaway prediction cycle (e.g., 1 week). This forms a rate dataset (i.e., the temperature rise rate for each charge needs to be statistically analyzed and aggregated for subsequent judgment). Within the prediction cycle, each time the battery pack is detected charging under high-temperature conditions, the temperature change over time is recorded in real time, and the temperature rise rate for each charge is calculated (e.g., temperature is recorded every 5 minutes, and the temperature rise rate at adjacent time points is calculated). After the cycle ends, the temperature rise rate of all high-temperature charging processes is statistically analyzed. If more than 80% of the rate values ​​are greater than a preset threshold (e.g., 0.5℃ / min), the indicator is considered abnormally exceeded. For example, if the prediction cycle is 1 week, and there are 5 high-temperature charging processes, the calculated temperature rise rates for each are 0.6℃ / min, 0.4℃ / min, 0.7℃ / min, 0.5℃ / min, and 0.8℃ / min (dataset). Four of the rate values ​​were ≥0.5℃ / min, accounting for 80%, which reached the threshold and indicated that the indicator was abnormally exceeded.

[0089] The aforementioned recovery rate after low-temperature charging refers to the rate at which the battery temperature recovers from the initial low temperature to the standard temperature (25℃) after charging begins, under all low-temperature conditions (e.g., battery pack temperature ≤ 5℃) within a prediction cycle. If the temperature exceeds 25℃ during the recovery process, only the recovery rate from the initial temperature to 25℃ is calculated; the portion exceeding 25℃ is not included in the calculation. Within the prediction cycle, for each low-temperature charging, the initial temperature (e.g., 5℃) and the time it takes for the temperature to recover to 25℃ are recorded. If the temperature does not exceed 25℃ during charging, the rate is calculated based on the actual recovery time; if it exceeds 25℃, only the time from the initial temperature to 25℃ is counted. After the cycle ends, if the average recovery rate of all low-temperature charging is lower than a preset threshold (e.g., 0.2℃ / min), the indicator is considered to have exceeded the limit abnormally. For example, if the prediction period is one week, and there are two low-temperature charging sessions: the first session starts at 5℃ and takes 100 minutes to rise to 25℃, with a rate of (25-5) / 100 = 0.2℃ / min; the second session starts at 3℃ and takes 110 minutes to rise to 25℃, with a rate of (25-3) / 110 ≈ 0.2℃ / min. The average rate is 0.2℃ / min, which reaches the lower limit of the threshold, indicating that this indicator is abnormally exceeded.

[0090] In this embodiment, the "multi-index screening" pre-process avoids unnecessary individual outlier identification processes triggered by a single temperature stratification (such as slight stratification caused by short-term environmental fluctuations), thus reducing the computational burden on the system. At the same time, the three indicators comprehensively assess the battery health status from the dimensions of "long-term temperature difference trend", "high temperature overheating risk" and "low temperature performance degradation". Only when multiple dimensions are abnormal will in-depth analysis be performed, thereby improving the pertinence and accuracy of thermal runaway prediction.

[0091] In one embodiment, predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data includes:

[0092] S61: Based on the first feature data, calculate the temperature deviation index of the temperature collected by the abnormal temperature acquisition unit.

[0093] The aforementioned temperature deviation index is used to quantify the difference between the temperature collected by the abnormal temperature acquisition unit and the surrounding normal temperature. Using the temperature collected by the abnormal temperature acquisition unit and the temperatures of surrounding units from the first feature data, the difference or ratio between the temperature collected by the abnormal temperature acquisition unit and the average temperature collected by the surrounding temperature acquisition units is calculated as the temperature deviation index. For example, in the above embodiment, the temperature collected by the abnormal temperature acquisition unit is 50℃, and the average temperature collected by the three surrounding temperature acquisition units is 45℃. The calculated difference is 5℃, and this difference is the temperature deviation index of the temperature collected by the abnormal temperature acquisition unit. Quantifying the temperature deviation collected by the abnormal temperature acquisition unit allows subsequent models to more accurately use this feature for thermal runaway prediction. Compared to qualitative judgment, quantitative indicators are more scientific and operable.

[0094] S62: Based on the second feature data, assess the degree of interference of population stratification on the overall temperature distribution.

[0095] The aforementioned interference level of group stratification on the overall temperature distribution refers to the impact of temperature stratification data from group-stratified regions on the analysis of the overall temperature distribution of the battery pack. Based on the second feature data (the remaining valid data after filtering out temperature stratification data from group-stratified regions), the changes in relevant statistics of the overall temperature distribution (such as temperature variance) before and after filtering are calculated. A larger change indicates a higher degree of interference. For example, before filtering temperature stratification data from group-stratified regions, the overall temperature variance of the battery pack was 8; after filtering, it was 5, a change of 3, indicating a high degree of interference from group stratification on the overall temperature distribution. Quantifying the interference effect of group stratification allows subsequent thermal runaway risk assessment models to reasonably weigh this factor, improving the accuracy of the model's thermal runaway prediction.

[0096] S63: Based on the third feature data, analyze the temperature consistency of the battery pack under extreme temperature conditions.

[0097] The temperature uniformity of the battery pack under the above extreme temperature conditions refers to the degree of temperature uniformity of each part of the battery pack under extreme temperature conditions.

[0098] The third characteristic data is the percentage of areas without temperature stratification under extreme temperatures. A higher percentage indicates that more areas maintain normal temperatures under extreme conditions, resulting in better temperature consistency. Other temperature statistics (such as the standard deviation of temperatures under extreme temperatures) can also be used to assist in the analysis. For example, in the above embodiment, the percentage of areas without temperature stratification under extreme temperatures was 57.1% (4 times without stratification and 3 times within a week). Combined with the corrected abnormal indicator judgment results: a 25% year-on-year increase in temperature difference at the end of the charging month (exceeding the limit), a high-temperature temperature rise rate of 80% ≥ 0.5℃ / min (exceeding the limit), and an average low-temperature recovery rate of 0.2℃ / min (exceeding the limit), all three are abnormal. This indicates that the battery pack not only has poor temperature consistency but also suffers from worsening monthly temperature differences, a significant trend of high-temperature overheating, and insufficient low-temperature recovery capability. A thermal runaway risk assessment model needs to be constructed by comprehensively considering four types of indicators (temperature deviation, interference level, temperature consistency, and corrected abnormal indicators) to improve the comprehensiveness of the prediction. This reflects the overall temperature performance of the battery pack under extreme conditions. Poor temperature consistency under extreme temperatures is often closely related to the risk of thermal runaway, providing important overall characteristics for thermal runaway prediction.

[0099] S64: Combining the temperature deviation index, interference level, and temperature consistency, construct a thermal runaway risk assessment model to predict the probability of thermal runaway of the battery.

[0100] By using temperature deviation, disturbance, and temperature consistency indicators as inputs, a thermal runaway risk assessment model, such as a weighted model or a deep learning model, is constructed. The thermal runaway probability is obtained through model training and calculation. For example, the calculated temperature deviation indicator (5℃), disturbance indicator (variance change of 3), and temperature consistency indicator (70% proportion and 4℃ standard deviation) are input into a simple weighted model. The model calculates a thermal runaway probability of 0.38 based on weights assigned according to historical thermal runaway data. By comprehensively considering multiple dimensions of indicators, including temperature anomalies collected by individual temperature acquisition units, group disturbances, and overall performance under extreme temperatures, the constructed model can more accurately predict the thermal runaway probability, providing strong support for battery safety decisions.

[0101] In this embodiment, the first, second, and third feature data are further analyzed in detail to obtain more targeted quantitative indicators. Then, a thermal runaway risk assessment model is constructed based on these indicators. Compared with existing methods that directly utilize raw feature data for simple model prediction, this method, through in-depth analysis of feature data and multi-indicator fusion, makes the features input to the model more representative and discriminative, thereby improving the accuracy and reliability of thermal runaway probability prediction and facilitating more precise battery thermal runaway risk management.

[0102] In one embodiment, the thermal runaway risk assessment model is constructed by combining the temperature deviation index, interference level, and temperature consistency to predict the thermal runaway probability S64 of the battery, including:

[0103] S641: Based on the temperature deviation index, calculate the Z-score value, probability value, and slope of the temperature collected by each abnormal temperature acquisition unit.

[0104] The Z-score mentioned above is a standardized statistic used to measure the degree of deviation of a data point from the mean of the dataset. It is calculated using Z = (data point temperature - mean temperature) / standard deviation. The probability value is the probability of the data point occurring, calculated based on the Z-score using a normal distribution with equal probability. The slope of the Z-score reflects the dynamic trend of the temperature deviation. The slope of the Z-score is calculated as the rate of change per unit time for the Z-score sequence. Temperature data collected by the temperature acquisition units during normal battery pack operation is collected, and their mean temperature and standard deviation are calculated. For temperatures collected by abnormal temperature acquisition units (data point temperatures), the Z-score value is calculated using the Z-score formula, and then the corresponding probability value is obtained according to the standard normal distribution table or related probability calculation methods. For example, if the mean temperature collected by each temperature acquisition unit during normal battery pack operation is 40℃ and the standard deviation is 3℃, and the temperature collected by the abnormal temperature acquisition unit is 50℃, then Z = (50-40) / 3 ≈ 3.33. Consulting the standard normal distribution table, a Z-score of 3.33 corresponds to a probability of approximately 0.0004, meaning the probability of this abnormal temperature occurring is extremely low. For the Z-score sequence Z=[Z... _1 Z _2 ,...,Z _m To calculate the rate of change (slope) per unit time, this embodiment uses a sliding window linear fitting method: select a time window [t, t+Δt] (e.g., Δt = 5 minutes), and calculate the Z-score value Z within the window. _t Z _t+1 ,...,Z_ t+Δt Perform linear regression to obtain the slope k Z The formula can be simplified to (approximate difference between adjacent time steps): k Z =(Z t+Δt -Z _t k / Δt, where Δt is the window time length. ZThe larger the value, the faster the "degree of temperature deviation from normal" increases, and the more significant the trend of thermal runaway risk. Z-score standardization eliminates the influence of differences in the mean and variance of temperatures under different battery packs or operating conditions, making the deviation of temperatures collected by the abnormal temperature acquisition unit more comparable. The probability value reflects the rarity of abnormal temperatures from a statistical probability perspective. Rare abnormal temperatures are often highly correlated with the risk of thermal runaway, providing more accurate statistical characteristics for thermal runaway prediction. The slope of the Z-score indicates the degree of temperature deviation from normal, also providing more accurate statistical characteristics for thermal runaway prediction.

[0105] S642: Based on the degree of interference, determine the proportion of remaining valid data after data filtering in the group-stratified stratified region.

[0106] The aforementioned percentage of remaining valid data refers to the ratio of the number of valid temperature data points remaining after filtering out the temperature stratification data from stratified areas within a group to the total number of data points. The percentage of remaining valid data is calculated by counting the total number of temperature data points for the battery pack (including temperature data collected by each temperature acquisition unit and regional temperature data), and then counting the number of valid data points remaining after filtering out the temperature stratification data from stratified areas within a group. For example, if the total number of temperature data points for an electric vehicle battery pack is 110 (100 individual cells + 10 regions), after filtering out the 30 temperature stratification data points from stratified areas within a group, the number of valid data points remaining is 80. Therefore, the percentage of remaining valid data is 80 / 110 ≈ 72.7%. This quantifies the impact of stratified areas within a group on data validity. A lower percentage of remaining valid data indicates more severe group stratification interference and a potentially higher risk of thermal runaway, providing a key quantitative indicator of group interference for thermal runaway prediction.

[0107] S643: Based on the temperature consistency, obtain the deviation value between the proportion of unstratified data under extreme temperatures and historical data for the same period;

[0108] The aforementioned deviation value refers to the difference between the percentage of batteries that did not exhibit temperature stratification under the current extreme temperature and the percentage of batteries that did not exhibit temperature stratification during the same historical period (e.g., the same extreme temperature environment last year). The deviation value is calculated by obtaining the percentage of battery packs that did not exhibit temperature stratification under the same extreme temperature during the same historical period, and then subtracting this percentage from the current percentage. A negative deviation value indicates that the current temperature consistency is worse than the historical average; a positive value indicates better consistency. For example, if the percentage of batteries that did not exhibit temperature stratification during extreme temperature charging in the same historical period (one week of the same season last year) is set at 80% (6 out of 7 monitoring tests showed no stratification, and 1 showed stratification), and this percentage is 57.1% during the current prediction period, then the deviation value = current percentage - historical percentage = 57.1% - 80% = -22.9%. A large absolute negative deviation value indicates that the frequency of temperature stratification in the current battery pack under extreme temperature charging scenarios is significantly higher than in the historical period, and the risk of thermal runaway is increasing. By unifying the statistical standards of "time period + scenario", the deviation value can more accurately reflect the changing trend of battery pack performance. Compared with the "comparison of data from different dimensions" in existing technologies, it is more in line with the technical logic of "trend analysis" in thermal runaway prediction, and provides more reliable trend indicators for risk assessment.

[0109] S644: Input the Z-score value, probability value, slope of Z-score, proportion of effective data and deviation value into the preset risk assessment model, generate a comprehensive thermal runaway risk score through weighted calculation, and predict the probability of thermal runaway of the battery based on the score.

[0110] The aforementioned pre-defined risk assessment model can be a weighted model, taking the Z-score, probability value, Z-score slope, effective data percentage, and deviation value as inputs. The weights of each indicator are assigned based on historical thermal runaway data, and a weighted calculation is performed to obtain a comprehensive thermal runaway risk score. Then, the thermal runaway probability is predicted based on the correspondence between the score and the thermal runaway probability (such as a function fitted through historical data). By integrating multiple indicators reflecting thermal runaway risk from different perspectives and comprehensively considering the influence of various factors through weighted calculations, the generated comprehensive thermal runaway risk score can more comprehensively and accurately reflect the true risk, thereby improving the accuracy of thermal runaway probability prediction and providing a more reliable basis for battery safety management.

[0111] In this embodiment, the indicators of each dimension are refined into more statistically significant and dynamic indicators (Z-score value, probability value, slope of Z-score, etc.), and then a risk assessment model is used to comprehensively score and predict the probability of thermal runaway. Compared with existing methods that use single or limited types of indicators for prediction, this method covers multiple dimensions of indicators such as statistical deviation, probability distribution, dynamic changes, data validity, and historical comparison, making the features of the model input richer and more accurate. This significantly improves the accuracy and comprehensiveness of thermal runaway probability prediction, helps to detect battery thermal runaway risks earlier and more accurately, and ensures battery safety.

[0112] In one embodiment, the above risk assessment model is as follows:

[0113]

[0114] In the formula, Z is the Z-score value, p is the probability value, R is the proportion of valid data, k is the slope of the Z-score, and ΔP is the bias value. , , , , The weighting coefficients are and satisfy the following conditions: + + + + =1, calibrated using historical thermal runaway data.

[0115] The above formula is a weighted summation model, which multiplies the absolute value of the Z-score |Z|, the negative logarithm of the probability value -Inp, the absolute value of the slope of the Z-score |K|, the complement of the effective data percentage 1 - R, and the absolute value of the bias value |ΔP| by their respective weighting coefficients. , , , , The results are then summed to obtain the comprehensive thermal runaway risk score S. The weighting coefficients are calibrated using historical thermal runaway data. Specifically, a large amount of historical data on battery thermal runaway and non-thermal runaway events is used, employing methods such as linear regression and grid search, to determine the weight values ​​that maximize the model's prediction accuracy. Existing weighted models for battery thermal runaway prediction often select only a limited number of indicators, primarily single, conventional indicators such as temperature and voltage. They lack comprehensive consideration of statistical characteristics related to temperature stratification (such as Z-scores and probability values), dynamic characteristics (the slope of the Z-score), data validity characteristics (the proportion of valid data), and historical comparison characteristics (deviation values). This formula, however, comprehensively integrates these characteristic indicators for temperature stratification scenarios, making it more targeted and comprehensive.

[0116] This embodiment integrates multiple indicators closely related to temperature stratification and thermal runaway, enabling a comprehensive characterization of battery thermal runaway risk from multiple dimensions. For example, |Z| reflects the degree of standardized deviation from abnormal temperatures, -Inp reflects the rarity of abnormal temperatures, |K| reflects the rate of temperature change, 1 - R reflects the degree of disturbance of population stratification, and |ΔP| reflects the performance change trend under extreme temperatures. Combined with weights calibrated using historical data, the model can more accurately calculate the comprehensive thermal runaway risk score, thereby improving the accuracy of thermal runaway probability prediction and providing a more reliable quantitative basis for battery safety management.

[0117] In another embodiment, the risk assessment model described above is a deep learning neural network model, which assesses the risk of thermal runaway using the following dynamic weight recursive formula:

[0118]

[0119] in, is the normalized feature vector; Z is the Z-score value, p is the probability value, R is the proportion of valid data, k is the slope of the Z-score, and ΔP is the bias value;

[0120] These are dynamic weights that are updated during training iterations. For adaptive learning rate, satisfy L is the loss function. , The model predicts probabilities, and α and γ are hyperparameters that balance positive and negative samples;

[0121] b(t) is the bias term in the neural network model that is dynamically updated during the training iteration process;

[0122] t represents the number of iterations for model training;

[0123] This is the Sigmoid activation function.

[0124] First, the feature vector Normalization is performed to ensure all features fall within the same numerical range, preventing the model from overemphasizing certain features due to differences in units or numerical values. Then, a deep learning neural network model is used to process the normalized feature vectors with dynamic weights. The update formula is updated iteratively with training: ,in For adaptive learning rate, satisfy This means the learning rate decays exponentially with the number of training iterations *t*, allowing the model to learn quickly in the early stages of training and fine-tune in the later stages. The loss function used is focus loss. This paper addresses the imbalance between positive and negative samples (few thermal runaway samples and many normal samples) in battery thermal runaway prediction. By adjusting the focus on positive and negative samples using α and γ, the model prioritizes learning from thermal runaway samples. The bias term b(t) is dynamically updated during training iterations, further enhancing the model's ability to fit complex features. Finally, the Sigmoid activation function maps the model output to a value between 0 and 1, representing the thermal runaway probability S. Existing neural network models for battery thermal runaway prediction often employ fixed weights or simple weight update strategies, with fixed learning rates and common cross-entropy loss functions, failing to adequately address the imbalance between positive and negative samples. Furthermore, the input features are often conventional indicators such as temperature and voltage, lacking specific features tailored to temperature stratification. This paper's model employs dynamic recursive weight updates, an adaptive learning rate, a focus loss function, and inputs specific feature vectors tailored to temperature stratification, representing innovations in both model structure and input features.

[0125] In this embodiment, dynamic weights and adaptive learning rates enable the model to flexibly adjust parameters based on errors during training, more accurately fitting the complex relationship between temperature stratification and thermal runaway. The focus loss function effectively addresses the imbalance between positive and negative samples, improving the ability to identify thermal runaway samples. Targeted input of characteristic feature vectors for temperature stratification allows the model to better capture thermal runaway risk factors related to temperature stratification. Combining these advantages, this deep learning neural network model can more accurately and robustly predict the probability of battery thermal runaway, demonstrating significantly improved prediction performance compared to existing models and providing stronger protection for battery safety.

[0126] In one embodiment, the above-mentioned analysis of whether temperature stratification has occurred within the battery pack based on the temperature data includes:

[0127] S211: Obtain the temperature sub-data collected by each temperature acquisition unit in the same frame of the battery pack from the temperature data, and calculate the temperature difference sequence of the temperatures collected by adjacent temperature acquisition units.

[0128] The temperature sub-data collected by each temperature acquisition unit in the battery pack within the same frame of temperature data refers to temperature data acquired in the same batch. For example, if temperature data is uploaded every 10 seconds, then the uploaded data is considered to be from the same frame. The temperature difference sequence refers to a sequence formed by arranging the differences in temperature values ​​collected by adjacent temperature acquisition units in order. Temperature data collected by each temperature acquisition unit in the battery pack is acquired, and then the temperature difference between adjacent temperature acquisition units is calculated sequentially. These differences are then arranged into a sequence according to the order of the temperature acquisition units. For example, if an electric vehicle power battery pack is equipped with 100 temperature acquisition units, numbered sequentially from 1 to 100, and the collected temperature data are T1, T2, ..., T100, the adjacent temperature differences are calculated as d1 = T2 - T1, d2 = T3 - T2, ..., d99 = T100 - T99, forming the temperature difference sequence [d1, d2, ..., d99]. By calculating the temperature difference between adjacent temperature acquisition units, the absolute temperature change is transformed into a relative difference, which better highlights the non-uniformity of temperature distribution and provides basic data for subsequent wavelet transform analysis of temperature abrupt changes.

[0129] S212: Perform wavelet transform on the temperature difference sequence to obtain wavelet coefficients.

[0130] Wavelet transform is a time-frequency analysis method that can localize signal analysis in the time and frequency domains. Wavelet coefficients, obtained after wavelet transform, reflect the characteristics of the signal at different scales and locations. By selecting an appropriate wavelet basis function (such as the db4 wavelet), wavelet transform can be applied to the temperature difference sequence to decompose it into wavelet coefficients at different scales. For example, using the db4 wavelet basis function to perform wavelet transform on the above temperature difference sequence [d1,d2,...,d99] yields wavelet coefficient matrices at multiple scales. Wavelet transform can effectively detect abrupt changes in signals. By applying wavelet transform to the temperature difference sequence, abrupt changes in temperature distribution (corresponding to the boundaries of temperature stratification) can be extracted, allowing for more accurate capture of local temperature abrupt changes compared to traditional signal analysis methods.

[0131] S213: Determine the location and amplitude of the temperature abrupt change based on the distribution of the modulus maxima of the wavelet coefficients.

[0132] The modulus maxima of wavelet coefficients refer to the maximum points of the modulus (absolute value) of the wavelet coefficients, and these points correspond to abrupt changes in the signal. The amplitude of a temperature abrupt change refers to the magnitude of the temperature change at the abrupt change location. By analyzing the distribution of the modulus maxima of wavelet coefficients, the locations corresponding to the modulus maxima points (i.e., the locations of temperature acquisition units) are found, and the amplitude of the temperature abrupt change is calculated based on the correspondence between wavelet coefficients and temperature changes. For example, in the distribution of the modulus maxima of wavelet coefficients, a modulus maxima point is found to correspond to the vicinity of the 20th temperature acquisition unit, and the calculated temperature abrupt change amplitude at this location is 4℃. Accurately locating the location of temperature abrupt changes and quantifying the abrupt change amplitude provides key abrupt change characteristic information for determining the existence of temperature stratification, making the determination of temperature stratification more targeted and accurate.

[0133] S214: When there is at least one temperature abrupt change location and the corresponding temperature abrupt change amplitude is greater than a preset threshold, and in conjunction with the spatial structure of the battery pack, it is determined that temperature stratification has occurred within the battery pack.

[0134] A preset temperature abrupt change threshold (e.g., 3℃) is used. When a temperature abrupt change occurs at a location exceeding this threshold, the battery pack's spatial structure (e.g., the arrangement of individual cells, heat dissipation channel distribution, etc.) is considered to determine whether these abrupt change locations form areas of abnormal temperature distribution. If so, temperature stratification is determined to exist. For example, if there is a temperature abrupt change location with a change of 4℃ (greater than the preset threshold of 3℃), and considering the battery pack's spatial structure, the individual cells near this abrupt change location are densely packed and have poor heat dissipation channels, then temperature stratification is determined to have occurred within the battery pack. By combining temperature abrupt change characteristics with battery pack spatial structure factors, the determination of temperature stratification more closely reflects the actual battery pack structure and thermal distribution, avoiding misjudgments due to simple temperature abrupt changes and improving the accuracy and reliability of temperature stratification determination.

[0135] In this embodiment, wavelet transform is used to determine whether temperature stratification exists within the battery pack. Compared to existing methods that rely on simple temperature statistics (such as average temperature and temperature variance) to identify abnormal temperature distribution, wavelet transform can more accurately capture local temperature abrupt changes. Combined with the spatial structure of the battery pack, this further improves the accuracy of temperature stratification detection. Accurate temperature stratification detection is a crucial prerequisite for subsequent thermal runaway prediction. Therefore, this embodiment provides a more reliable foundation for the entire thermal runaway prediction process, helping to improve the overall accuracy of thermal runaway prediction and ensuring battery safety.

[0136] In another embodiment, the above-mentioned determination of whether temperature stratification exists within the battery pack based on the temperature data includes:

[0137] S221: Obtain temperature sub-data of each location in the same frame of the battery pack from the temperature data, and calculate the temperature difference between adjacent temperature acquisition points.

[0138] Temperature acquisition points refer to specific locations within the battery pack used to collect temperature data, including the surface of individual battery cells and local areas of the battery pack. The temperature difference between adjacent acquisition points refers to the difference in temperature values ​​collected by two spatially adjacent acquisition points. Through an array of temperature acquisition units deployed within the battery pack, temperature data from each acquisition point is collected in real time (e.g., every 5 seconds). Then, the temperature difference between adjacent acquisition points is calculated sequentially according to their spatial location, forming a temperature difference sequence. For example, a power battery pack for an electric vehicle might have 30 individual battery cells arranged in a "3×10" matrix, with one temperature acquisition point at the center of each cell, for a total of 30 acquisition points. Temperatures at time t1 were collected as follows: T1=32℃, T2=33℃, T3=31℃…T30=32.5℃. Adjacent temperature differences were calculated as follows: d1=T2-T1=1℃, d2=T3-T2=-2℃, d3=T4-T3=0.8℃…d29=T30-T29=0.5℃, resulting in a sequence of 29 temperature differences. Converting absolute temperature data into adjacent temperature difference data amplifies subtle differences in local temperature distribution, providing more accurate baseline data for subsequent standard deviation calculations that reflect temperature non-uniformity and prevents local anomalies from being masked by overall temperature shifts.

[0139] S222: Calculate the standard deviation of the temperature difference based on the temperature difference between the adjacent temperature acquisition points.

[0140] The standard deviation of temperature difference is a statistical measure of the dispersion of temperature difference data between adjacent temperature sampling points. Based on the temperature difference sequence obtained in the above steps, the average value of all temperature differences is first calculated, and then substituted into the standard deviation formula to calculate the dispersion. The larger the value, the more drastic the temperature difference fluctuation between adjacent sampling points, and the more uneven the temperature distribution. Quantifying the dispersion of temperature difference through standard deviation transforms the "non-uniformity" of temperature distribution into a calculable and comparable numerical indicator. Compared with qualitative descriptions, this is more objective and operable, providing a quantitative basis for determining temperature stratification voids.

[0141] S223: Compare the standard deviation of the temperature difference with a preset threshold. If the standard deviation of the temperature difference is greater than the preset threshold, it is determined that temperature stratification has occurred in the battery pack.

[0142] The preset threshold is a judgment boundary set based on the standard deviation range of temperature difference during normal battery pack operation. It is determined through historical normal operating data statistics (such as the upper limit of the 95% confidence interval) or industry standards and is used to distinguish between normal temperature distribution and abnormal temperature stratification. The calculated standard deviation of temperature difference is compared with the preset threshold. If the standard deviation exceeds the threshold, it indicates that the temperature fluctuation between adjacent sampling points exceeds the normal range, and there are areas within the battery pack with significantly deviated local temperatures, thus indicating the presence of temperature stratification. Conversely, if the standard deviation is within the threshold, the temperature distribution is considered uniform. For example, for electric vehicle power battery packs, the standard deviation threshold for temperature difference is set at 0.8℃ based on historical normal data. If the currently calculated standard deviation is 1.2℃ > 0.8℃, then temperature stratification has occurred within the battery pack. This "standard deviation + preset threshold" judgment logic enables rapid and objective determination of temperature stratification gaps, avoiding the subjectivity and lag of traditional methods relying on manual experience or single temperature values.

[0143] In this embodiment, a three-step process of "temperature difference calculation → standard deviation quantification → threshold determination" is used to accurately identify temperature stratification voids. Standard deviation is used to convert temperature non-uniformity into numerical values, avoiding ambiguity in qualitative judgments. The standard deviation of adjacent temperature differences can capture subtle local temperature stratifications (such as a 1-2℃ local temperature difference caused by a micro-short circuit in a single unit), making it easier to detect early anomalies compared to the overall temperature threshold. Based on adjacent temperature difference analysis, interference from the overall environmental temperature rise on the judgment can be eliminated.

[0144] Reference Figure 3 This invention also provides a battery thermal runaway prediction device based on temperature stratification, comprising:

[0145] The acquisition unit 10 is used to acquire temperature data at various designated locations within the battery pack through multiple temperature acquisition units, wherein the temperature data is data within a battery thermal runaway prediction cycle;

[0146] The judgment unit 20 is used to analyze whether temperature stratification has occurred in the battery pack based on the temperature data. Temperature stratification refers to the temperature difference between a local area and the surrounding area in the battery pack reaching a preset condition.

[0147] The first acquisition unit 30 is used to determine whether there is a single outlier temperature acquisition unit if temperature stratification has occurred. If there is, the temperature acquisition unit is marked as an outlier and the first feature data is obtained. The single outlier refers to the temperature acquired by a single temperature acquisition unit deviating from the temperature acquired by other nearby temperature acquisition units and reaching a first deviation threshold.

[0148] The second acquisition unit 40 is used to determine whether there is a group-layered stratification region if temperature stratification has occurred. If it exists, the temperature stratification data of this part is filtered out to obtain the second feature data. The group-layered stratification refers to the temperature of multiple adjacent temperature acquisition units that are similar to form a temperature region, and the temperature of this temperature region deviates from the temperature of other regions and reaches the second deviation threshold.

[0149] The third acquisition unit 50 is used to count the percentage of the battery pack that does not exhibit temperature stratification under extreme temperatures, and obtain the third feature data, wherein the extreme temperature refers to the temperature that exceeds the normal operating temperature range of the battery.

[0150] The prediction unit 60 is used to predict the probability of thermal runaway of the battery based on the first feature data, the second feature data and the third feature data.

[0151] The above-mentioned battery thermal runaway prediction device based on temperature stratification is a device for implementing the above-mentioned battery thermal runaway prediction method based on temperature stratification. The implementation method is the same as the above embodiments, and will not be repeated here.

[0152] Reference Figure 4 The present invention also provides a computer device, the internal structure of which can be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor is designed to provide computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores operating devices, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores temperature data, etc. The network interface is used to communicate with external terminals via a network connection. Furthermore, the computer device may also include input devices and a display screen. When the computer program is executed by the processor, it implements the battery thermal runaway prediction method based on temperature stratification described in any of the above embodiments. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0153] One embodiment of this application also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the battery thermal runaway prediction method based on temperature stratification described in any of the above embodiments. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0154] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0155] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0156] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A battery thermal runaway prediction method based on temperature stratification, characterized in that, The method includes: Temperature data at designated locations within the battery pack are acquired through multiple temperature acquisition units, wherein the temperature data is data within a battery thermal runaway prediction cycle; Based on the temperature data, analyze whether temperature stratification has occurred within the battery pack. Temperature stratification refers to the temperature difference between a local area and the surrounding area within the battery pack reaching a preset condition. If temperature stratification has occurred, based on the data corresponding to the occurrence of temperature stratification, it is determined whether there is a single outlier temperature acquisition unit. If so, the temperature acquisition unit is marked as an anomaly, and first feature data is obtained. Here, the single outlier refers to a temperature acquired by a single temperature acquisition unit that deviates from the temperatures acquired by other nearby temperature acquisition units, reaching a first deviation threshold. Determine whether there is a group-stratified region. If so, filter out the temperature stratification data of that part to obtain the second feature data. The group-stratified region refers to a temperature region formed by multiple adjacent temperature acquisition units acquiring similar temperatures, and the temperature of this temperature region deviates from the temperature of other regions and reaches a second deviation threshold. The percentage of battery packs that did not exhibit temperature stratification under extreme temperatures was statistically analyzed to obtain the third characteristic data, where extreme temperatures refer to temperatures exceeding the normal operating temperature range of the battery. Based on the first feature data, the second feature data, and the third feature data, the probability of thermal runaway of the battery is predicted. The step of predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data includes: Based on the first feature data, calculate the temperature deviation index of the temperature collected by the abnormal temperature acquisition unit. Based on the second feature data, assess the degree of interference of population stratification on the overall temperature distribution; Based on the third feature data, the temperature consistency of the battery pack under extreme temperature conditions is analyzed. By combining the temperature deviation index, interference level, and temperature consistency, a thermal runaway risk assessment model is constructed to predict the probability of thermal runaway of the battery. The risk assessment model is a deep learning neural network model, which assesses the risk of thermal runaway using the following dynamic weight recursive formula: in, is the normalized feature vector; Z is the Z-score value, p is the probability value, R is the proportion of valid data, k is the slope of the Z-score, and ΔP is the bias value; These are dynamic weights that are updated during training iterations. For adaptive learning rate, satisfy L is the loss function. , The model predicts probabilities, and α and γ are hyperparameters that balance positive and negative samples. λ is the initial learning rate, and λ is the decay constant. b(t) is the bias term in the neural network model that is dynamically updated during the training iteration process; t represents the number of iterations for model training; This is the Sigmoid activation function.

2. The battery thermal runaway prediction method based on temperature stratification according to claim 1, characterized in that, Before predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data, the following steps are included: Based on the first feature data, the second feature data, and the third feature data, multiple temperature anomaly detection indicators are constructed, including: Year-on-year temperature difference at the end of the charging cycle: The change in temperature difference between each temperature acquisition unit / area of ​​the battery pack at the end of the current charging cycle and the same period of the previous month. High-temperature temperature rise rate: The rate at which the temperature rises, as measured by the temperature acquisition unit, during the charging process in an environment exceeding the first temperature. Low-temperature charging recovery rate: the rate at which the temperature collected by the temperature acquisition unit recovers to the normal operating temperature during the charging process in an environment below the second temperature; wherein, the first temperature is greater than the second temperature; If multiple temperature anomaly detection indicators exceed preset over-limit thresholds, then proceed to the step of predicting the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data.

3. The battery thermal runaway prediction method based on temperature stratification according to claim 1, characterized in that, The thermal runaway risk assessment model is constructed by comprehensively considering the temperature deviation index, interference level, and temperature consistency to predict the probability of thermal runaway of the battery, including: Based on the temperature deviation index, calculate the Z-score value, probability value, and slope of the temperature collected by each abnormal temperature acquisition unit. Based on the degree of interference, determine the proportion of remaining valid data after data filtering in the group-stratified stratified region; The Z-score value, probability value, slope of the Z-score, percentage of valid data, and deviation value are input into a preset risk assessment model. A comprehensive thermal runaway risk score is generated through weighted calculation, and the probability of thermal runaway of the battery is predicted based on the score.

4. The battery thermal runaway prediction method based on temperature stratification according to claim 1, characterized in that, The step of analyzing whether temperature stratification has occurred within the battery pack based on the temperature data includes: Obtain temperature sub-data collected by each temperature acquisition unit in the same frame of the battery pack from the temperature data, and calculate the temperature difference sequence of temperatures collected by adjacent temperature acquisition units. Perform wavelet transform on the temperature difference sequence to obtain wavelet coefficients; Based on the distribution of the modulus maxima of the wavelet coefficients, determine the location and amplitude of the temperature abrupt change; When there is at least one temperature abrupt change location, and the corresponding temperature abrupt change amplitude is greater than a preset threshold, and combined with the spatial structure of the battery pack, it is determined that temperature stratification has occurred within the battery pack.

5. The battery thermal runaway prediction method based on temperature stratification according to claim 1, characterized in that, The step of determining whether temperature stratification exists within the battery pack based on the temperature data includes: Acquire temperature sub-data for each location in the battery pack within the same frame of the temperature data, and calculate the temperature difference between adjacent temperature acquisition points; Calculate the standard deviation of the temperature difference based on the temperature difference between the adjacent temperature collection points; The standard deviation of the temperature difference is compared with a preset threshold. If the standard deviation of the temperature difference is greater than the preset threshold, it is determined that temperature stratification has occurred in the battery pack.

6. A battery thermal runaway prediction device based on temperature stratification, used to execute the battery thermal runaway prediction method based on temperature stratification as described in any one of claims 1-5, characterized in that, include: The acquisition unit is used to acquire temperature data at various designated locations within the battery pack through multiple temperature acquisition units, wherein the temperature data is data within a battery thermal runaway prediction cycle; The judgment unit is used to analyze whether temperature stratification has occurred in the battery pack based on the temperature data. Temperature stratification refers to the temperature difference between a local area and the surrounding area in the battery pack reaching a preset condition. The first acquisition unit is used to determine whether there is a single outlier temperature acquisition unit if temperature stratification has occurred. If there is, the temperature acquisition unit is marked as an outlier and the first feature data is obtained. The single outlier refers to the temperature acquired by a single temperature acquisition unit deviating from the temperature acquired by other nearby temperature acquisition units and reaching a first deviation threshold. The second acquisition unit is used to determine whether there is a group-layered stratification region if temperature stratification has occurred. If it exists, the temperature stratification data of this part is filtered out to obtain the second feature data. The group-layered stratification refers to the temperature of multiple adjacent temperature acquisition units that are similar to form a temperature region, and the temperature of this temperature region deviates from the temperature of other regions and reaches the second deviation threshold. The third acquisition unit is used to count the percentage of the battery pack that does not exhibit temperature stratification under extreme temperatures, and obtain the third feature data, wherein the extreme temperature refers to the temperature that exceeds the normal operating temperature range of the battery. The prediction unit is used to predict the probability of thermal runaway of the battery based on the first feature data, the second feature data, and the third feature data.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the battery thermal runaway prediction method based on temperature stratification as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the battery thermal runaway prediction method based on temperature stratification as described in any one of claims 1 to 5.

Images