Early warning system, method, battery pack and electric vehicle for battery pack

Abstract

This application provides a battery pack early warning system, method, battery pack, and electric vehicle. It relates to the field of battery testing technology. The method includes: acquiring failure characteristics of the battery pack under test; inputting the failure characteristics into a preset failure prediction model to obtain a predicted risk level; determining the corresponding target early warning strategy from the mapping relationship between the preset risk level and the preset early warning strategy; and generating an early warning command. The above scheme trains a preset failure prediction model based on sample failure characteristics and corresponding risk levels. This allows for coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, avoiding misjudgments and missed judgments caused by random fluctuations in single failure characteristics. Based on the predicted risk level output by the model, the corresponding target early warning strategy is accurately matched, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

Description

Technical Field

[0001] This application relates to the field of battery detection technology, and in particular to a warning system, method, battery pack, and electric vehicle for a battery pack. Background Technology

[0002] In the application of power battery packs in new energy vehicles, early warning of coolant leakage is of great significance for the safety of battery pack application.

[0003] In related technologies, warnings are issued based on the real-time operating parameters of the battery pack using a fixed threshold method, but this method suffers from low accuracy. Summary of the Invention

[0004] This application provides a battery pack warning system, method, battery pack, and electric vehicle to improve the accuracy of battery pack warnings.

[0005] In a first aspect, embodiments of this application provide a warning system for a battery pack. The warning system for the battery pack is used to issue a warning to a battery pack under test in order to determine whether the battery pack under test is working properly.

[0006] The battery pack's warning system is configured to perform the following steps:

[0007] Obtain the failure characteristics of the battery pack under test;

[0008] Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model;

[0009] Based on the predicted risk level, the target early warning strategy corresponding to the predicted risk level is determined from the mapping relationship between the preset risk level and the preset early warning strategy.

[0010] A warning instruction is generated according to the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high voltage circuit, and shutdown protection.

[0011] The preset failure prediction model is determined after training based on the failure characteristics of all samples and the risk level corresponding to each failure characteristic.

[0012] Secondly, embodiments of this application provide a warning method for a battery pack, comprising:

[0013] Obtain the failure characteristics of the battery pack under test;

[0014] Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model;

[0015] Based on the predicted risk level, the target early warning strategy corresponding to the predicted risk level is determined from the mapping relationship between the preset risk level and the preset early warning strategy.

[0016] A warning instruction is generated according to the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high voltage circuit, and shutdown protection.

[0017] The preset failure prediction model is determined after training based on the failure characteristics of all samples and the risk level corresponding to each failure characteristic.

[0018] Thirdly, embodiments of this application provide a battery pack, which is obtained through a warning system for the battery pack described in the first aspect.

[0019] Fourthly, embodiments of this application provide an electric vehicle that includes at least the battery pack described in the third aspect.

[0020] Fifthly, embodiments of this application provide an electrical device that includes at least the battery pack described in the third aspect.

[0021] Sixthly, embodiments of this application provide a warning device for a battery pack, comprising:

[0022] The acquisition module is used to acquire the failure characteristics of the battery pack under test.

[0023] The input module is used to input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model.

[0024] The determination module is used to determine the target early warning strategy corresponding to the predicted risk level from the mapping relationship between preset risk levels and preset early warning strategies, based on the predicted risk level.

[0025] The generation module is used to generate a warning instruction according to the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high-voltage circuit, and shutdown protection.

[0026] In a seventh aspect, embodiments of this application provide an electronic device, including: a memory and a processor;

[0027] The memory stores computer-executed instructions;

[0028] The processor executes computer execution instructions stored in the memory, causing the processor to perform the implementation method described in the second aspect above.

[0029] Eighthly, embodiments of this application provide a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the embodiments described in the second aspect above.

[0030] Ninthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the implementation methods described in the second aspect above.

[0031] The battery pack early warning system, method, battery pack, and electric vehicle provided in this application embodiment train a preset failure prediction model based on sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. Based on the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning. Attached Figure Description

[0032] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0033] Figure 1 A schematic diagram illustrating an application scenario of a battery pack early warning method provided in this application embodiment;

[0034] Figure 2 A flowchart illustrating a battery pack early warning method provided in an embodiment of this application;

[0035] Figure 3 A flowchart illustrating another early warning method for a battery pack provided in an embodiment of this application;

[0036] Figure 4 A schematic diagram of the coupled vibration and shock conditions provided in the embodiments of this application;

[0037] Figure 5 A schematic diagram illustrating the acquisition of failure characteristics provided in an embodiment of this application;

[0038] Figure 6 A schematic diagram of a battery pack provided in an embodiment of this application;

[0039] Figure 7 A schematic diagram of an electric vehicle provided in an embodiment of this application;

[0040] Figure 8 This is a schematic diagram of the structure of a battery detection device provided in an embodiment of this application;

[0041] Figure 9 This is a schematic diagram of another battery detection device provided in an embodiment of this application;

[0042] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0043] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0044] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0045] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0046] It should be noted that the phrase "at...time" in the embodiments of this application can refer to the instant at which a certain situation occurs, or to a period of time after the occurrence of a certain situation; the embodiments of this application do not specifically limit this. Furthermore, the display interface provided in the embodiments of this application is merely an example, and the display interface may include more or less content.

[0047] It should be noted that the warning system, method, battery pack, and electric vehicle of this application can be used in the field of battery testing technology, or in any field other than battery testing. The application fields of the warning system, method, battery pack, and electric vehicle of this application are not limited.

[0048] Figure 1 This is a schematic diagram illustrating an application scenario of a battery pack early warning method provided in this application embodiment. The scenario illustrated is as follows: the cooling pipes of the battery pack are inspected, and corresponding early warning commands are executed based on the inspection results to ensure battery pack safety.

[0049] In practical applications, power battery packs generally adopt liquid cooling heat dissipation structure, which removes the heat generated by the battery pack through coolant circulation, effectively improving the overall temperature control uniformity of the battery pack, and has better heat dissipation efficiency and thermal management capabilities compared with air cooling structure.

[0050] However, under long-term vehicle vibration, impact, and alternating high and low temperature conditions, liquid-cooled battery packs are prone to damage such as micro-cracks and loosening of cooling pipes and sealing joints, which can lead to coolant leakage. After coolant enters the battery pack, it gradually wets electrical components such as high-voltage copper busbars, terminals, and cell surfaces, causing deterioration problems such as insulation attenuation, micro-short circuits, and localized overheating. Ultimately, it can easily induce thermal runaway, which is one of the safety hazards of liquid-cooled battery packs.

[0051] In related technologies, warnings are issued based on a single real-time operating parameter of the battery pack by using a fixed threshold.

[0052] However, parameters for a single operating condition can fluctuate randomly, and the early warning schemes of related technologies may have errors, leading to misjudgments or missed judgments.

[0053] The method for early warning of battery packs provided in this application aims to solve the above-mentioned technical problems in related technologies.

[0054] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0055] Figure 2 This is a flowchart illustrating a battery pack early warning method provided in an embodiment of this application. The method includes the following steps:

[0056] S201. Obtain the failure characteristics of the battery pack under test.

[0057] The battery pack warning system is used to provide warnings to the battery pack under test in order to determine whether the battery pack is working properly.

[0058] For example, failure characteristics are feature data characterizing the potential fault state of a battery pack under test, specifically including feature parameters related to coolant leakage. The process of acquiring failure characteristics involves extracting feature quantities that reflect signs of leakage from the operational data of the battery pack under test. During operation, the Battery Management System (BMS) continuously collects raw data such as voltage, current, temperature, and insulation resistance. The early warning system reads this raw data through a communication interface. Leakage characteristics may be hidden in signals such as a slow decreasing trend in insulation resistance, intermittent fluctuations in liquid cooling circuit pressure, and abnormal changes in the cell temperature rise rate.

[0059] For example, the normal operating state of the battery pack under test serves as the benchmark for determining whether a warning is needed. When there is no coolant leakage, the various operating parameters of the battery pack remain within the normal range, and the failure characteristics are not obvious. When coolant leakage begins, characteristics such as decreased insulation resistance, decreased liquid cooling circuit pressure, and increased cell temperature difference gradually appear, and these characteristics constitute the specific content of the failure characteristics.

[0060] For example, when acquiring failure characteristics, real-time data from the Battery Management System (BMS) of the battery pack under test can be analyzed, or historical operating data of the battery pack under test can be combined for joint analysis. When the insulation resistance value exhibits a step-like decay, the failure characteristics may include the rate of decrease in insulation resistance, the current insulation resistance value, and the duration of the decay phenomenon; when the liquid cooling circuit pressure exhibits abnormal fluctuations, the failure characteristics may include the pressure fluctuation amplitude, fluctuation frequency, and changes in coolant flow rate.

[0061] S202. Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model.

[0062] For example, the preset failure prediction model is a classification or regression model pre-trained using sample data. It is used to calculate the risk level of thermal runaway in the battery pack under test based on the input failure features. After the failure features are input into the model, the model processes the feature data and outputs the predicted risk level. The predicted risk level characterizes the current risk level of the battery pack under test; a higher risk level indicates a greater likelihood of thermal runaway or that the battery pack is closer to thermal runaway.

[0063] Optionally, the preset failure prediction model can be a model built based on machine learning algorithms, such as gradient boosting trees, random forests, and support vector machines (SVMs), or a model built based on deep learning algorithms, such as convolutional neural networks (CNNs) and long short-term memory networks (LSTMs).

[0064] For example, the input to the preset failure prediction model is the failure characteristics obtained in S201, and the output is a discrete risk level label or a continuous risk score. When the output is a discrete label, the predicted risk level can be low risk, medium risk, high risk, etc. When the output is a continuous score, the score can be further mapped to the corresponding risk level through threshold division.

[0065] For example, the model's inference process involves operations such as feature vectorization, weight calculation, and nonlinear mapping. After training, the model learns the mapping relationship between failure features and risk levels. Therefore, when used online, you only need to input the real-time acquired failure features into the model according to the training format, and the model can output the corresponding predicted risk level.

[0066] Optionally, before inputting failure features into the preset failure prediction model, they can undergo preprocessing such as standardization and normalization to unify the dimensional range of different features and improve the stability of model prediction. For example, the insulation resistance feature has a large value range, while the temperature rise rate feature has a small value range. After normalization, the contribution weight of each feature in the model is more balanced.

[0067] For example, the failure characteristics of the battery pack under test may involve feature data in multiple dimensions, such as structural deformation features, insulation degradation features, and temperature anomaly features. These features need to be spliced ​​and format converted before being input into the model to meet the input requirements of the model.

[0068] S203. Based on the predicted risk level, determine the target early warning strategy corresponding to the predicted risk level from the mapping relationship between the preset risk level and the preset early warning strategy.

[0069] For example, the mapping relationship between preset risk levels and preset early warning strategies is a pre-configured correspondence table or correspondence rule, used to map different predicted risk levels to early warning strategies of different intensities. The risk level classification in the mapping relationship is consistent with the predicted risk level output in S202.

[0070] For example, the target early warning strategy is an early warning strategy corresponding to the current predicted risk level. Different risk levels correspond to different early warning strategies, realizing tiered early warning. The mapping relationship can be stored in the storage module of the early warning system in the form of a table, or it can be implemented in the form of a rule engine.

[0071] For example, the mapping relationship can be constructed by: statistically analyzing historical fault data to determine which early warning strategies can effectively prevent thermal runaway under different risk levels, or by pre-defining the correspondence between risk levels and early warning strategies according to the battery pack's safety design specifications. Low risk levels correspond to the "fault indication and continuous monitoring" strategy, medium risk levels correspond to the "limiting charging and discharging power" and "enhancing heat dissipation" strategies, and high risk levels correspond to the "locking high-voltage circuit" and "shutdown protection" strategies.

[0072] Optionally, the mapping relationship is stored in the early warning system's storage module. After obtaining the predicted risk level, the early warning system determines the target early warning strategy by looking up the mapping relationship. The mapping relationship can be updated based on actual operational data to continuously optimize the adaptability of the early warning strategy.

[0073] Optionally, the warning strategies in the mapping relationship can be used in combination, meaning that one risk level can correspond to a combination of multiple warning strategies. A medium-risk level can simultaneously correspond to both "limiting charging and discharging power" and "enhancing heat dissipation" strategies to reduce battery pack heat generation and accelerate heat dissipation, thus slowing down the rate of fault evolution. When determining the target warning strategy, the warning system uses all strategies corresponding to that risk level in the mapping relationship as the target warning strategy. When the predicted risk level is low, the target warning strategy can only include "fault indication and continuous monitoring," prompting the user to pay attention to the battery pack status without affecting normal use; when the predicted risk level is high, the target warning strategy can simultaneously include "locking the high-voltage circuit" and "shutdown protection," immediately cutting off the high-voltage circuit and controlling the vehicle to safely shut down, preventing thermal runaway.

[0074] S204. Generate a warning instruction based on the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high-voltage circuit, and shutdown protection.

[0075] The preset failure prediction model is determined after training based on the failure characteristics of all samples and the risk level corresponding to each failure characteristic.

[0076] For example, a warning command is a control instruction issued by the warning system to the battery management system. This instruction contains the specific content and execution parameters of the target warning strategy. After receiving the warning command, the battery management system parses the instruction content and executes the corresponding warning operation.

[0077] Optionally, warning commands can be transmitted to the battery management system via Controller Area Network (CAN) bus, Ethernet, or wireless communication.

[0078] Optionally, when generating the warning instruction, the instruction field is filled in according to the type of the target warning strategy. When the target warning strategy includes "locking the high-voltage circuit", the warning instruction includes a high-voltage circuit lock indicator bit and a lock delay parameter; when the target warning strategy includes "limiting charging and discharging power", the warning instruction includes a power limit ratio and a limit duration.

[0079] Optionally, after executing the warning command, the battery management system shall send the execution result back to the warning system.

[0080] For example, the failure characteristics of the samples are derived from the characteristic data recorded during the failure evolution of historical faulty battery packs, including but not limited to the time series changes of parameters such as insulation resistance, liquid cooling circuit pressure, cell temperature, and gas concentration.

[0081] The battery pack early warning method provided in this application embodiment trains a preset failure prediction model based on sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and missed judgment problems caused by random fluctuations of single failure characteristics in related technologies, improving the accuracy of battery pack failure risk identification. Based on the mapping relationship between preset risk levels and early warning strategies, the model outputs the predicted risk level and accurately matches the corresponding target early warning strategy. This determines appropriate protection strategies for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0082] Based on any of the above embodiments, the following, in conjunction with Figure 3 The detailed process of the early warning method for the battery pack is explained.

[0083] Figure 3 This is a flowchart illustrating another early warning method for a battery pack provided in an embodiment of this application. Figure 3 As shown, the method includes the following steps:

[0084] S301. Retrieve multi-dimensional real-time operating data of the battery pack under test from the battery management system of the battery pack under test.

[0085] For example, multi-dimensional real-time working data includes various parameters collected by the battery management system during the operation of the battery pack under test, covering multiple dimensions such as electrical, temperature, and liquid cooling circuits.

[0086] The electrical dimensions include parameters such as individual cell voltage, total voltage, charging and discharging current, and insulation resistance; the temperature dimensions include parameters such as cell temperature, module temperature, and coolant inlet and outlet temperatures; and the liquid cooling circuit dimensions include parameters such as coolant flow rate, liquid cooling circuit pressure, and coolant temperature.

[0087] Optionally, the retrieval process is implemented through a communication interface. The early warning system sends a data request to the battery management system, and the battery management system returns multi-dimensional real-time working data within the current time or recent time window to the early warning system.

[0088] For example, the acquisition frequency of multi-dimensional real-time working data can be set as needed. For instance, the normal acquisition frequency is no less than 10Hz. At critical points of thermal runaway, the acquisition frequency can be increased to 100Hz. High-frequency acquisition can capture transient parameter changes such as sudden changes in insulation resistance and instantaneous pressure fluctuations, providing sufficient data density for subsequent failure feature extraction.

[0089] Optionally, when retrieving multi-dimensional real-time working data, the operating status identifiers of the battery pack under test are also recorded simultaneously. These identifiers include charging, discharging, and resting states. The baseline parameter values ​​of the battery pack differ under different operating states. During charging, the voltage gradually increases; during discharging, the voltage gradually decreases; and during resting, the voltage tends to stabilize. Recording the operating status identifiers helps to accurately identify the validity of data during subsequent operating condition screening and avoids misjudgments caused by changes in operating states.

[0090] S302. Perform noise reduction and working condition filtering on multi-dimensional real-time working data to obtain filtered working data.

[0091] For example, noise reduction is used to filter out random disturbances introduced into multi-dimensional real-time working data due to factors such as electromagnetic interference, sensor noise, and communication jitter. Noise reduction can be implemented using filtering algorithms, such as low-pass filtering, Kalman filtering, and median filtering.

[0092] Taking median filtering as an example, sorting the data within the sliding window by size and taking the median value as the output value of the current point can effectively eliminate impulse noise while preserving the trend information in the data. The data after noise reduction is smoother, which can reduce misjudgments in the subsequent feature extraction process. Low-pass filtering is suitable for attenuating high-frequency noise components and preserving low-frequency signals that reflect slow trend changes, and is suitable for processing parameters such as insulation resistance.

[0093] For example, operating condition screening is used to remove data segments that are irrelevant to coolant leakage analysis or contain serious interference. The data characteristics of the battery pack under test differ under different operating conditions. Under rapid acceleration conditions, drastic current changes may cause abnormal fluctuations in voltage and temperature data. Operating condition screening identifies and removes these abnormal data segments based on preset operating condition judgment rules, retaining data under valid operating conditions. Operating condition judgment rules can be set based on conditions such as operating status indicators, parameter change rates, and data integrity. The filtered working data is high-quality data after noise reduction and operating condition screening.

[0094] For example, the order of noise reduction and operating condition screening can be adjusted.

[0095] Optionally, in practical applications, the configuration can be flexibly adjusted based on the amount of data and computing resources. The filtered working data is stored in the cache area of ​​the early warning system for use in subsequent feature extraction steps. The filtered working data also records the corresponding data acquisition time for time-series analysis during feature extraction.

[0096] S303. Based on the preset feature extraction rules, extract data from the filtered working data to obtain failure features related to coolant leakage.

[0097] For example, the preset feature extraction rules are predefined algorithms and judgment conditions used to identify signs of coolant leakage from working data.

[0098] For example, coolant leakage manifests in various forms in battery pack operating data. The feature extraction process sequentially examines various parameters in the filtered working data, identifies parameter change patterns that match leakage symptoms based on the judgment conditions in the feature extraction rules, and uses the identification results as failure features.

[0099] For example, the preset feature extraction rules include structural deformation feature extraction rules, leakage and infiltration time sequence feature extraction rules, and temperature anomaly feature extraction rules.

[0100] Structural deformation feature extraction rules are used to identify characteristics of cooling pipe damage from pressure and flow parameters in the liquid cooling circuit; leakage and wetting sequence feature extraction rules are used to identify characteristics of coolant wetting of electrical components from high-voltage insulation resistance and dynamic parameters of the liquid cooling circuit; temperature anomaly feature extraction rules are used to identify characteristics of thermal imbalance from parameters such as cell temperature rise rate and battery module temperature difference. Each feature extraction rule includes specific calculation methods and threshold conditions, such as thresholds for judging pressure fluctuation amplitude, insulation resistance decay rate, and cell temperature rise rate.

[0101] For example, in a coolant leakage scenario, the initial failure characteristics of the leak typically manifest as a slow decrease in insulation resistance and intermittent fluctuations in the liquid cooling circuit pressure, among other weak signals. By using pre-defined feature extraction rules, these subtle signal variations can be accurately extracted from multi-dimensional operational data. When the liquid cooling circuit pressure exhibits continuous fluctuations and deviates from the normal pressure range, the structural deformation feature extraction rules determine this as an abnormal deviation condition and extract the damage characteristics.

[0102] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0103] A feasible implementation method involves extracting damage features using the following steps: Based on structural deformation feature extraction rules, identify continuous fluctuations and abnormal deviations in liquid cooling circuit pressure and flow parameters from the screened working data to obtain damage features; based on leakage and infiltration sequence feature extraction rules, identify stepwise decay of insulation resistance and abnormal fluctuations in liquid cooling parameters from the screened working data to obtain coolant failure features; based on temperature anomaly feature extraction rules, identify battery pack thermal imbalance, localized overheating, and abnormal temperature rise conditions from the screened working data including cell temperature rise rate, battery module temperature difference, and charging / discharging electrical parameters to obtain thermal failure features; and finally, determine the damage features, coolant failure features, and thermal failure features as the final failure features.

[0104] For example, structural deformation feature extraction rules are used to identify deformation and damage characteristics of cooling pipes caused by mechanical damage. The pressure and flow parameters of the liquid cooling circuit directly reflect the structural integrity of the cooling pipes. When microcracks develop in the cooling pipes due to vibration fatigue, coolant leaks at the cracks, causing a drop in liquid cooling circuit pressure; the cracks open and close periodically under vibration, causing continuous pressure fluctuations.

[0105] During feature extraction, time-domain analysis is performed on the pressure parameters of the liquid cooling circuit to extract indicators such as pressure mean, pressure variance, and pressure fluctuation frequency. These indicators are then compared with preset fluctuation thresholds. When parameters fluctuate continuously and deviate from the normal pressure range, it is determined to be an abnormal deviation condition, and damage characteristics are obtained.

[0106] For example, damage characteristics include pressure fluctuation indicators, pressure deviation, and the corresponding duration. Microcracks caused by vibration fatigue at pipe weld locations have small crack openings, slow coolant leakage rates, and only minor pressure fluctuations in the liquid cooling circuit. Through cracks caused by impact at pipe bends have larger crack openings, faster coolant leakage rates, and correspondingly larger pressure fluctuations in the liquid cooling circuit. Structural deformation feature extraction rules can distinguish different degrees of damage by setting different fluctuation thresholds.

[0107] For example, if the pressure fluctuation is less than 5% of the normal pressure, it is considered a minor damage; if the pressure fluctuation is greater than 15% of the normal pressure, it is considered a serious damage.

[0108] For example, when extracting structural deformation features, the flow rate parameters of the liquid cooling circuit can also be introduced for joint judgment. When a cooling pipe is damaged, coolant leaks at the leak point, and the flow rate in the liquid cooling circuit decreases accordingly. By monitoring the flow rate change trend, the authenticity of the damage feature can be confirmed, avoiding misjudging normal pressure fluctuations caused by sensor malfunction or gas-liquid two-phase flow as damage.

[0109] For example, the leakage infiltration time-series feature extraction rule is used to identify insulation failure characteristics caused by coolant leakage into the internal electrical components of the battery pack. Coolant contains conductive ions; when coolant infiltrates the surface of high-voltage electrical components, the high-voltage insulation resistance decreases, and as the coolant continues to infiltrate, the insulation resistance exhibits a step-like decay. Dynamic parameters of the liquid cooling circuit include coolant flow rate and liquid cooling circuit pressure. When coolant leakage occurs, these dynamic parameters exhibit abnormal fluctuations. During feature extraction, trend analysis is performed on the high-voltage insulation resistance time series to identify whether the insulation resistance exhibits a step-like decay pattern; simultaneously, fluctuation analysis is performed on the dynamic parameters of the liquid cooling circuit to identify whether abnormal fluctuations exist.

[0110] For example, the method for determining the step-like decay of insulation resistance is as follows: divide the high-voltage insulation resistance time series into multiple consecutive time windows and calculate the average insulation resistance value within each window. When the average insulation resistance value of adjacent windows shows a gradual decrease and remains relatively stable within each window, it is determined to be a step-like decay. Criteria for determining the dynamic parameters of the liquid cooling circuit include the trend of decreasing coolant flow rate and the intermittent fluctuation of liquid cooling circuit pressure. Coolant failure characteristics include the insulation resistance decay rate, the step amplitude of the step-like decay, and the amplitude and frequency of abnormal fluctuations in liquid cooling circuit parameters.

[0111] For example, temperature anomaly feature extraction rules are used to identify characteristics of battery pack temperature anomalies caused by coolant leakage. After coolant leakage, the cooling effect of the coolant on the battery cells decreases, and the heat generated by the cells cannot be dissipated in time, leading to an abnormally high rate of temperature rise in the cells. At the same time, coolant wetting of electrical components may cause micro-short circuits, and the Joule heating generated by the micro-short circuits further exacerbates local overheating.

[0112] During feature extraction, the cell temperature rise rate is monitored and compared with a preset rate threshold; the battery module temperature difference is analyzed to identify whether there is a thermal imbalance state with excessive local temperature difference; and the charging and discharging electrical parameters are analyzed to identify whether there is additional heat generation caused by abnormal current.

[0113] For example, thermal failure characteristics include the cell temperature rise rate, battery module temperature difference, localized overheating temperatures, and corresponding changes in electrical parameters. The normal range for cell temperature rise rate is typically 0.2℃ / min to 0.5℃ / min; a rate exceeding 0.5℃ / min is considered abnormal. The normal range for battery module temperature difference is typically less than 3℃; a difference exceeding 5℃ is considered thermal imbalance. Localized overheating caused by micro-short circuits usually manifests as a significantly higher temperature in a particular cell or module compared to other areas, forming localized overheating spots. Charging and discharging electrical parameters include abnormal fluctuations in charging and discharging current and increased differences in individual cell voltages; these parameters can help identify the authenticity and severity of thermal failure characteristics.

[0114] For example, the identification of abnormal temperature rise conditions can also be corrected by incorporating ambient temperature. In high-temperature environments, the heat dissipation of the battery cell deteriorates, and the normal temperature rise rate is already high; therefore, the threshold for determining abnormal temperature rise needs to be adjusted based on the ambient temperature. In low-temperature environments, the heat dissipation of the battery cell is better, making abnormal temperature rise easier to identify. The temperature anomaly feature extraction rule improves the accuracy of thermal failure feature extraction by introducing an ambient temperature correction coefficient.

[0115] For example, damage characteristics, coolant failure characteristics, and thermal failure characteristics describe the signs of coolant leakage from three dimensions: structural deformation, insulation degradation, and thermal anomalies, respectively. By collectively identifying these three types of characteristics as failure characteristics, a comprehensive characterization of the leakage state of the battery pack under test can be achieved. These failure characteristics serve as input data for a pre-defined failure prediction model, which calculates the predicted risk level based on these characteristics.

[0116] For example, when performing feature fusion, features can be organized according to their type and time information. For instance, damage features, coolant failure features, and thermal failure features within the same time period can be treated as independent grouped features, and each grouped feature can be assigned a corresponding data type identifier. When the default failure prediction model reads the failure features, it parses the features based on the data type identifier.

[0117] For example, the significance of the three types of features varies at different stages of leakage. In the early stage of leakage, damage features appear first, coolant failure features begin to emerge, and thermal failure features are not yet obvious; in the middle stage of leakage, damage features persist, coolant failure features gradually become significant, and thermal failure features begin to appear; in the later stage of leakage, all three types of features are highly significant, indicating a high risk of thermal runaway. Using all three types of features as failure characteristics enables the pre-defined failure prediction model to comprehensively perceive the fault state of the battery pack.

[0118] In this feasible implementation, by extracting damage features, coolant failure features, and thermal failure features respectively, the fault signs of coolant leakage are comprehensively covered from three dimensions: mechanical structure, insulation performance, and thermal state. This effectively improves the completeness of the failure feature information, enabling the preset failure prediction model to comprehensively assess the risk status of the battery pack under test, thereby improving the accuracy of battery pack early warning.

[0119] One feasible approach is to determine the final failure characteristics by performing time-series alignment processing on the damage characteristics, coolant failure characteristics, and thermal failure characteristics, and removing isolated data with time-series mismatches to obtain the failure characteristics.

[0120] For example, damage features, coolant failure features, and thermal failure features may have temporal differences. Damage features appear earlier, coolant failure features appear slightly later, and thermal failure features appear latest. Temporal alignment processing is used to align the three types of features on a unified time axis, enabling joint analysis of the three types of features at the same time.

[0121] For isolated data that cannot be matched with other features in time, such as when only thermal failure features appear at a certain moment while there are no corresponding data for damage features and coolant failure features, the isolated data may be caused by sensor noise or accidental disturbances and should be removed.

[0122] For example, the specific method of time-series alignment includes: unifying the timestamps of damage features, coolant failure features, and thermal failure features to the same time base, and aligning the feature data in segments according to a preset time window length. Using a 1-minute time window, the three types of features within each window are concatenated to form the failure features within that time window. If certain feature categories are missing within a time window, that window can be marked. If multiple consecutive windows have missing features, these windows are marked as invalid and discarded.

[0123] For example, the timing alignment process also considers potential delays during feature extraction. Structural deformation features are derived from real-time analysis of liquid cooling circuit pressure parameters, resulting in low data latency; coolant failure features are derived from trend analysis of insulation resistance changes, resulting in relatively high data latency. Timing alignment can unify these features into a consistent timing relationship, avoiding feature misalignment caused by feature extraction delays and improving the overall quality of failure features.

[0124] In this feasible implementation, by performing time-series alignment processing on damage features, coolant failure features, and thermal failure features, and eliminating isolated data with time-series mismatches, it is possible to eliminate feature pseudo-associations caused by factors such as sensor sampling delay and data jitter, improve the data quality of failure features, enable the preset failure prediction model to make predictions based on time-series consistent failure features, further improve the accuracy of predicting risk levels, and thus improve the accuracy of battery pack early warning.

[0125] S304. Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model.

[0126] In one feasible implementation, before inputting failure features into a preset failure prediction model, the early warning method for the battery pack further includes: responding to an early warning request from a liquid-cooled battery pack, identifying multiple abnormal battery packs, wherein the multiple abnormal battery packs are obtained by performing different failure simulations on multiple normal battery packs respectively; acquiring multiple sample failure features of the multiple abnormal battery packs and acquiring multiple sample risk levels of the multiple abnormal battery packs, wherein the multiple sample failure features include: damage features, coolant failure features, and thermal failure features; and training the model based on the multiple sample failure features and the multiple sample risk levels to obtain a preset failure prediction model.

[0127] For example, during model training, different failure simulations are performed on multiple normal battery packs to obtain multiple abnormal battery packs. Failure simulation is used to artificially construct the failure state of the battery packs, making the normal battery packs exhibit failure characteristics consistent with real failures. Optionally, failure simulation methods include mechanical damage simulation, coolant injection simulation, and thermal load simulation. Vibration loads are applied to the cooling pipes of normal battery packs to generate fatigue cracks, or a set flow rate of coolant is injected into the battery pack to simulate leakage failure, or the battery pack is placed in a high-temperature environment to simulate thermal failure. The abnormal battery packs obtained after failure simulation have failure characteristics consistent with real failures.

[0128] Optionally, failure simulation can be performed using a simulation system to obtain multiple abnormal battery packs.

[0129] The simulation system includes a vibration table, an impact table, a leak detection device, a liquid injection device, and an environmental chamber.

[0130] For example, a vibration table is used to simulate alternating vibration loads caused by road surface unevenness during vehicle operation, while an impact table is used to simulate instantaneous impact loads such as chassis gravel impacts and bottoming out. Leakage detection equipment is used to reproduce the damaged state of battery pack piping and to calibrate the leakage characteristics corresponding to different damage modes, simulating leakage conditions after micro-leakage and cracking of cooling pipes during actual vehicle operation. A coolant injection device is used to controllably deliver coolant to the cooling pipes, simulating the actual vehicle failure evolution process of continuous coolant leakage and gradual wetting of internal battery pack components, and conducting graded leakage and wetting tests. An environmental chamber is used to provide a temperature and humidity-controlled test environment to trigger and observe the thermal runaway evolution process.

[0131] For example, mechanical stress is applied to the cooling pipes of a normal battery pack by means of a vibration table and an impact table, according to the mechanical loading method of superimposed vibration load and impact load.

[0132] As an example, various coupling schemes for vibration and shock are illustrated:

[0133] Option 1: The test sequence is continuous vibration followed by minor impact. This simulates long-term urban road driving, where the vehicle is frequently hit by road debris and subjected to low-speed, gentle chassis scrapes. This represents the most common driving conditions for new energy vehicles.

[0134] The test parameters are as follows: triaxial random vibration: 12 hours for each axis, frequency band 5~20Hz, matching standard PSD parameters; 2 hours of static loading after loading (simulating the stress relaxation and coolant hysteresis infiltration process of a real vehicle); 150 joules of slight impact, and three vertical impacts at the front, middle and rear points of the bottom of the battery pack.

[0135] The damage evolution characteristics are that long-term alternating vibration causes fatigue micropores in the pipe welds and rubber seals, which, combined with a single slight impact, form shallow closed microcracks with extremely small crack openings and slow coolant penetration rates, closely matching the characteristics of long-term hidden leaks in real vehicles.

[0136] The operating conditions are consistent with conventional civilian road conditions, and the damage is minor and controllable with no structural penetration damage; however, the extent of defect expansion is limited and the fault evolution gradient is insufficient, making it unsuitable for advanced thermal runaway observation tests.

[0137] Option 2: The test sequence is continuous vibration followed by severe impact. This simulates a single, sudden impact from a deep pothole and a hard, protruding rock after a vehicle has been driving at high speeds on national highways for an extended period. This is a combination of conventional durability testing and a single, extreme bottoming-out incident, resulting in a severe road condition.

[0138] The test parameters are as follows: vibration parameters are the same as in Scheme 1, and the load is left to stand for 2 hours; 300 Joule heavy impact, and three vertical impacts at three points on the bottom side.

[0139] The damage evolution characteristics are: pre-accumulation of alternating fatigue stress by vibration, instantaneous amplification of stress concentration effect by severe impact, tearing of weak points in the pipeline, forming large-aperture through cracks, and uncontrollable leakage with no transition stage.

[0140] The feasibility is moderate; the impact is highly destructive, easily causing irreversible deformation of the battery pack casing, resulting in poor controllability of micro-leakage and low test repeatability.

[0141] Option 3: The test sequence is vibration, minor impact, and secondary vibration. This simulates the entire process of a vehicle driving on paved roads, experiencing minor chassis scrapes, and continuing to serve for a long time without maintenance, thus replicating the degradation logic of a passenger vehicle throughout its entire life cycle and representing the most frequent combined damage conditions on real roads.

[0142] The test parameters were segmented vibration: two segments, each lasting 6 hours, for a total duration of 12 hours, using standard PSD parameters; a slight impact of 150 joules was coupled in the middle, with three impacts at each of the three points; the system was then left to stand for 2 hours after loading.

[0143] The damage evolution characteristics are segmented vibration: two segments, each lasting 6 hours, for a total duration of 12 hours, using standard PSD parameters; a slight impact of 150 joules is coupled in the middle, with three impacts at each of the three points; and the system is left to stand for 2 hours after loading.

[0144] The damage is gradual, the leakage volume is stable, and the working conditions are highly consistent, which can accurately reproduce the hidden leakage fault in the actual vehicle and is suitable for subsequent thermal runaway evolution observation tests.

[0145] Option 4: The test sequence is vibration, severe impact, and secondary vibration. This simulates a vehicle experiencing severe bottoming out during long-distance driving, followed by failure to repair the vehicle promptly and continued driving on the road, which constitutes a high-risk and severe vehicle operating condition.

[0146] The test parameters were the same as those in Scheme 3, with 12-hour segmented vibration parameters; a 300-joule heavy impact was coupled in the middle, with three impacts at each of the three points; the system was left to stand for 2 hours after loading.

[0147] The damage evolution characteristics are that the severe impact directly caused structural damage to the pipeline, the secondary vibration accelerated the crack penetration and expansion, and a large amount of leakage occurred in a short period of time. The fault span was large and there was no gradual process.

[0148] The damage gradient span is large and the micro-leakage retention time is short, which is not conducive to observing the complete evolution process from leakage to insulation failure and thermal runaway.

[0149] Option 5: The test sequence is slight impact followed by continuous vibration. This simulates a minor chassis scrape during the initial use of a new vehicle, carrying initial minor defects and undergoing long-term vibration service, which is a condition where the defects evolve slowly after they have occurred.

[0150] The test parameters were: a pre-load of 150 joules of slight impact to pre-fabricate micro-defects, followed by 12 hours of standard triaxial vibration, and then 2 hours of static rest after loading.

[0151] The damage evolution characteristics are that the initial deformation defect slowly creeps and expands under constant vibration stress, the crack initiation cycle is long, the probability of micro-leakage detection is low, and the damage is highly random.

[0152] The test cycle is lengthy, the degradation rate is slow, the test efficiency is low, and it is not suitable for standardized batch testing.

[0153] Option 6: The test sequence is heavy impact followed by continuous vibration. This simulates a vehicle continuing to drive after suffering severe bottom damage, representing an extreme driving condition following an accident.

[0154] The test parameters were: a 300-joule heavy impact to cause structural damage, followed by 12 hours of standard triaxial vibration, and then 2 hours of static rest after loading.

[0155] The damage evolution characteristics are that the impact directly caused the casing to dent and the pipeline to break, and the subsequent vibration accelerated the rapid leakage of coolant, resulting in a large area of ​​liquid accumulation in the battery pack in a short period of time.

[0156] It is too destructive, lacks a micro-leakage transition phase, and cannot simulate a progressive coolant leakage failure.

[0157] Option 7: The test sequence is vibration synchronous superposition intermittent impact (conducted simultaneously). This simulates the vehicle driving on bumpy, rough, and unpaved roads, with the vehicle body continuously vibrating while the chassis is constantly subjected to random impacts from gravel and protrusions, representing the most complex and severe combined coupled working condition in real driving.

[0158] The test parameters were as follows: during a 12-hour standard triaxial vibration process, a slight impact of 150 joules was intermittently superimposed, with the test points rotating every 2 hours, 3 times per test, and then left to stand for 2 hours after loading.

[0159] The damage evolution characteristics are that vibration generates continuous alternating stress, intermittent impacts apply concentrated stress instantaneously to the fatigue weak points of the pipeline, stress coupling and superposition accelerate lattice slip, microcrack initiation rate is the fastest, crack distribution closely matches the random damage characteristics of actual vehicles, and is mostly concentrated in weld seams and pipeline bending stress concentration areas.

[0160] It has the highest accuracy in terms of working conditions, the fastest defect generation efficiency, and crack distribution that closely matches the actual vehicle; however, the testing equipment has a high threshold, requiring a dedicated vibration and impact composite test bench, and the impact point is highly random, making it difficult to control the consistency of the test and guarantee the repeatability of the test.

[0161] Below, combining Scheme 3 and Figure 4 The coupled operating conditions of vibration and shock are explained.

[0162] Figure 4 This is a schematic diagram illustrating the coupled vibration and shock conditions provided in an embodiment of this application. Figure 4As shown, a segmented triaxial random vibration test was conducted. First, a 6-hour random vibration test was performed. Then, a 150J slight impact was applied, with three vertical impacts each at three points on the bottom of the battery pack (front, middle, and rear). A second 6-hour random vibration test followed. After the mechanical load was applied, the battery was allowed to stand for 2 hours. A segmented load application logic was adopted, which closely matches the damage patterns throughout the entire lifecycle of a passenger vehicle. Controllable micro-leakage was generated through segmented stress accumulation, preventing sudden penetration damage. The micro-leakage duration was sufficient to facilitate complete observation of the entire process of coolant wetting, insulation degradation, high-voltage creepage, and thermal runaway propagation, balancing test reliability, repeatability, and cost.

[0163] By alternating continuous vibration and impact, the cooling pipes are subjected to alternating fatigue stress and instantaneous impact stress, resulting in rapid fatigue failure. The crack distribution closely matches the random failure characteristics of real vehicles, improving the realism of the mechanical loading test and thus enhancing the accuracy of failure simulation.

[0164] For example, a normal battery pack can be leak-detected and its damage characteristics can be calibrated using a leak detection device to obtain the leaking battery pack and its damage characteristics.

[0165] Specifically, under preset constant temperature and cooling circulation pressure conditions, coolant is injected into the cooling pipes of the battery pack under test. The coolant includes the fluorescent medium of the fluorescent tracer detection device. The coolant leakage rate of the cooling pipes is detected. When the leakage rate is greater than or equal to the preset rate, the leakage point and the damage characteristics of the leakage point are located by ultraviolet light. The damage characteristics include at least one of the following: leakage point size, three-dimensional coordinates, and morphology.

[0166] The leak detection equipment includes fluorescent tracer detection equipment and ultraviolet lamps.

[0167] For example, the preset ambient temperature condition can be 40℃±3℃, and the cooling circulation pressure condition is 0.8 bar to 0.9 bar. When injecting coolant into the cooling pipeline, the coolant injection flow rate is controlled at 20 mL / min±1 mL / min, and the circulation time is greater than 48 hours. A fluorescent medium is uniformly mixed in the coolant; this medium emits visible fluorescence under ultraviolet light to indicate the location of coolant leakage. A liquid coolant leakage rate greater than 2.67 mL / min is considered a valid leak.

[0168] For example, the coolant leakage rate can be detected in real time using a flow sensor or a level sensor. The flow sensor is installed at the return end of the cooling pipe to measure the difference between the return flow rate and the injection flow rate; this difference is the coolant leakage rate. When the leakage rate is less than a preset rate, it indicates that an effective leakage path has not yet formed in the cooling pipe, and the load needs to be increased or the test duration extended. When the leakage rate is greater than or equal to the preset rate, it is determined that an effective leakage path has formed in the cooling pipe.

[0169] For example, when locating a leak point using a UV lamp, the operation is carried out in a dark room. The UV lamp irradiates the outer surface of the battery pack cooling pipes, and the fluorescent medium accumulates and emits fluorescence at the leak point. The leak point is determined based on the location of the fluorescence. The dimensions of the leak point, including crack length, width, and depth, are measured using calipers or a 3D scanning device. The 3D coordinates of the leak point are recorded, and the morphological characteristics of the leak point, such as crack shape, edge smoothness, and opening status, are recorded using a microscopic imaging device.

[0170] By injecting coolant containing fluorescent medium into the cooling pipes and detecting the coolant leakage rate, the leak point is located and the damage characteristics are calibrated using an ultraviolet lamp when the leakage rate reaches a preset rate. This achieves precise location of the leak point and quantitative calibration of the damage characteristics, thereby improving the accuracy of failure simulation.

[0171] Optionally, if the leakage rate is less than the preset rate within a preset time period, the cooling pipe is subjected to continuous vibration repeatedly through the vibration table until the leakage rate is greater than or equal to the preset rate.

[0172] For example, the preset duration is a set leakage detection observation period, such as 2 to 6 hours. Within the preset duration, if the coolant leakage rate remains lower than the preset rate, it indicates that although microcracks exist in the cooling pipes, they have not yet formed a through-hole leakage channel, or the microcracks have self-closed during the leakage detection process. At this time, continuous vibration is applied to the cooling pipes again using a vibration table, with vibration parameters the same as in the mechanical loading stage, using alternating stress to cause the microcracks to reopen or further expand.

[0173] For example, the number of repeated vibrations and the duration of each vibration can be dynamically adjusted according to changes in the leakage rate. After each vibration loading is completed, the coolant leakage rate is re-detected. When the leakage rate is greater than or equal to a preset rate, the repeated vibration stops and the damage characteristic calibration stage begins. During repeated vibration, vibration acceleration and pipeline strain are monitored simultaneously to prevent vibration parameters from exceeding preset ranges and causing structural damage to non-preset areas.

[0174] For example, a coolant grading leakage and wetting test is performed on a leaking battery pack using a liquid injection device to obtain the coolant failure characteristics of the leaking battery pack.

[0175] Specifically, coolant is injected into the cooling pipes in stages according to multiple preset rates, and the pipes are left to stand and soak for a preset time after each injection.

[0176] The coolant failure characteristics corresponding to each preset rate are obtained respectively. The coolant failure characteristics include at least one of the following: coolant diffusion path, change in insulation resistance of the leaking battery pack, and immersion sequence of electrical components of the leaking battery pack.

[0177] For example, multiple preset rates include three levels: low, medium, and high. The injection rate for the low-speed level is 5 mL / min to 15 mL / min, for the medium-speed level it is 15 mL / min to 30 mL / min, and for the high-speed level it is 30 mL / min to 50 mL / min. During each injection stage, the injection device injects coolant into the cooling lines at a constant rate, maintaining the pressure within the cooling lines at 1.2 bar ± 0.2 bar. After each injection stage is completed, the injection is stopped, and the leaking battery pack is sealed and left to soak for 120 minutes. During the soaking process, the ambient temperature and relative humidity are maintained at 25℃ ± 2℃ and 50% ± 5%RH, simulating the slow diffusion process of coolant inside the battery pack under real vehicle parking conditions.

[0178] For example, after each stage of leakage is completed, the coolant failure characteristics at the preset rate are acquired. The coolant diffusion path is collected using high-speed imaging and fluorescence tracing technology, recording the diffusion trajectory of the coolant along the bottom protective plate of the battery pack, module gaps, wiring harnesses, etc. Insulation resistance changes are collected in real time by the insulation monitoring module, reflecting the impact of coolant wetting degree on the deterioration of the battery pack's electrical insulation performance. The wetting sequence of electrical components is determined by the combined monitoring of high-speed imaging and insulation resistance, recording the time sequence in which the coolant contacts electrical components such as low-voltage wiring harnesses, high-voltage connectors, and cell terminals from the leak point, in order to analyze the degree of impact of coolant wetting on electrical safety.

[0179] Optionally, during the graded leakage test, a miniature humidity sensor can be used to monitor the local humidity field distribution inside the battery pack. When coolant leaks into a certain area, the humidity in that area rises rapidly. The humidity field distribution can help infer the diffusion path and wetting boundary of the coolant.

[0180] By injecting coolant into the cooling pipes at multiple preset rates and allowing them to stand for a preset immersion time after each injection, the coolant failure characteristics corresponding to each preset rate are obtained. This enables the quantitative characterization of coolant diffusion path, insulation resistance change and electrical component immersion sequence under different leakage levels, covering a variety of complex scenarios and thus improving the accuracy of failure simulation.

[0181] For example, based on the usage scenario of the battery pack under test, a thermal runaway trigger test is conducted on the leaking battery pack using an environmental chamber to obtain the thermal failure characteristics of the leaking battery pack.

[0182] Specifically, determine the usage scenario corresponding to the leaking battery pack; determine the corresponding temperature and humidity conditions based on the usage scenario; control the temperature and humidity through an environmental chamber based on the temperature and humidity conditions, and collect the thermal failure characteristics of the leaking battery pack. The thermal failure characteristics include at least one of the following: the surface temperature change trend of the leaking battery pack, the cell temperature rise rate of the leaking battery pack, the thermal runaway trigger duration, the thermal deformation state of the casing of the leaking battery pack, and the internal thermal diffusion range of the leaking battery pack.

[0183] For example, the usage scenarios corresponding to the leaking battery pack are determined based on the intended application vehicle model and operating area. Usage scenarios include normal temperature scenarios, high temperature scenarios, and humid heat scenarios. The normal temperature scenario corresponds to a temperature and humidity condition of 25℃±2℃ and a relative humidity of 50%±5%RH, simulating the normal operation of the vehicle under mild climatic conditions. The high temperature scenario corresponds to a temperature and humidity condition of 55℃±2℃, simulating the high-temperature environment inside the battery pack after exposure to summer sun. The humid heat scenario corresponds to a temperature and humidity condition of 40℃ and a relative humidity of 85%RH, simulating the conditions where condensation is exacerbated in high humidity environments.

[0184] For example, after determining the temperature and humidity conditions, the temperature and humidity inside the test chamber are adjusted to the target conditions using an environmental chamber, while keeping the leaking battery pack in a sealed environment. During the test, the battery management system of the leaking battery pack remains in a low-voltage energized and high-voltage standby state, simulating the actual working state of the vehicle. The environmental chamber operates continuously according to the target temperature and humidity conditions, while simultaneously collecting the thermal failure characteristics of the leaking battery pack.

[0185] For example, surface temperature change trends are collected using an array of temperature sensors deployed on the surface of the battery pack. These sensors are arranged in a grid pattern to depict the temperature field distribution and temperature gradient changes on the battery pack surface. Cell temperature rise rates are collected using temperature sensors at the cell terminals, with a sampling frequency of at least 10Hz, upgraded to 100Hz at critical thermal runaway points. Thermal runaway trigger duration is recorded as the time interval from the start of the test to the moment thermal runaway is triggered. The thermal deformation state of the casing is captured using a combination of strain gauges and high-speed imaging, recording the expansion, bulging, and rupture processes of the casing during thermal runaway. The internal heat diffusion range is inferred from the temperature gradient distribution between cells, used to determine the propagation path and speed of thermal runaway within the cell module.

[0186] Based on different usage scenarios, corresponding temperature and humidity conditions are determined, and the test environment conditions are precisely controlled through an environmental chamber. This allows for the simulation of the thermal runaway triggering process of leaking battery packs under different climate regions and operating conditions. Multi-dimensional thermal failure characteristics are collected, such as surface temperature change trends, cell temperature rise rate, thermal runaway triggering time, shell thermal deformation state, and internal thermal diffusion range. This matches the actual usage scenarios of the battery pack, thereby improving the accuracy of failure simulation.

[0187] Below, in conjunction with Figure 5 The failure characteristics of the acquired samples are explained.

[0188] Figure 5 This is a schematic diagram illustrating the acquisition of sample failure characteristics provided in an embodiment of this application. Figure 5 As shown, a normal battery pack first undergoes a leakage detection and damage feature calibration process to acquire damage features such as the size, three-dimensional coordinates, and morphology of the leakage point, thus identifying the leaking battery pack. The leaking battery pack then undergoes a coolant graded leakage and wetting test to collect coolant failure features such as coolant diffusion path, insulation resistance changes, and the wetting sequence of electrical components. A thermal runaway triggering test is then performed to record thermal failure features such as the battery pack surface temperature change trend, cell temperature rise rate, thermal runaway trigger duration, casing thermal deformation state, and internal thermal diffusion range. These damage features, coolant failure features, and thermal failure features collectively constitute the overall sample failure features of the battery pack.

[0189] As an example, during the simulation system setup, each device underwent metrological calibration before testing to ensure controllable experimental errors and accurate data. The triaxial vibration table has a frequency range of 5Hz to 200Hz, matched with a standard power spectral density vibration spectrum, supports the import of real vehicle road spectra and triaxial synchronous vibration, and has a frequency and acceleration calibration error of less than or equal to ±5%, used to simulate alternating vibrations during vehicle operation and induce pipeline fatigue damage. The bottom impact table has an impact energy range of 50 Joules to 300 Joules, equipped with both hemispherical and wedge-shaped impact heads, with impact angles ranging from 0 degrees to 90 degrees, and an energy calibration error of less than or equal to ±3 Joules. Among these, 0 to 150 Joules covers low-speed gravel and minor bottoming-out conditions, while 150 to 300 Joules covers rough roads and severe bottoming-out extreme conditions. The coolant precision injection system has a flow rate range of 0 mL / min to 500 mL / min, an effective simulation range for routine tests of 5 mL / min to 50 mL / min, a pressure range of 0 bar to 5 bar, and is compatible with ethylene glycol aqueous solutions with a volume concentration of 40% to 60%. It features three injection modes: fixed-point, pulse, and continuous. The high and low temperature humidity chamber has a temperature range of -40°C to 70°C and a humidity range of 30% to 95% relative humidity, with temperature and humidity control accuracies of ±2°C and ±5% relative humidity, respectively. The multi-parameter monitoring system includes electrical acquisition, temperature detection, gas analysis, and high-speed camera modules, with an acquisition frequency of at least 10 Hz, which can be upgraded to 100 Hz for key thermal runaway points. It covers monitoring dimensions such as voltage, current, insulation resistance, gas concentration, and transient morphology. The data acquisition system synchronously summarizes all monitoring data with a data latency of less than or equal to 10 milliseconds, supporting curve plotting, data export, and test replay.

[0190] For example, multiple sample failure features and multiple sample risk levels are obtained for multiple abnormal battery packs. Sample failure features include damage features, coolant failure features, and thermal failure features, corresponding to three dimensions: structural deformation, insulation degradation, and temperature anomaly, respectively. The sample risk level is a label indicating the severity of the fault corresponding to each sample failure feature, which can be determined based on the actual fault evolution results of the abnormal battery packs. After failure simulation, if an abnormal battery pack experiences only minor coolant leakage without causing insulation failure, the corresponding sample risk level is low risk; if an abnormal battery pack experiences a large coolant leak leading to a sharp drop in insulation resistance, the corresponding sample risk level is high risk.

[0191] For example, after obtaining multiple sample failure features and multiple sample risk levels, the model is trained based on these sample data. During training, the sample failure features are used as input to the model, and the sample risk levels are used as supervision labels for the model. By iteratively optimizing the model parameters, the model learns the mapping relationship between failure features and risk levels.

[0192] In this feasible implementation, abnormal battery packs are obtained through failure simulation, and sample failure characteristics and risk levels are acquired. This solves the problem of difficulty in obtaining real failure samples and provides sufficient and labeled training data for training the preset failure prediction model. Model training based on multiple sample failure characteristics and risk levels enables the preset failure prediction model to accurately establish the mapping relationship between failure characteristics and risk levels, improving the accuracy of online prediction and thus enhancing the early warning accuracy of the battery pack.

[0193] One feasible approach is to train the model using the following method: extracting multi-dimensional parameters from the failure features of multiple samples to obtain multi-dimensional failure features, which include at least two of the following: mechanical features, electrical features, temperature features, gas features, and image features; and training the model based on the multi-dimensional failure features and the risk levels of multiple samples to obtain a predictive failure model.

[0194] For example, multi-dimensional parameter extraction refers to extracting feature parameters of different physical dimensions from the failure features of the original sample.

[0195] The mechanical characteristics include structural parameters such as cooling pipe crack size and number, shell deformation, and vibration acceleration, used to characterize the degree of mechanical damage. Electrical characteristics include electrical parameters such as insulation resistance, short-circuit current, and voltage deviation, used to characterize the degree of insulation degradation and electrical faults. Temperature characteristics include temperature parameters such as cell temperature rise rate, module temperature difference, and maximum temperature, used to characterize the degree of thermal anomalies. Gas characteristics include gas parameters such as hydrogen concentration, carbon monoxide concentration, and volatile organic compound (VOC) concentration, used to characterize the degree of electrolyte decomposition and thermal side reactions. Imaging characteristics include image features acquired by high-speed cameras, such as smoke, fire, and cell rupture, used to characterize the transient morphology of thermal runaway.

[0196] With scenario examples, multi-dimensional failure features can comprehensively describe the fault state of a battery pack from multiple physical dimensions. In the early stages of coolant leakage, the insulation resistance value in the electrical features begins to decrease, but the temperature feature is not yet obvious; as the leakage intensifies, the changes in electrical features become more significant, and the temperature feature gradually emerges; when thermal runaway is imminent, the concentrations of hydrogen and volatile organic compounds (VOCs) in the gas features begin to rise, and smoke appears in the imaging features. The features of each dimension complement each other, together forming a complete description of the failure features. By training the model based on multi-dimensional failure features and multiple sample risk levels, the model can make risk predictions based on more comprehensive feature information.

[0197] For example, during model training, multi-dimensional failure features are used as the input feature matrix. The model automatically learns the contribution weights of different features to the risk level during training. When sample data shows a high correlation between insulation resistance decay rate and high risk level, the model increases the weight of electrical features in risk prediction. Through multiple rounds of iterative training, the model gradually optimizes the feature weights, ultimately obtaining a predictive failure model capable of accurately predicting risk levels.

[0198] As an example, this illustrates the correlation patterns between multi-dimensional features.

[0199] Electrical-insulation linkage characteristics: In the early stage of minor coolant leakage, the total voltage of the battery pack does not change significantly, and the insulation resistance decreases linearly first; when the insulation resistance drops below 500Ω / V, creepage occurs in the high-voltage part and induces micro-short circuits, and the voltage difference between individual cells increases, which can be used as a basis for early leakage judgment.

[0200] Temperature-electric coupling characteristics: Micro-short circuit continuously generates Joule heat, and the cell temperature rise rate is maintained at 0.5~2℃ / min; at 70~75℃, the SEI film is damaged and decomposed; after the temperature exceeds 85℃, the electrolyte undergoes violent side reactions, the cell voltage drops rapidly, forming a fixed fault sequence of insulation attenuation, temperature rise, and voltage collapse.

[0201] Gas-temperature abrupt change characteristics: In the early stage of the thermal side reaction, hydrogen gas is released and its concentration first exceeds 500 ppm; when the temperature of the electrode inside the cell rises to 140~160℃, the electrolyte decomposes in large quantities, and CO and VOC gases increase explosively. The gas peak value lags behind the temperature peak value, which can be used as a distinguishing indicator of irreversible thermal runaway.

[0202] Mechanical-electrical correlation characteristics: The crack opening of the pipeline, the deformation of the shell, and the insulation attenuation rate are positively correlated. The higher the degree of mechanical damage, the faster the coolant wetting speed and the shorter the insulation breakdown time.

[0203] Image-multi-parameter synchronous features: The appearance features such as condensation, carbonization, and smoke captured by high-speed cameras occur simultaneously with insulation drop, hot spot protrusion, and hydrogen evolution, achieving two-way mutual verification of morphology and parameters.

[0204] In this feasible implementation, multi-dimensional parameters are extracted from the sample failure features to obtain multi-dimensional failure features such as mechanical, electrical, temperature, gas, and imaging. This fully utilizes different types of fault information, enabling the preset failure prediction model to more comprehensively perceive the fault state of the battery pack, effectively improving the accuracy of risk prediction, and thus improving the accuracy of battery pack early warning.

[0205] One feasible approach is to train the model using the following method: determining the application scenario for each abnormal battery pack; and training the model based on multi-dimensional failure characteristics, multiple sample risk levels, and the application scenario of each abnormal battery pack to obtain a failure prediction model.

[0206] For example, application scenarios are used to describe the vehicle usage environment simulated by abnormal battery packs, such as urban commuting scenarios, highway scenarios, and bumpy road scenarios. The operating conditions of the battery pack, such as vibration loads, impact loads, and ambient temperature, differ in different application scenarios.

[0207] For example, application scenarios have a significant impact on the failure evolution of battery packs. In high-temperature scenarios, the evaporation rate of coolant after leakage is faster, and the insulation resistance decays more quickly; in bumpy road scenarios, cooling pipe cracks propagate faster under continuous vibration loads, and the leakage volume increases significantly. By inputting application scenarios as part of the training data into the model, the model can learn the influence of different application scenarios on the mapping relationship between failure characteristics and risk levels.

[0208] Similarly, when issuing warnings for the battery pack under test, the application scenario of the battery pack under test is also collected as the input to the preset failure prediction model.

[0209] For example, when training a model based on multi-dimensional failure characteristics, multiple sample risk levels, and the application scenario of each abnormal battery pack, the application scenario can be encoded and converted into numerical features usable by the model.

[0210] In this feasible implementation, by introducing the application scenario of each abnormal battery pack into the model training, the preset failure prediction model can learn the influence of different usage scenarios on the battery pack failure process. When predicting the risk level, it combines the application scenario factors to make a comprehensive judgment, further improving the matching degree between the prediction results and the actual working conditions, thereby improving the early warning accuracy of the battery pack.

[0211] One feasible approach is to train the model using the following method: based on multi-dimensional failure characteristics, risk levels of multiple samples, and the application scenario of each abnormal battery pack, group the abnormal battery packs according to their operating conditions to obtain a training dataset. The training dataset includes data samples of multiple operating conditions, including normal operating conditions, minor leakage operating conditions, moderate wetting failure operating conditions, and thermal runaway precursor operating conditions. The model is then iteratively trained based on the training dataset, continuously optimizing the model feature weights and correlation parameters until the model meets the preset convergence conditions, thus obtaining a failure prediction model.

[0212] For example, sample data with the same or similar operating conditions are grouped together. Operating condition types include normal operating condition, minor leakage condition, moderate wetting failure condition, and pre-thermal runaway condition. Normal operating condition corresponds to a fault-free battery pack operation; minor leakage condition corresponds to a state where there is a small amount of coolant leakage but the electrical performance has not yet significantly deteriorated; moderate wetting failure condition corresponds to a state where coolant has seeped into electrical components and the insulation resistance has significantly decreased; and pre-thermal runaway condition corresponds to a state where the cell temperature rises abnormally and gas evolution indicates that thermal runaway is about to occur. The grouped training dataset contains data samples of multiple operating conditions, covering the complete fault evolution stage from the normal state to pre-thermal runaway.

[0213] For example, during iterative training, the model reads samples from the training dataset in batches, calculates predicted values ​​forward, propagates errors backward, and updates model parameters. After each iteration, the model's prediction accuracy on the validation set is calculated. If the prediction accuracy does not meet the preset convergence condition, the next iteration continues. Model feature weights and correlation parameters refer to the weight coefficients and bias parameters used in the model to calculate the prediction results, which are continuously optimized during iterative training.

[0214] Optionally, the preset convergence condition can be that the loss function value is less than a preset threshold, the prediction accuracy reaches a preset level, or the number of iterations reaches a preset maximum value.

[0215] In this feasible implementation, by grouping the training dataset according to the type of working condition, the preset failure prediction model can learn the differences in failure characteristics under different working conditions. Through iterative training, the model feature weights and correlation parameters are continuously optimized, enabling the model to accurately distinguish between normal working conditions, minor leakage working conditions, moderate wetting failure working conditions, and thermal runaway precursor working conditions, thereby improving the accuracy of risk prediction and thus improving the early warning accuracy of the battery pack.

[0216] S305. Based on the predicted risk level, determine the target early warning strategy corresponding to the predicted risk level from the mapping relationship between the preset risk level and the preset early warning strategy.

[0217] It should be noted that the execution process of S305 is the same as that of S203, and will not be repeated here.

[0218] S306. Generate a warning instruction based on the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high-voltage circuit, and shutdown protection.

[0219] One feasible implementation method, after generating the early warning instruction, may further include: acquiring the real-time operating condition of the battery pack under test and the historical operating condition and historical predicted risk level of the battery packs in the same batch under test; performing correlation analysis on the real-time operating condition, historical operating condition, historical predicted risk level, predicted risk level and target early warning strategy to obtain the failure evolution law; and updating the mapping relationship according to the failure evolution law to obtain the updated mapping relationship.

[0220] For example, real-time operating conditions refer to the current operating conditions of the battery pack under test, including charging, discharging, stationary, driving, and parking conditions. Historical operating conditions and historical predicted risk levels of battery packs from the same batch record changes in operating conditions and risk levels during historical operation. By obtaining this information, the correlation between the battery pack under test and other battery packs from the same batch can be analyzed in terms of failure behavior.

[0221] For example, correlation analysis refers to the joint analysis of real-time operating conditions, historical operating conditions, historical predicted risk levels, predicted risk levels, and target early warning strategies to uncover the correlation patterns between various data.

[0222] Specifically, we can analyze the similarity between the predicted risk level of the battery pack under test under current real-time operating conditions and the historical predicted risk level of the same batch of battery packs under similar operating conditions, and analyze the changing trend of the predicted risk level after the target early warning strategy is implemented, thereby obtaining the failure evolution law. The failure evolution law reflects the path and conditions for the battery pack to evolve from its current state to a more severe failure state, such as the time and triggering conditions required for it to evolve from a slight leakage condition to a moderate wetting failure condition.

[0223] For example, newly discovered failure evolution patterns are fed back into the mapping relationship between preset risk levels and preset early warning strategies, adjusting the correspondence between risk levels and early warning strategies. If correlation analysis reveals that the time for a minor leak to evolve into a pre-thermal runaway condition is significantly shorter under high-temperature conditions than under normal-temperature conditions, then in the mapping relationship, the early warning strategy corresponding to the minor leak risk level under high-temperature conditions is upgraded by one level, changing from "fault indication and continuous monitoring" to "limiting charging and discharging power" and "enhancing heat dissipation." After the update process is completed, an updated mapping relationship is obtained, which is used to determine the target early warning strategy when issuing warnings subsequently.

[0224] In this feasible implementation, by acquiring the real-time operating conditions of the battery pack under test and the historical operating conditions and historical predicted risk levels of the same batch of battery packs, and performing correlation analysis to obtain the failure evolution law, it is possible to discover the evolution trend of risk levels under different operating conditions, and update the mapping relationship accordingly, so that the determination of the early warning strategy is more in line with the actual use environment of the battery pack, continuously improving the adaptability of the early warning strategy, thereby improving the early warning accuracy of the battery pack.

[0225] This application provides a battery pack, which is a normal battery pack determined through an early warning process. The early warning process is performed through the early warning system of the battery pack.

[0226] See Figure 6 , Figure 6 A schematic diagram of the battery pack structure is shown in some embodiments of this application. The battery pack 200 is disposed inside the electric vehicle and can be installed at the bottom, front, or rear of the electric vehicle. The battery pack 200 can provide power support to the electric vehicle, for example, acting as the operating power source for the electric vehicle. The electric vehicle may also be equipped with a controller and a motor. The controller is used to schedule the battery pack to deliver electrical energy to the motor to meet the power needs of the electric vehicle during starting, navigation, and driving.

[0227] The battery pack 200, as a rechargeable battery, is the power source for new energy vehicles. The battery pack includes a housing 210 and multiple individual battery cells housed within the housing 210. The housing 210 provides space for the individual battery cells and other components, while also providing dustproof, waterproof, and protective functions to the internal components, thereby reducing the adverse effects of external liquids or foreign objects on the effectiveness and performance of the individual battery cells and other components, effectively extending the battery pack's lifespan. The shape of the housing is not limited; for example, it can be a cuboid, cylinder, etc. The material of the housing can also be flexibly selected; for example, copper, iron, aluminum, stainless steel, aluminum alloy, and plastic are all suitable.

[0228] The battery pack 200 in this embodiment includes multiple battery groups 230, with one battery group 230 disposed in each sub-accommodating space. Each battery group 230 includes multiple stacked individual cells, and the stacking direction of the individual cells can be parallel to the plane of the base plate. The battery pack 200 also includes a battery management system (BMS), a thermal management system, an electrical connection system (high-voltage / low-voltage connectors, wiring harnesses, etc.), structural components (casing, brackets, etc.), and protective components, etc., and the above components are placed in the housing 210 and sealed by the cover plate 220 to form a complete functional unit that can directly output electrical energy.

[0229] The battery management system (BMS) is used to detect the operating status of the battery pack and its individual cells, and to manage the battery module and its individual cells. The BMS includes a Battery Management Unit (BMU), a Cell Measurement Circuit (CMC), sensors, and various electronic control devices. The BMS includes at least one processor and a memory. The memory can be built into the BMS or externally located, and can also be remotely configured and connected to the BMS via a network.

[0230] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0231] This application provides an electric vehicle that includes at least the aforementioned battery pack.

[0232] The following example uses an electric vehicle as a specific example of an electrical device. See [link / reference] Figure 7 , Figure 7 The diagram illustrates the structure of an electric vehicle in some embodiments of this application. This electric vehicle can be a new energy vehicle, encompassing various types such as pure electric vehicles, hybrid electric vehicles, and range-extended electric vehicles. The battery device 500 can be located at the bottom, front, or rear of the electric vehicle 100, providing power support to the electric vehicle, for example, acting as its operating power source. Furthermore, the electric vehicle 100 typically also includes a controller 300, which manages the discharge process of the battery device 500 to cover various power needs during vehicle startup, navigation, and operation.

[0233] In some embodiments, the battery device 500 comprises one or more battery packs, which can be connected in series, parallel, or a hybrid configuration. A hybrid configuration refers to the simultaneous presence of series and parallel connections in the connection of multiple battery packs. In other embodiments, the battery device 500 is a cluster-level battery architecture composed of multiple battery packs connected in series, wherein the number of battery packs in each cluster is strictly configured according to voltage and capacity requirements. More specifically, the battery unit of the battery device 500 includes multiple batteries, some of which are connected in series to form a cluster that meets a preset power supply voltage requirement, and at least one spare battery among the multiple batteries is bypassed.

[0234] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0235] This application provides an electrical device that includes at least the battery pack described above.

[0236] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0237] Figure 8 This is a schematic diagram of the structure of a warning device for a battery pack provided in an embodiment of this application. Figure 8 As shown, the warning device 80 of the battery pack may include: an acquisition module 81, an input module 82, a determination module 83, and a generation module 84.

[0238] The acquisition module 81 is used to acquire the failure characteristics of the battery pack under test.

[0239] Input module 82 is used to input failure characteristics into a preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model.

[0240] The determination module 83 is used to determine the target early warning strategy corresponding to the predicted risk level from the mapping relationship between the preset risk level and the preset early warning strategy, based on the predicted risk level.

[0241] The generation module 84 is used to generate warning instructions according to the target warning strategy. The warning instructions are used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high voltage circuit, and shutdown protection.

[0242] Optionally, module 81 can be executed. Figure 2 S201 in the embodiment.

[0243] Optionally, input module 82 can execute Figure 2 S202 in the embodiment.

[0244] Optionally, module 83 can be executed. Figure 2 S203 in the embodiment.

[0245] Optionally, the generation module 84 can be executed. Figure 2 S204 in the embodiment.

[0246] In one possible implementation, the acquisition module 81 is specifically used for:

[0247] Retrieve multi-dimensional real-time operating data of the battery pack under test from its battery management system.

[0248] Noise reduction and working condition filtering are performed on multi-dimensional real-time work data to obtain filtered work data;

[0249] Based on the preset feature extraction rules, data is extracted from the filtered working data to obtain failure features related to coolant leakage.

[0250] In one possible implementation, the acquisition module 81 is specifically used for:

[0251] Based on the structural deformation feature extraction rules, the continuous fluctuation and abnormal deviation of parameters in the liquid cooling circuit pressure and flow parameters in the screened working data are identified to obtain the damage characteristics.

[0252] Based on the leakage and infiltration time sequence feature extraction rules, the step-like decay of insulation resistance and abnormal fluctuation of liquid cooling parameters are identified from the high-voltage insulation resistance and liquid cooling circuit dynamic parameters in the screened working data, and the coolant failure characteristics are obtained.

[0253] Based on the temperature anomaly feature extraction rules, the battery pack thermal imbalance, local overheating, and abnormal temperature rise conditions are identified from the cell temperature rise rate, battery module temperature difference, and charging and discharging electrical parameters in the screened working data, and thermal failure features are obtained.

[0254] Damage characteristics, coolant failure characteristics, and thermal failure characteristics are identified as failure characteristics.

[0255] In one possible implementation, the acquisition module 81 is specifically used for:

[0256] The failure features, coolant failure features, and thermal failure features are time-series aligned, and isolated data with time mismatch are removed to obtain the failure features.

[0257] It should be noted that the warning device for the battery pack shown in the embodiments of this application can execute the technical solution shown in the above method embodiments, and its implementation principle and beneficial effects are similar, so they will not be described again here.

[0258] Figure 9 This is a schematic diagram of the structure of a warning device for another battery pack provided in an embodiment of this application. Figure 8 Based on the illustrated embodiments, as Figure 9 As shown, the warning device 80 of the battery pack also includes: a training module 85 and an update module 86.

[0259] Training module 85 is used for:

[0260] In response to the warning request of the liquid-cooled battery pack, multiple abnormal battery packs were identified. These multiple abnormal battery packs were obtained by performing different failure simulations on multiple normal battery packs.

[0261] Multiple sample failure features of multiple abnormal battery packs and multiple sample risk levels of multiple abnormal battery packs are obtained. The multiple sample failure features include: damage features, coolant failure features and thermal failure features.

[0262] The model is trained based on the failure characteristics and risk levels of multiple samples to obtain a preset failure prediction model.

[0263] In one possible implementation, the training module 85 is specifically used for:

[0264] Multidimensional parameters are extracted from the failure features of multiple samples to obtain multidimensional failure features. The multidimensional failure features include at least two of the following: mechanical features, electrical features, temperature features, gas features, and image features.

[0265] A failure prediction model is obtained by training the model based on multi-dimensional failure characteristics and multiple sample risk levels.

[0266] In one possible implementation, the training module 85 is specifically used for:

[0267] Determine the application scenario for each abnormal battery pack;

[0268] A failure prediction model is obtained by training the model based on multi-dimensional failure characteristics, multiple sample risk levels, and the application scenario of each abnormal battery pack.

[0269] In one possible implementation, the training module 85 is specifically used for:

[0270] Based on multi-dimensional failure characteristics, risk levels of multiple samples, and application scenarios of each abnormal battery pack, the abnormal battery packs are grouped according to their operating conditions to obtain a training dataset. The training dataset includes data samples of multiple operating conditions, including normal operating conditions, minor leakage operating conditions, moderate wetting failure operating conditions, and thermal runaway precursor operating conditions.

[0271] The model is iteratively trained based on the training dataset, continuously optimizing the model feature weights and correlation parameters until the model meets the preset convergence conditions, thus obtaining the prediction failure prediction model.

[0272] Update module 86, used for:

[0273] Obtain the real-time operating status of the battery pack under test, as well as the historical operating status and historical predicted risk level of the battery packs in the same batch under test;

[0274] Correlation analysis was performed on real-time operating conditions, historical operating conditions, historical predicted risk levels, predicted risk levels, and target early warning strategies to obtain the failure evolution law;

[0275] Based on the failure evolution law, the mapping relationship is updated to obtain the updated mapping relationship.

[0276] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 10 As shown, the electronic device includes:

[0277] The electronic device includes a processor 291 and a memory 292; it may also include a communication interface 293 and a bus 294. The processor 291, memory 292, and communication interface 293 can communicate with each other via the bus 294. The communication interface 293 can be used for information transmission. The processor 291 can invoke logical instructions stored in the memory 292 to execute the methods of the above embodiments.

[0278] Furthermore, the logic instructions in the aforementioned memory 292 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0279] The memory 292, as a non-volatile computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this application. The processor 291 executes functional applications and data processing by running the software programs, instructions, and modules stored in the memory 292, that is, it implements the methods in the above-described method embodiments.

[0280] The memory 292 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 292 may include high-speed random access memory and may also include non-volatile memory.

[0281] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0282] This application provides a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method as described in the foregoing embodiments.

[0283] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0284] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in the foregoing embodiments.

[0285] Based on the above implementation methods, a preset failure prediction model is trained according to the sample failure characteristics and corresponding risk levels. This model can perform coupled and comprehensive analysis of multi-dimensional failure characteristics of the battery pack under test, such as damage, coolant failure, and thermal failure. This avoids the misjudgment and omission problems caused by random fluctuations of single failure characteristics in related technologies, and improves the accuracy of battery pack failure risk identification. According to the mapping relationship between preset risk levels and early warning strategies, the corresponding target early warning strategy can be accurately matched according to the predicted risk level output by the model. Appropriate protection strategies can be determined for different degrees of battery pack failure and degradation, avoiding over-protection or under-protection, thereby improving the accuracy of battery pack early warning.

[0286] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0287] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps; they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages, which do not necessarily complete at the same time but can be executed at different times. The execution order of these sub-steps or stages is also not necessarily sequential but can be alternated or carried out in turn with other steps or at least some of the sub-steps or stages of other steps.

[0288] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0289] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0290] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. The processor can be any suitable hardware processor, such as CPU, GPU, FPGA, DSP, and ASIC. The storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0291] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0292] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0293] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only.

[0294] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A warning system for a battery pack, characterized in that, The battery pack warning system is used to provide warnings to the battery pack under test in order to determine whether the battery pack under test is working properly; The battery pack's warning system is configured to perform the following steps: Obtain the failure characteristics of the battery pack under test; Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model; Based on the predicted risk level, the target early warning strategy corresponding to the predicted risk level is determined from the mapping relationship between the preset risk level and the preset early warning strategy. A warning instruction is generated according to the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high voltage circuit, and shutdown protection. The preset failure prediction model is determined after training based on the failure characteristics of all samples and the risk level corresponding to each failure characteristic.

2. The system according to claim 1, characterized in that, Before inputting the failure characteristics into the preset failure prediction model, the method further includes: In response to an early warning request from a liquid-cooled battery pack, multiple abnormal battery packs are identified, which are obtained by performing different failure simulations on multiple normal battery packs respectively. Multiple sample failure features of the multiple abnormal battery packs are obtained and multiple sample risk levels of the multiple abnormal battery packs are obtained. The multiple sample failure features include: damage features, coolant failure features and thermal failure features. The model is trained based on the failure characteristics and risk levels of the multiple samples to obtain the preset failure prediction model.

3. The system according to claim 2, characterized in that, The model is trained based on the failure characteristics and risk levels of the multiple samples to obtain the preset failure prediction model, including: Multi-dimensional parameters are extracted from the failure features of the multiple samples to obtain multi-dimensional failure features, which include at least two of the following: mechanical features, electrical features, temperature features, gas features, and image features. The model is trained based on the multi-dimensional failure characteristics and the risk levels of the multiple samples to obtain the failure prediction model.

4. The system according to claim 3, characterized in that, The model is trained based on the multi-dimensional failure characteristics and the multiple sample risk levels to obtain the failure prediction model, including: Determine the application scenario for each abnormal battery pack; The model is trained based on the multi-dimensional failure characteristics, the risk levels of the multiple samples, and the application scenario of each abnormal battery pack to obtain the predictive failure model.

5. The system according to claim 4, characterized in that, The model is trained based on the multi-dimensional failure characteristics, the risk levels of the multiple samples, and the application scenario of each abnormal battery pack to obtain the failure prediction model, including: Based on the multi-dimensional failure characteristics, the risk levels of the multiple samples, and the application scenario of each abnormal battery pack, the abnormal battery pack is grouped according to its operating conditions to obtain a training dataset. The training dataset includes data samples of multiple types of operating conditions, including normal operating conditions, minor leakage operating conditions, moderate wetting failure operating conditions, and thermal runaway precursor operating conditions. The model is iteratively trained based on the training dataset, and the model feature weights and correlation parameters are continuously optimized until the model meets the preset convergence condition, thus obtaining the prediction failure prediction model.

6. The system according to claim 1, characterized in that, Obtaining the failure characteristics of the battery pack under test includes: Retrieve multi-dimensional real-time operating data of the battery pack under test from its battery management system. The multi-dimensional real-time working data is subjected to noise reduction and working condition filtering to obtain filtered working data. According to the preset feature extraction rules, data is extracted from the filtered working data to obtain the failure features related to coolant leakage.

7. The system according to claim 6, characterized in that, According to preset feature extraction rules, data is extracted from the filtered working data to obtain the failure features related to coolant leakage, including: Based on the structural deformation feature extraction rules, continuous fluctuations and abnormal deviations in parameters of the liquid cooling circuit pressure and flow rate parameters in the filtered working data are identified to obtain the damage characteristics. Based on the leakage and infiltration timing feature extraction rules, the step-like decay of insulation resistance and abnormal fluctuation of liquid cooling parameters are identified from the high-voltage insulation resistance and liquid cooling circuit dynamic parameters in the screened working data, and the coolant failure characteristics are obtained. Based on the temperature anomaly feature extraction rules, from the cell temperature rise rate, battery module temperature difference, and charging and discharging electrical parameters in the filtered working data, the thermal imbalance, local overheating, and abnormal temperature rise conditions of the battery pack are identified, and thermal failure features are obtained. The damage feature, the coolant failure feature, and the thermal failure feature are identified as the failure features.

8. The system according to claim 7, characterized in that, The damage characteristics, the coolant failure characteristics, and the thermal failure characteristics are identified as the failure characteristics, including: The damage features, coolant failure features, and thermal failure features are time-series aligned, and isolated data with time mismatches are removed to obtain the failure features.

9. The system according to any one of claims 1-8, characterized in that, After generating the warning instruction, the method further includes: Obtain the real-time operating status of the battery pack under test, as well as the historical operating status and historical predicted risk level of the battery packs in the same batch under test; A correlation analysis is performed on the real-time operating conditions, the historical operating conditions, the historical predicted risk level, the predicted risk level, and the target early warning strategy to obtain the failure evolution law; Based on the failure evolution law, the mapping relationship is updated to obtain an updated mapping relationship.

10. A method for early warning of a battery pack, characterized in that, include: Obtain the failure characteristics of the battery pack under test; Input the failure characteristics into the preset failure prediction model to obtain the predicted risk level output by the preset failure prediction model; Based on the predicted risk level, the target early warning strategy corresponding to the predicted risk level is determined from the mapping relationship between the preset risk level and the preset early warning strategy. A warning instruction is generated according to the target warning strategy. The warning instruction is used to instruct the battery management system corresponding to the battery pack under test to execute the warning strategy. The warning strategy includes: fault indication and continuous monitoring, limiting charging and discharging power, enhancing heat dissipation, locking the high voltage circuit, and shutdown protection. The preset failure prediction model is determined after training based on the failure characteristics of all samples and the risk level corresponding to each failure characteristic.

11. A battery pack, characterized in that, The battery pack is a normal battery pack determined through a warning process, which is performed by the warning system of the battery pack according to any one of claims 1-9.

12. An electric vehicle, characterized in that, It includes at least the battery pack as described in claim 11.

13. An electrical appliance, characterized in that, It includes at least the battery pack as described in claim 11.

Images