Battery fault early warning method, battery management system and vehicle

Abstract

The invention discloses a battery fault early warning method, a battery management system and a vehicle, and the battery fault early warning method comprises the steps: obtaining the consistency key characteristic value of each battery cell in a battery pack in each complete charge-discharge cycle of the battery pack, and calculating the deviation degree of the consistency key characteristic value of each battery cell and a health baseline, determining health baselines based on statistical distribution of consistency key characteristic values; obtaining a track of the deviation degree of each battery cell changing along with the time sequence as a deviation track; performing time sequence analysis on the deviation track of each battery cell to obtain dynamic early warning feature information of each battery cell; and when the dynamic early warning feature information satisfies an early warning condition, performing early warning processing. According to the method provided by the invention, dynamic evolution characteristic identification of consistency abnormity of each battery cell in the battery pack is realized, potential abnormity identification before further development of battery degradation is facilitated, and the timeliness of early warning is improved.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and in particular to a battery fault early warning method, a battery management system, and a vehicle. Background Technology

[0002] Currently, Battery Management Systems (BMS) primarily rely on instantaneous values ​​of parameters such as voltage and temperature, combined with preset fixed thresholds, to monitor battery pack inconsistencies. For example, when the voltage difference or temperature of a cell exceeds a set threshold, a corresponding fault alarm is triggered. However, this static judgment method, based on comparing instantaneous parameters with fixed thresholds, only focuses on whether the parameters exceed the limit at a certain moment and cannot detect abnormal trends below the threshold. When cell parameters have not yet exceeded the threshold but are already showing a continuous deterioration trend, the system struggles to detect the anomaly in a timely manner, easily missing the optimal opportunity to provide early warning of potential faults. Summary of the Invention

[0003] This invention aims to at least solve one of the technical problems existing in the prior art. Therefore, one objective of this invention is to propose a battery fault early warning method. This method enables the identification of dynamic evolution characteristics of inconsistencies in the individual cells of a battery pack, which is beneficial for identifying potential anomalies before further battery degradation, thus improving the timeliness of early warning.

[0004] The second objective of this invention is to provide a battery management system.

[0005] The third objective of this invention is to provide a vehicle.

[0006] To achieve the above objectives, a battery fault early warning method according to a first aspect of the present invention includes: obtaining a consistency key feature value of each cell in the battery pack during each complete charge-discharge cycle of the battery pack; calculating the deviation of the consistency key feature value of each cell from a health baseline, wherein the health baseline is determined based on the statistical distribution of the consistency key feature value; obtaining a trajectory of the deviation of each cell changing over time as a deviation trajectory; performing time-series analysis on the deviation trajectory of each cell to obtain dynamic early warning feature information of each cell; and performing early warning processing when the dynamic early warning feature information meets the early warning conditions.

[0007] According to the battery fault early warning method of this invention, in each complete charge-discharge cycle, firstly, a consistency key feature value is extracted for each cell in the battery pack. This consistency key feature value is a characteristic quantity that can comprehensively reflect the electrochemical performance and working state of the cell within a cycle, thus it can relatively stably characterize the health state and performance level of the cell under that cycle. Simultaneously, a health baseline is established based on the statistical distribution of the consistency key feature values ​​of all cells. This health baseline reflects the overall consistency level and normal fluctuation range of the battery pack in the current operating stage. Then, the consistency key feature value of each cell is compared with the health baseline, and its deviation is calculated. This quantifies the degree of deviation of each cell from the overall state of the battery pack. This deviation directly reflects whether the cell has begun to deviate from the overall cluster and generate potential inconsistency degradation, thereby transforming the originally difficult-to-perceive slight inconsistency into statistically significant deviation information. Based on this, by continuously recording the changes in the deviation of each cell in chronological order and forming a deviation trajectory, the evolution process of the cell's consistency state with the number of cycles can be reflected, thereby revealing whether there is a trend of continuously expanding or gradually deteriorating inconsistency. Then, time-series analysis is performed on the deviation trajectory to extract dynamic early warning feature information, enabling the system to identify potential anomalies based solely on their deviation trends and evolution characteristics before key consistency feature values ​​show significant exceedances. Thus, this invention, through the aforementioned logical chain of "feature value—deviation degree—deviation trajectory—time-series analysis," achieves the identification of dynamic evolution characteristics of consistency anomalies in each cell of the battery pack. Compared to existing technologies that rely solely on instantaneous parameters and fixed thresholds, this approach can capture the nascent stages of battery degradation earlier, thereby providing early warning before further degradation and fundamentally improving the timeliness of fault early warning.

[0008] In some embodiments, the key consistency features include at least one of the following: the charging end voltage, the voltage change rate over a preset battery state of charge range, the resting voltage drop, and the highest temperature and maximum temperature rise during a complete charge-discharge cycle.

[0009] In some embodiments, the battery fault warning method further includes: obtaining the consistency key feature value of each cell in multiple charge-discharge cycles to form a health cluster; calculating the average value and standard deviation of each consistency key feature value in the health cluster; and establishing the health baseline based on the average value and the standard deviation.

[0010] In some embodiments, the battery fault warning method further includes updating the health baseline at preset intervals.

[0011] In some embodiments, calculating the deviation of the consistency key feature value of each cell from the health baseline includes obtaining the deviation based on the consistency key feature value, the average value corresponding to the consistency key feature value, and the standard deviation corresponding to the consistency key feature value.

[0012] In some embodiments, the dynamic warning feature information includes at least one of the absolute value of the deviation and the slope of the deviation trajectory; the warning condition includes at least one of a first warning condition and a second warning condition; the first warning condition includes the absolute value of the deviation of the cell exceeding a deviation threshold in a plurality of consecutive charge-discharge cycles; the second warning condition includes the slope of the deviation trajectory of the cell remaining positive over time.

[0013] In some embodiments, when the dynamic early warning feature information meets the early warning conditions, early warning processing is performed, including: executing a graded early warning strategy based on the situation where the dynamic early warning feature information meets the early warning conditions.

[0014] In some embodiments, the battery fault early warning method further includes: collecting the voltage, temperature and total current of each cell in the battery pack and storing them as historical data with time tags, so as to extract the consistency key feature value of each cell from the historical data during the charge-discharge cycle.

[0015] To achieve the above objectives, a battery management system according to a second aspect of the present invention includes: a plurality of sensors for collecting the total current of the battery pack and the voltage and current of each cell; and a battery manager connected to the plurality of sensors for implementing the battery fault early warning method described in the above embodiment.

[0016] According to the battery management system of this invention, the battery fault early warning method described in the above embodiment is used. In each complete charge-discharge cycle, firstly, a consistency key feature value is extracted for each cell in the battery pack. This consistency key feature value is a characteristic quantity that can comprehensively reflect the electrochemical performance and working state of the cell within a cycle, thus it can relatively stably characterize the health state and performance level of the cell under that cycle. At the same time, a health baseline is established based on the statistical distribution of the consistency key feature values ​​of all cells. This health baseline reflects the overall consistency level and normal fluctuation range of the battery pack in the current operating stage. Then, the consistency key feature value of each cell is compared with the health baseline and its deviation is calculated. This can quantify the degree of deviation of each cell from the overall state of the battery pack. This deviation can directly reflect whether the cell has begun to deviate from the overall cluster and generate potential inconsistency degradation, thereby transforming the originally difficult-to-perceive slight inconsistency into statistically significant deviation information. On this basis, by continuously recording the changes in the deviation of each cell in chronological order and forming a deviation trajectory, the evolution process of the cell's consistency state with the number of cycles can be reflected, thereby revealing whether there is a trend of continuous expansion or gradual deterioration of inconsistency development. Then, time-series analysis is performed on the deviation trajectory to extract dynamic early warning feature information, enabling the system to identify potential anomalies based solely on their deviation trends and evolution characteristics before key consistency feature values ​​show significant exceedances. Thus, this invention, through the aforementioned logical chain of "feature value—deviation degree—deviation trajectory—time-series analysis," achieves the identification of dynamic evolution characteristics of consistency anomalies in each cell of the battery pack. Compared to existing technologies that rely solely on instantaneous parameters and fixed thresholds, this approach can capture the nascent stages of battery degradation earlier, thereby providing early warning before further degradation and fundamentally improving the timeliness of fault early warning.

[0017] To achieve the above objectives, a vehicle according to a third aspect of the present invention includes a battery pack and a battery management system as described in the above embodiment, wherein the battery management system is connected to the battery pack, and the battery pack includes a plurality of cells connected in series and parallel.

[0018] According to the vehicle of the present invention, the battery management system described in the above embodiment is used. By connecting the battery management system to the battery pack, the dynamic evolution characteristics of the consistency abnormality of each cell in the battery pack can be identified. This is beneficial to identify potential abnormalities before the battery deteriorates further and improves the timeliness of the early warning.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a battery fault early warning method according to an embodiment of the present invention; Figure 2 This is an overall flowchart of a battery fault early warning method according to an embodiment of the present invention; Figure 3 This is a block diagram of a battery management system according to an embodiment of the present invention; Figure 4 This is a block diagram of a vehicle according to an embodiment of the present invention.

[0021] Figure label: 100 vehicles; Battery Management System 1; Battery Pack 2; Multiple sensors 11; battery manager 12. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. The embodiments of the present invention are described in detail below.

[0023] With the rapid development of electric vehicles and large-scale energy storage, the safety and reliability of power battery packs have become a focus of industry attention. Battery packs can consist of hundreds or even thousands of cells connected in series and parallel. Due to differences in manufacturing processes, initial conditions, usage environments, and aging rates, inconsistencies are unavoidable among the cells. These inconsistencies manifest as differences in parameters such as voltage, temperature, and internal resistance, and are key factors affecting the overall performance, lifespan, and even safety of the battery pack. If not effectively managed, these inconsistencies will intensify with each cycle, leading to overcharging, over-discharging, or overheating of some cells, potentially causing serious safety incidents such as thermal runaway. Therefore, real-time and effective monitoring and diagnosis of battery pack inconsistencies is one of the core functions of a battery management system. Currently, Battery Management Systems (BMS) primarily rely on instantaneous values ​​of parameters such as voltage and temperature, combined with preset fixed thresholds, to monitor battery pack inconsistencies. For example, when the voltage difference or temperature of a cell exceeds a set threshold, a corresponding fault alarm is triggered. However, this static judgment method, based on comparing instantaneous parameters with fixed thresholds, only focuses on whether the parameter exceeds the limit at a certain moment, completely ignoring the entire process of the parameter's slow deterioration before reaching the threshold. Therefore, when the cell's parameters have not yet exceeded the threshold but have already shown a continuous and slow deterioration trend, the system struggles to detect the anomaly in a timely manner, easily missing the best opportunity to provide early warning of potential faults, resulting in a significant delay in warnings.

[0024] In addition, existing technologies also employ a scheme based on equivalent circuit models, attempting to diagnose through model prediction. However, the model parameters often fail to accurately depict the unique trajectory of each individual cell in the battery pack as it ages. When a cell begins to deviate from the overall degradation path, its abnormal signal changes are small and easily drowned out by model errors and noise, resulting in extreme insensitivity to early, weak consistency faults.

[0025] Therefore, existing technologies cannot effectively identify small, gradual performance degradation trends within battery packs, resulting in late warnings of faults caused by inconsistent evolution, or even triggering alarms only when a fault has already occurred or is about to cause serious safety risks.

[0026] To address the above issues, this invention proposes a battery fault early warning method. This method does not rely on fixed thresholds or equivalent circuit models. By capturing and analyzing the dynamic evolution characteristics of battery pack inconsistencies, it achieves early, accurate, and predictive early warning of potential cell faults, thereby significantly advancing the fault detection point and improving the safety and reliability of the battery pack.

[0027] The following is for reference. Figures 1-2 A battery fault early warning method according to an embodiment of the present invention is described.

[0028] Figure 1 This is a flowchart of a battery fault early warning method according to an embodiment of the present invention, such as... Figure 1 As shown, the battery fault warning method of this embodiment of the invention includes at least steps S1-S4, as detailed below.

[0029] S1. In each complete charge-discharge cycle of the battery pack, obtain the consistency key characteristic value of each cell in the battery pack, and calculate the deviation of the consistency key characteristic value of each cell from the health baseline, wherein the health baseline is determined based on the statistical distribution of the consistency key characteristic value.

[0030] In some embodiments, a complete charge-discharge cycle can refer to the battery pack gradually completing the charging process from a lower state of charge to reaching a preset charging termination condition, and then completing the discharging process to reach a preset discharging termination condition, or vice versa, allowing the battery cell to undergo a complete, continuous, and comparable energy input and release process. Compared to incomplete cycles that only include the charging process or only include the discharging process, a complete charge-discharge cycle can cover the comprehensive behavioral characteristics of the battery cell under different states of charge, different current directions, and different operating stress conditions, thereby more comprehensively reflecting the electrochemical reactions and thermal behavior changes of the battery cell.

[0031] Therefore, by using a complete charge-discharge cycle as the basic period for feature extraction, it is possible to ensure that the obtained key consistency features cover the main electrochemical reaction stages and thermal behavior changes of the cell within one operating cycle. This avoids the bias and random operating condition interference caused by extracting data only during the charging or discharging phases. Compared to incomplete charge-discharge processes, key consistency features obtained under complete charge-discharge cycles have better stability, repeatability, and comparability, and are more conducive to reflecting the true differences in performance and degradation trends between cells. Therefore, they are more suitable for consistency analysis and subsequent deviation calculations.

[0032] In some embodiments, the key consistency feature value for each cell can refer to a feature quantity extracted from the time-series operating data of the cell during a complete charge-discharge cycle, which can comprehensively reflect the electrochemical performance and operating state of the cell. This feature value is not simply instantaneous sampled data, but is obtained by summarizing and extracting the behavior of the cell under specific operating conditions or stages, thus more realistically characterizing the cell's health status and performance changes. By extracting key consistency feature values ​​on a cycle scale, the performance differences between different cells can be presented in a comparable form, providing a reliable basis for subsequent consistency analysis and deviation judgment.

[0033] In some embodiments, a health baseline can refer to a reference benchmark established based on the statistical distribution of key consistency characteristic values ​​of each cell within the battery pack. This benchmark characterizes the overall health level and normal consistency status of the battery pack at the current stage. Determining the health baseline based on the statistical distribution of key consistency characteristic values ​​indicates that statistical analysis of the key consistency characteristic values ​​of each cell within the battery pack yields statistical parameters that represent the group's centrality and dispersion, thus forming a reference benchmark reflecting the overall state of the battery pack. The health baseline is not a fixed absolute threshold but dynamically reflects the normal consistency level at the current stage as the overall state of the battery pack changes.

[0034] In some embodiments, after establishing a healthy baseline, for each subsequent complete charge-discharge cycle, the degree of difference in the behavior of a single cell relative to the overall battery pack can be quantitatively characterized by calculating the deviation of each cell's key consistency characteristic value from the healthy baseline. This deviation reflects the degree to which an individual cell deviates from the group average level in a statistical sense, serving as a core quantitative indicator for characterizing the evolution of inconsistencies, rather than a simple comparison of parameter magnitudes. By introducing the calculation of deviation, cells that have not yet reached a fixed threshold but have begun to continuously deviate from the group's behavior can be effectively identified, thus providing a foundation for subsequently constructing deviation trajectories and analyzing the dynamic evolution process of inconsistencies. It is precisely because of this deviation-based quantification method that the system can shift its focus from "whether the limit is exceeded" to "whether there is deviation and how it evolves," laying a crucial foundation for achieving early warning.

[0035] S2, obtain the trajectory of the deviation of each cell changing over time, as the deviation trajectory.

[0036] In some embodiments, obtaining the trajectory of the deviation of each cell over time, as the deviation trajectory, is to extend the static deviation result under a single charge-discharge cycle into a temporal representation that can reflect the process of cell consistency change. Since the degradation of cell performance and the formation of inconsistencies is a slow, gradual, and dynamic process, the deviation calculated in a single cycle only reflects the instantaneous deviation of the cell from the healthy baseline in that cycle, making it difficult to distinguish between temporary deviations caused by random fluctuations and persistent anomalies caused by cell degradation. However, by recording the deviation changes of the same cell in chronological order over multiple consecutive charge-discharge cycles, the resulting deviation trajectory can intuitively characterize the evolution trend of that cell relative to the overall consistency level of the battery pack.

[0037] Therefore, deviation trajectories can reflect whether the deviation behavior of a battery cell exhibits sporadic, periodic, or continuous and cumulative characteristics. When the deviation trajectory of a battery cell shows a continuous increase or a long-term maintenance at a high deviation level over time, it means that the performance of that cell has begun to systematically deviate from the overall state of the battery pack, posing a potential risk of degradation or failure. Compared to judging based solely on a single deviation, constructing and analyzing deviation trajectories can effectively reveal the dynamic process of inconsistency evolving gradually with the number of cycles, thus providing basic data support for subsequent time-series analysis and dynamic early warning feature identification.

[0038] S3 performs time-series analysis on the deviation trajectory of each battery cell to obtain dynamic early warning characteristic information for each battery cell.

[0039] In some embodiments, time-series analysis of the deviation trajectory of each cell can refer to performing mathematical or statistical processing on the deviation trajectory formed by the cell in multiple consecutive charge-discharge cycles over time to identify its variation patterns and evolution trends. Since the deviation trajectory reflects the process of cell consistency evolving relative to a healthy baseline over cycles, time-series analysis of this trajectory helps to distinguish between random fluctuations, short-term disturbances, and abnormal changes with persistence and directionality.

[0040] In some embodiments, time-series analysis can employ methods such as linear fitting, moving average, and abrupt change detection. Linear fitting, by fitting the deviation trajectory, can obtain the overall trend of the deviation degree changing with the number of cycles, characterizing the long-term direction and rate of change of the cell's consistency characteristics. Moving average processing of the deviation trajectory can reduce the influence of occasional fluctuations or measurement noise on the trajectory morphology, making the potential long-term trend clearer. Abrupt change detection can be used to detect whether there are obvious state transitions or abrupt behaviors in the deviation trajectory, reflecting the possible phased changes in the cell's consistency state. Therefore, through the above-mentioned time-series analysis methods, diagnostically significant evolutionary features can be extracted from the deviation trajectory.

[0041] In some embodiments, dynamic early warning feature information can refer to comprehensive information obtained from the time-series analysis results of deviation trajectories, which can characterize the evolutionary features of cell consistency status. Dynamic early warning feature information can comprehensively reflect the degree of deviation, trend, or persistence of cell consistency anomalies, and is used to characterize whether cell consistency anomalies have a tendency to continue developing or worsening. By obtaining dynamic early warning feature information, the evolutionary patterns implicit in the deviation trajectory can be transformed into characteristic criteria that can be used for judgment, thereby providing a basis for determining subsequent early warning conditions.

[0042] S4: When the dynamic early warning feature information meets the early warning conditions, early warning processing is performed.

[0043] In some embodiments, the warning conditions are used to determine whether the battery cell has deviated from its healthy state and has the risk of further deterioration. Specifically, the warning conditions can be set based on the deviation trajectory change pattern reflected in the dynamic warning feature information, and are used to comprehensively reflect whether the deviation of the key characteristic value of battery cell consistency from the healthy baseline shows abnormal characteristics such as continuous deviation or unstable evolution trend.

[0044] In some embodiments, performing early warning processing when dynamic early warning feature information meets early warning conditions helps to identify potential consistency anomalies before a cell exhibits a significant fault. By performing early warning processing, risk warning information specific to a particular cell can be provided to the battery management system to support subsequent maintenance decisions, operational strategy adjustments, or the triggering of safety protection measures, thereby avoiding overall battery pack performance degradation or safety hazards caused by the spread of degradation in a single cell.

[0045] According to the battery fault early warning method of this invention, in each complete charge-discharge cycle, firstly, a consistency key feature value is extracted for each cell in the battery pack. This consistency key feature value is a characteristic quantity that can comprehensively reflect the electrochemical performance and working state of the cell within a cycle, thus it can relatively stably characterize the health state and performance level of the cell under that cycle. Simultaneously, a health baseline is established based on the statistical distribution of the consistency key feature values ​​of all cells. This health baseline reflects the overall consistency level and normal fluctuation range of the battery pack in the current operating stage. Then, the consistency key feature value of each cell is compared with the health baseline, and its deviation is calculated. This quantifies the degree of deviation of each cell from the overall state of the battery pack. This deviation directly reflects whether the cell has begun to deviate from the overall cluster and generate potential inconsistency degradation, thereby transforming the originally difficult-to-perceive slight inconsistency into statistically significant deviation information. Based on this, by continuously recording the changes in the deviation of each cell in chronological order and forming a deviation trajectory, the evolution process of the cell's consistency state with the number of cycles can be reflected, thereby revealing whether there is a trend of continuously expanding or gradually deteriorating inconsistency. Then, time-series analysis is performed on the deviation trajectory to extract dynamic early warning feature information, enabling the system to identify potential anomalies based solely on their deviation trends and evolution characteristics before key consistency feature values ​​show significant exceedances. Thus, this invention, through the aforementioned logical chain of "feature value—deviation degree—deviation trajectory—time-series analysis," achieves the identification of dynamic evolution characteristics of consistency anomalies in each cell of the battery pack. Compared to existing technologies that rely solely on instantaneous parameters and fixed thresholds, this approach can capture the nascent stages of battery degradation earlier, thereby providing early warning before further degradation and fundamentally improving the timeliness of fault early warning.

[0046] In some embodiments, key consistency characteristics include at least one of the following: charging end voltage, voltage change rate over a preset battery state of charge range, resting voltage drop, and the highest temperature and maximum temperature rise during a complete charge-discharge cycle. These characteristics are specific features that are highly sensitive to consistency changes, effectively condensing the cell's health status information.

[0047] In some embodiments, the charging terminal voltage value can refer to the terminal voltage of the battery cell at the end of the constant current charging phase, which is highly sensitive to changes in cell capacity and internal resistance. Specifically, under constant current charging conditions, the terminal voltage of the battery cell can be considered to be jointly determined by the open-circuit voltage of the cell and the voltage drop across the cell's internal resistance caused by the charging current. The open-circuit voltage is closely related to the effective capacity and electrochemical activity state of the cell. As the cell ages, its capacity decay leads to changes in the open-circuit voltage characteristics under the corresponding state of charge. Simultaneously, the cell's internal resistance increases with the number of cycles, resulting in a larger voltage drop across the internal resistance under the same charging current. Therefore, the charging terminal voltage value comprehensively reflects the combined effects of changes in cell capacity and internal resistance, and can serve as an important characteristic parameter characterizing changes in cell consistency and aging state.

[0048] In some embodiments, the voltage change rate within a preset battery state of charge (SOC) range can refer to the rate of change of cell voltage with varying SOC or charge / discharge capacity within a selected SOC range, reflecting the cell's polarization characteristics within that SOC range. The preset SOC range can be a range where cell voltage changes relatively smoothly (e.g., 20%-80%), or a range where the voltage change slope is significant (e.g., 80%-100%). Within different SOC ranges, the cell's voltage change characteristics are closely related to its internal ohmic polarization, electrochemical polarization, and concentration polarization processes. By calculating the voltage change rate within the preset SOC range, the degree of cell polarization and its changing trend within that range can be reflected. As the number of cycles increases or cell performance deteriorates, polarization characteristics will change, causing the voltage change rate to shift between different cycles. Therefore, the voltage change rate within a preset SOC range can be used to characterize the evolution of the cell's internal polarization state and is an important characteristic quantity reflecting cell consistency.

[0049] In some embodiments, the resting voltage drop refers to the voltage change that occurs when a battery cell, after completing its charging process, is left to rest for a preset period of time under no-current conditions, causing its terminal voltage to drop from its initial value back to a stable value. The resting voltage drop process reflects the reduction of polarization within the battery cell and the re-equilibrium of the electrochemical reaction. Its magnitude and rate of change are closely related to the stability of the electrochemical processes within the battery cell. When a battery cell suffers from aging, structural degradation, or uneven internal reactions, its voltage relaxation characteristics during the resting process will change, causing the resting voltage drop to deviate from the normal level. Therefore, the resting voltage drop can be used to characterize the electrochemical stability and consistency within the battery cell.

[0050] In some embodiments, the highest temperature and maximum temperature rise during a complete charge-discharge cycle are used to characterize the thermal properties and heat dissipation of the battery cell during operation. The highest temperature reflects the extreme thermal state reached by the battery cell during charge-discharge, while the maximum temperature rise reflects the degree of heat generation relative to the initial temperature. The internal heating behavior of the battery cell is closely related to its internal resistance, electrochemical reaction efficiency, and energy loss. When the battery cell ages or exhibits abnormal consistency, its internal resistance increases or side reactions intensify, leading to increased heat generation.

[0051] Therefore, by monitoring and analyzing the highest temperature and maximum temperature rise during a complete charge-discharge cycle, changes in cell consistency can be reflected from the perspective of thermal behavior, providing an auxiliary basis for identifying potential cell failures or degradation risks.

[0052] In some embodiments, the battery fault early warning method further includes: obtaining the consistency key feature values ​​of each cell in multiple charge-discharge cycles to form a health cluster; calculating the mean and standard deviation of each consistency key feature value in the health cluster; and establishing a health baseline based on the mean and standard deviation.

[0053] In embodiments of the present invention, multiple charge-discharge cycles can be selected as several complete charge-discharge cycles during the initial use of the battery pack (e.g., the first 100 cycles). During this stage, the cells have not yet shown significant aging or abnormal degradation, and their electrochemical performance and thermal characteristics are generally in a relatively stable state, making them suitable as reference samples for characterizing the health status of the cells. By collecting consistent key characteristic values ​​of each cell in multiple charge-discharge cycles during this stage, the impact of random fluctuations in a single charge-discharge cycle on the statistical results can be effectively reduced, making the subsequently established health baseline more representative and robust.

[0054] In some embodiments, a health cluster can refer to a data set formed by the consistent key characteristic values ​​of each cell in a healthy state during multiple charge-discharge cycles. As the basic data set for establishing a health baseline, the health cluster aims to avoid relying solely on a single measurement result for judgment, thereby improving the stability and reliability of battery fault warnings.

[0055] In some embodiments, a health baseline (μ, δ) is established based on the mean (μ) and standard deviation (δ) of each consistency key feature value in the health cluster. Specifically, statistical analysis is performed on each type of consistency key feature value in the health cluster. The mean value is used to characterize the central level of the feature under healthy conditions, and the standard deviation is used to characterize the normal fluctuation range of the feature under healthy conditions, thereby forming a statistical reference benchmark for subsequent consistency assessment. By using statistical distribution characteristics to construct the health baseline, the health baseline can adapt to the actual characteristic differences of different battery packs and different operating conditions, providing a more objective and flexible reference basis for deviation calculation and dynamic evolution analysis, which is beneficial to improving the accuracy and timeliness of battery fault early warning.

[0056] In some embodiments, updating the health baseline at preset intervals can refer to dynamically adjusting the health baseline during continuous operation of the battery pack, based on the consistency key characteristic values ​​obtained by the same battery pack in subsequent charge-discharge cycles. The update of the health baseline is not based on data from other battery packs, but on the operating data of the battery pack when it is determined to be still in a healthy state, thereby reflecting the normal evolution trend of the battery pack's consistency characteristics over time.

[0057] In some embodiments, the health baseline can be updated by re-statistically analyzing the distribution characteristics of key consistency features within a preset time interval. For example, within a new time window or cyclic window, the corresponding statistical parameters for the key consistency features of each cell can be recalculated, and the original health baseline can be corrected without affecting the sensitivity of anomaly identification, so that the health baseline can be updated in line with the slow changes in the overall performance of the battery pack.

[0058] In some embodiments, the deviation of the consistency key feature value of each cell from the healthy baseline is calculated, including obtaining the deviation based on the consistency key feature value, the average value corresponding to the consistency key feature value, and the standard deviation corresponding to the consistency key feature value.

[0059] Specifically, the key consistency feature value can be compared with the average value established based on the healthy cluster, and the difference can be scaled by combining the standard deviation corresponding to the key consistency feature value, thereby obtaining a deviation index that reflects the degree to which an individual cell deviates from the center of the group. By introducing the standard deviation as a normalization factor, not only can the influence of different feature dimensions and numerical ranges be eliminated, but the deviation results can also be made comparable and stable.

[0060] In some embodiments, deviation can be characterized using a standard score, calculated as follows: deviation equals the difference between the key consistency feature value and the mean, divided by the standard deviation. This method allows deviation to be used as a unified scale to intuitively reflect the degree of deviation of a single cell from the healthy baseline during the current charge-discharge cycle. This provides a unified and adaptive benchmark for evaluating different cells and different aging stages, thus providing a unified and quantifiable basic input for subsequent deviation trajectory construction, time-series analysis based on deviation trajectories, and extraction of dynamic early warning feature information.

[0061] In summary, by employing methods such as Z-Score (standard score) based on cluster statistics to calculate the deviation, the evaluation benchmark (mean μ and standard deviation δ) is derived from the health baseline established by the battery pack itself. This means that the warning threshold is essentially a statistical quantity relative to the current health state of the system, rather than an absolutely fixed physical value. It adaptively adjusts as the battery pack ages. Therefore, this invention avoids the awkward situation of a fixed threshold being too lenient for new batteries and too sensitive for older batteries, enabling the system to maintain high sensitivity and low false alarm rate throughout the entire battery lifespan, possessing strong adaptive capabilities.

[0062] In some embodiments, the dynamic warning feature information includes at least one of the absolute value of the deviation and the slope of the deviation trajectory.

[0063] The absolute value of deviation characterizes the magnitude of a single cell's deviation from the healthy baseline during the current charge-discharge cycle. A larger absolute value indicates a more significant statistical deviation of the cell's key consistency characteristics from the healthy baseline, suggesting that the cell's electrochemical performance or operating state has significantly deviated from the normal behavior range of most cells in the battery pack. Therefore, the absolute value of deviation directly reflects the severity of cell consistency anomalies and is an important quantitative indicator for determining whether a cell exhibits abnormalities.

[0064] In some embodiments, the slope of the deviation trajectory is used to characterize the trend of deviation over the number of charge-discharge cycles or time. The slope obtained by fitting or trend analysis of the deviation trajectory can reflect whether the cell is deviating from the healthy baseline in a state of continuous deterioration.

[0065] The warning conditions include at least one of a first warning condition and a second warning condition. The first warning condition includes the absolute value of the cell's deviation exceeding a deviation threshold over multiple consecutive charge-discharge cycles. This first warning condition indicates that the cell has been consistently and significantly deviating from the overall consistency level of the battery pack for a long period, rather than experiencing occasional transient fluctuations. This situation reflects a persistently high deviation characteristic of the cell, meaning that its key characteristics such as electrochemical performance, thermal properties, or internal resistance have substantially differed from those of healthy cells, posing a high potential risk of failure.

[0066] In some embodiments, the deviation threshold can be set based on statistical principles, such as the standard deviation range of key consistency features in the healthy baseline, such as setting 2δ or 3δ as the deviation threshold, in order to suppress false alarms caused by random noise while ensuring the sensitivity of anomaly detection.

[0067] In some embodiments, the second warning condition includes a continuously positive slope for the deviation trajectory of the battery cell over time. This second warning condition indicates that the consistency deviation of the battery cell is not stable, but rather shows a continuously aggravating trend across multiple charge-discharge cycles. Specifically, when the slope of the deviation trajectory remains positive over multiple cycles, it means that the deviation is continuously increasing over time, i.e., the difference between the battery cell and the healthy cluster is continuously widening. This situation reflects that the internal performance degradation process of the battery cell is progressing, with obvious trend and directional characteristics. Even if the current deviation has not yet reached a severely abnormal level, it can be regarded as an early signal of potential failure, thereby achieving a proactive warning of potential failures.

[0068] In some embodiments, when the dynamic early warning feature information meets the early warning conditions, early warning processing is performed, including: executing a graded early warning strategy based on the situation where the dynamic early warning feature information meets the early warning conditions.

[0069] Specifically, the system can differentiate and set warning levels based on the degree and trend of cell consistency anomalies reflected in the dynamic warning feature information. For example, when only a cell is detected to have an absolute deviation exceeding the deviation threshold in multiple consecutive charge-discharge cycles, indicating that the cell is in a state of continuous high deviation but has not yet shown a significant accelerated degradation trend, a lower-level Level 1 warning can be triggered to prompt maintenance personnel to focus on and continuously monitor the cell. However, when a continuous high deviation is detected accompanied by a continuously positive deviation trajectory slope, indicating that the cell's consistency anomaly is worsening over time, a higher-level Level 2 warning can be triggered to indicate that the cell has a high potential failure risk. In addition, after triggering a Level 2 warning, preventive suggestions such as reducing the cell's charging current, limiting the charge-discharge rate of suspicious cells, arranging inspections and planned maintenance, and preparing spare parts in advance can be provided. This expands the battery fault warning from a simple anomaly indication to a graded response mechanism that combines risk levels, achieving an orderly connection from anomaly identification to warning indication and subsequent handling suggestions. Therefore, this invention can proactively prevent the deterioration of potential faults and even the occurrence of safety accidents, while reducing unexpected downtime through preventative recommendations, thereby greatly improving the safety and operational reliability of the entire battery pack.

[0070] In some embodiments, by establishing a hierarchical early warning strategy mechanism based on dynamic early warning feature information, the system can issue early warnings based solely on the deterioration trend of key consistency feature values ​​before they exceed any fixed safety threshold, thereby achieving a substantial forward shift of the early warning node.

[0071] In some embodiments, the battery fault early warning method further includes: collecting the voltage, temperature and total current of each cell in the battery pack and storing them as time-stamped historical data, so as to extract consistent key feature values ​​of each cell from the historical data during charge-discharge cycles.

[0072] In some embodiments, the voltage, temperature, and total current of each cell in the battery pack can be acquired using standard sensors set in the battery management system (BMS). This eliminates the need for dedicated detection hardware such as electrochemical impedance spectroscopy (EIS) and avoids reliance on massive amounts of comprehensive fault data for model training. This significantly reduces the cost and complexity of the technology, making it easy to deploy directly on existing BMS hardware platforms via software upgrades. It also facilitates large-scale promotion and application in fields such as electric vehicles and large-scale energy storage, thereby improving engineering feasibility and applicability.

[0073] In some embodiments, storing the collected data as time-stamped historical data can refer to attaching corresponding collection time information to each data point when collecting and recording various raw parameters, and storing them in chronological order, so that each parameter forms a continuous and ordered data sequence in the time dimension. By setting time tags for the collected raw data, a complete time-series data record can be formed, enabling accurate tracing of the state changes of each cell under different charge-discharge cycles and different operating stages. This historical data is not only used to extract key consistency feature values ​​within a single charge-discharge cycle, but also provides basic data support for subsequent deviation trajectory construction and time-series analysis, thereby ensuring the continuity and consistency of the dynamic analysis process.

[0074] In some embodiments, introducing time-stamped historical data is also beneficial for distinguishing the chronological order of changes in different key consistency features over time, thereby assisting in the location of abnormal cells and the analysis of root causes of failures. Specifically, by sorting the voltage and temperature characteristics of the same cell at different time points and constructing deviation trajectories, the system can determine the order in which different features deviate. For example, when the deviation trajectory corresponding to the voltage-related features of a cell deviates earlier than the temperature-related features, it can be determined that the consistency anomaly of the cell is first manifested in the voltage characteristics; conversely, when the deviation of the temperature-related features occurs before the voltage features, it can be considered that the thermal characteristic anomaly may be a precursor signal of the abnormal evolution. Through this time-stamped time-series comparative analysis, the present invention can not only identify "whether there is an anomaly," that is, accurately locate the individual cells or modules in the battery pack that show the earliest deterioration trend, but also further distinguish the chronological order of deviations of different features, providing a basis for subsequent analysis of the causes of anomalies (such as loose connections, increased internal resistance, and poor cooling), thereby improving the pertinence and effectiveness of fault diagnosis.

[0075] Therefore, this invention overcomes the shortcomings of existing technologies that can only alarm "there is a problem with the system" but cannot locate "where the problem is" or "what the problem may be", and achieves accurate fault location and root cause diagnosis, thereby improving maintenance efficiency.

[0076] Figure 2 This is an overall flowchart of a battery fault early warning method according to an embodiment of the present invention, as follows: Figure 2 As shown, the overall process of the battery fault early warning method includes at least the following steps, as detailed below.

[0077] S10 collects the voltage, temperature, and total current of each cell in the battery pack and stores it as historical data with time tags.

[0078] S11. In each complete charge-discharge cycle of the battery pack, obtain at least one of the following key consistency characteristics of each cell in the battery pack: the charging end voltage value, the voltage change rate of the preset battery state of charge range, the static voltage drop value, and the highest temperature and maximum temperature rise value in a complete charge-discharge cycle.

[0079] S12, obtain the consistency key feature value of each cell in multiple charge and discharge cycles as a health cluster, calculate the average value and standard deviation of each consistency key feature value in the health cluster, establish a health baseline based on the average value and standard deviation, and update the health baseline at preset time intervals.

[0080] S13, the deviation is obtained based on the consistency key feature value, the mean value corresponding to the consistency key feature value, and the standard deviation value corresponding to the consistency key feature value.

[0081] S14, obtain the trajectory of the deviation degree of each cell changing over time as the deviation trajectory, and perform time-series analysis on the deviation trajectory of each cell to obtain dynamic warning feature information of each cell. The dynamic warning feature information includes at least one of the absolute value of the deviation degree and the slope of the deviation trajectory.

[0082] S15, if the dynamic early warning feature information meets the early warning conditions, execute the graded early warning strategy.

[0083] In summary, this invention continuously tracks and quantifies the deviation of key consistency characteristic values ​​of each cell in a battery pack from a healthy baseline, constructs the deviation trajectory of each cell, and identifies potential anomalies from its subtle changes over time, thereby achieving early detection and warning of battery faults. Compared with existing technologies that rely solely on instantaneous parameters or fixed thresholds for judgment, this invention significantly advances the fault detection time from "during" or "after" to "before," achieving true early warning.

[0084] Furthermore, at the engineering implementation level, this invention makes full use of the voltage, temperature and current data collected by standard sensors in existing battery management systems (BMS), without the need to introduce dedicated hardware such as electrochemical impedance spectroscopy (EIS) or rely on massive amounts of fault samples for model training. This allows the method to be deployed in software on existing BMS platforms, with low implementation costs and good engineering feasibility, making it suitable for large-scale applications.

[0085] In addition, this invention can not only trigger fault warnings, but also accurately locate abnormal cells based on the timing characteristics of deviations from the trajectory, and identify their degradation trend characteristics, providing maintenance personnel with clear diagnostic basis and assisting in making targeted maintenance decisions.

[0086] Therefore, this invention achieves early warning of battery failure in terms of early warning effect, improves the active protection capability of the system in terms of safety, achieves accurate location and trend identification of abnormal cells in terms of diagnostic capability, has the ability to adaptively update with the operating status in terms of performance, and takes into account both economy and scalability in engineering applications, thereby significantly improving the overall safety and reliability of the power battery system.

[0087] The following is for reference. Figure 3 A battery management system according to an embodiment of the present invention is described.

[0088] Figure 3 This is a block diagram of a battery management system according to an embodiment of the present invention, such as... Figure 3 As shown, the battery management system 1 includes: multiple sensors 11 and a battery manager 12.

[0089] In some embodiments, multiple sensors 11 are used to collect the total current of the battery pack and the voltage and current of each cell.

[0090] In some embodiments, the plurality of sensors 11 may include, but are not limited to, a voltage sensor, a temperature sensor, and a current sensor. The current sensor can be used to collect the total current information of the battery pack during charging and discharging; the voltage sensor can be used to collect the individual cell voltage of each cell within the battery pack; and the temperature sensor can be used to collect the temperature information of each cell or the module containing the cell, to reflect the thermal characteristics and heat dissipation of the cell during operation.

[0091] In some embodiments, the multiple sensors 11 are standard configuration sensors in the battery management system 1, and the data they collect can be directly used to implement the battery fault early warning method described in the above embodiments without the need for additional dedicated detection hardware. By utilizing the sensor resources already deployed in the existing BMS hardware platform, the present invention can identify the dynamic evolution characteristics of abnormal consistency of each cell in the battery pack without changing the original hardware structure, simply by deploying the corresponding software algorithm in the battery manager 12. This eliminates the need for additional dedicated detection hardware such as electrochemical impedance spectroscopy (EIS) and the need to rely on massive and complete fault data for model training, thereby reducing system implementation costs and improving engineering feasibility.

[0092] In some embodiments, the battery manager 12 may refer to the main control unit or controller in the battery management system 1, which includes a processor and a memory for processing and analyzing data collected by multiple sensors 11.

[0093] In some embodiments, the battery manager 12 is connected to multiple sensors 11 to implement the battery fault early warning method described in the above embodiments, including operations such as extracting consistent key feature values, establishing and updating health baselines, calculating deviation, constructing deviation trajectories, performing time series analysis, early warning judgment, and executing graded early warning strategies.

[0094] According to the battery management system 1 of this embodiment, the battery fault early warning method described in the above embodiment is used. In each complete charge-discharge cycle, firstly, a consistency key feature value is extracted for each cell in the battery pack. This consistency key feature value is a characteristic quantity that can comprehensively reflect the electrochemical performance and working state of the cell in a cycle, and therefore can relatively stably characterize the health state and performance level of the cell in that cycle. At the same time, a health baseline is established based on the statistical distribution of the consistency key feature values ​​of all cells. This health baseline reflects the overall consistency level and normal fluctuation range of the battery pack in the current operating stage. Then, the consistency key feature value of each cell is compared with the health baseline and its deviation is calculated. This can quantify the degree of deviation of each cell from the overall state of the battery pack. This deviation can directly reflect whether the cell has begun to deviate from the overall cluster and generate potential inconsistency degradation, thereby transforming the originally difficult-to-perceive slight inconsistency into statistically significant deviation information. On this basis, by continuously recording the changes in the deviation of each cell in chronological order and forming a deviation trajectory, the evolution process of the cell's consistency state with the number of cycles can be reflected, thereby revealing whether there is a trend of continuous expansion or gradual deterioration of inconsistency development. Then, time-series analysis is performed on the deviation trajectory to extract dynamic early warning feature information, so that the system can identify potential anomalies based solely on their deviation trend and evolution characteristics before the key consistency feature values ​​show obvious overruns.

[0095] Therefore, this invention, through the above-mentioned logical chain consisting of "feature value - deviation degree - deviation trajectory - time series analysis", realizes the identification of the dynamic evolution characteristics of the consistency abnormality of each cell in the battery pack. Compared with the existing technology that judges based only on instantaneous parameters and fixed thresholds, it can capture the budding stage of battery degradation earlier, thereby completing the early warning before the battery degradation develops further, fundamentally improving the timeliness of fault warning.

[0096] The following is for reference. Figure 4 A vehicle according to an embodiment of the present invention is described.

[0097] Figure 4 This is a block diagram of a vehicle according to an embodiment of the present invention, such as... Figure 4 As shown, the vehicle 100 includes a battery pack 2 and a battery management system 1 as described in the above embodiment.

[0098] In some embodiments, the battery pack 2 in the vehicle 100 serves as the main energy supply unit for the entire vehicle, providing electrical energy to the vehicle 100's drive system, on-board control system, and on-board electrical equipment. For example, the battery pack 2 can provide driving energy for the drive motor and provide stable power support for low-voltage or high-voltage electrical units such as vehicle controllers, sensors, communication modules, and thermal management systems. By configuring the battery management system 1 described in the above embodiments in the vehicle 100, the battery management system 1 is connected to the battery pack 2, and performs real-time monitoring and management of the operating status of each cell in the battery pack 2, ensuring the safety, reliability, and service life of the battery pack 2 during charging, discharging, and driving.

[0099] In some embodiments, battery pack 2 includes multiple cells connected in series and parallel. By connecting multiple cells in series, the output voltage of battery pack 2 can be increased to meet the voltage level requirements of the vehicle's drive motor and high-voltage electrical system; by connecting multiple cells in parallel, the capacity and output current capability of battery pack 2 can be increased, thereby improving the driving range and power performance of vehicle 100. Simultaneously, the series-parallel combination can also distribute the load of individual cells to a certain extent, coping with high-power charging and discharging conditions. Since cell variations are inevitable in manufacturing, aging, and usage environments, inconsistency issues are more likely to arise after multiple cells are connected in series and parallel. Therefore, using the battery fault early warning method described in the above embodiments to identify and warn of the dynamic evolution characteristics of abnormal cell inconsistency is beneficial for identifying potential anomalies before further battery degradation, improving the timeliness of early warning.

[0100] In some embodiments, vehicle 100 may be a type of vehicle that uses a power battery as its primary or important energy source, such as a pure electric vehicle, a plug-in hybrid electric vehicle, or a range-extended electric vehicle. Furthermore, vehicle 100 may also be an electric bus, an electric truck, an electric special-purpose vehicle, or even engineering machinery or special-purpose vehicles equipped with an on-board power battery system. This invention does not limit the specific form of vehicle 100; as long as vehicle 100 is equipped with a battery pack 2 composed of multiple cells connected in series and parallel, and the battery pack 2 is managed by a battery management system 1, the technical solution of this invention can be applied, thereby improving the timeliness of early warning of battery consistency anomalies and the overall vehicle operational safety.

[0101] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0102] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A battery fault early warning method, characterized in that, include: In each complete charge-discharge cycle of the battery pack, the consistency key feature value of each cell in the battery pack is obtained, and the deviation of the consistency key feature value of each cell from the health baseline is calculated, wherein the health baseline is determined based on the statistical distribution of the consistency key feature value; Obtain the trajectory of the deviation degree of each of the battery cells changing sequentially over time, as the deviation trajectory; A time-series analysis is performed on the deviation trajectory of each battery cell to obtain dynamic early warning feature information for each battery cell; When the dynamic early warning feature information meets the early warning conditions, early warning processing is performed.

2. The battery fault early warning method according to claim 1, characterized in that, The key consistency features include at least one of the following: the charging end voltage value, the voltage change rate of the preset battery state of charge range, the static voltage drop value, and the highest temperature and maximum temperature rise value in a complete charge-discharge cycle.

3. The battery fault early warning method according to claim 1, characterized in that, The battery fault early warning method also includes: Obtain the consistency key feature value of each cell in multiple charge-discharge cycles as a healthy cluster; Calculate the mean and standard deviation of each of the consistency key features in the healthy cluster; The health baseline is established based on the mean and the standard deviation.

4. The battery fault early warning method according to claim 3, characterized in that, The battery fault early warning method further includes updating the health baseline at preset intervals.

5. The battery fault early warning method according to claim 1, characterized in that, The calculation of the deviation of the consistency key characteristic value of each of the battery cells from the health baseline includes: The deviation is obtained based on the consistency key feature value, the average value corresponding to the consistency key feature value, and the standard deviation corresponding to the consistency key feature value.

6. The battery fault early warning method according to claim 1, characterized in that, The dynamic early warning feature information includes at least one of the absolute value of the deviation and the slope of the deviation trajectory; The warning conditions include at least one of the first warning conditions and the second warning conditions; The first warning condition includes the absolute value of the deviation of the battery cell exceeding the deviation threshold in multiple consecutive charge-discharge cycles; The second warning condition includes the slope of the deviation trajectory of the battery cell remaining positive over time.

7. The battery fault early warning method according to claim 1, characterized in that, When the dynamic early warning feature information meets the early warning conditions, early warning processing is performed, including: Based on the dynamic early warning feature information meeting the early warning conditions, a tiered early warning strategy is executed.

8. The battery fault early warning method according to any one of claims 1-7, characterized in that, The battery fault early warning method further includes: collecting the voltage, temperature and total current of each cell in the battery pack and storing them as historical data with time tags, so as to extract the consistency key feature value of each cell from the historical data during the charge and discharge cycle.

9. A battery management system, characterized in that, include: Multiple sensors are used to collect the total current of the battery pack and the voltage and current of each cell; A battery manager connected to a plurality of the sensors for implementing the battery fault warning method according to any one of claims 1-8.

10. A vehicle, characterized in that, The vehicle includes a battery pack and a battery management system as described in claim 9, the battery management system being connected to the battery pack, the battery pack including a plurality of cells connected in series and parallel.

Images