A method and system for intelligent detection of short circuit faults in lithium batteries

Abstract

This invention relates to the technical field of lithium battery short-circuit fault detection, and provides an intelligent detection method and system for lithium battery short-circuit faults. The method includes: acquiring the operating parameters and environmental parameters corresponding to the battery cells of a lithium battery; establishing and continuously updating a personalized operating characteristic record for each battery cell based on the operating parameters and environmental parameters; predicting the normal behavior parameters of each battery cell under the current operating conditions based on the personalized operating characteristic record, operating parameters, and environmental parameters; comparing the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and performing cumulative analysis on the real-time deviation value to identify abnormal deviation patterns of the battery cell; triggering a graded early warning based on the abnormal deviation pattern, and executing measures corresponding to the graded early warning. This invention improves the sensitivity and accuracy of fault detection.

Description

Technical Field

[0001] This invention relates to the technical field of lithium battery short-circuit fault detection, and specifically to an intelligent method and system for detecting lithium battery short-circuit faults. Background Technology

[0002] In the power systems of new energy vehicles, the safe operation of lithium battery packs is crucial. However, due to long-term use and complex operating environments, the cells inside lithium battery packs are prone to abnormal connections or deterioration in insulation performance, leading to short-circuit faults. Traditional detection methods often rely on a single voltage criterion, which frequently results in false alarms in practical applications and makes it difficult to detect slowly developing short-circuit risks with inconspicuous initial characteristics under varying operating conditions.

[0003] When a battery cell begins to experience very early, slow-developing insulation aging or a micro-short circuit—for example, when lithium dendrites slowly grow and gradually pierce the separator, forming a tiny, high-impedance localized short circuit point—the voltage drop signal caused by this short circuit is extremely weak, possibly only a few millivolts. Furthermore, due to the instability of the short circuit path and environmental fluctuations, this signal is intermittent and unstable. In actual operation, this weak and unstable fault signal is not only completely drowned out by the normal large voltage fluctuations caused by daily driving behavior and changes in ambient temperature, but it is also highly likely to be ignored as ordinary noise by the filtering programs in the battery management control unit designed for stable operating conditions. Even if this signal manages to pass through the filtering stage, its amplitude is far from reaching the intentionally high alarm threshold set to avoid false alarms. Therefore, the battery management system is completely unable to effectively identify and warn of this gradual, initially subtle short circuit risk in its nascent stage. The existence of this "blind spot" causes the system to miss the best time to intervene, greatly increasing the potential risk of more serious and sudden short-circuit failures later on, and even thermal runaway.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] This application discloses an intelligent detection method and system for short-circuit faults in lithium batteries, aiming to solve the technical problems of high false alarm rate and difficulty in detecting slowly developing short-circuit risks in the early stage under complex operating conditions in existing lithium battery short-circuit fault detection methods.

[0006] The technical solution of this application is as follows:

[0007] In a first aspect, this application discloses an intelligent detection method for short-circuit faults in lithium batteries, including:

[0008] Obtain the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery;

[0009] Based on operating parameters and environmental parameters, a corresponding personalized operating characteristic record is established and continuously updated for each battery cell;

[0010] Based on personalized operating characteristic records, operating parameters, and environmental parameters, predict the normal behavior parameters of each battery cell under the current operating conditions;

[0011] The actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. The real-time deviation value is then accumulated and analyzed to identify the abnormal deviation pattern of the battery cell.

[0012] Based on the abnormal deviation pattern, a tiered early warning is triggered, and the corresponding measures are executed.

[0013] Through this technical solution, this application can effectively distinguish between normal fluctuations and abnormal deviations by establishing personalized operating characteristic records and predicting normal behavior parameters. This improves the accuracy of early short-circuit fault identification under complex and ever-changing operating conditions, avoids false alarms and missed alarms caused by traditional methods based on a single threshold judgment, and significantly improves the safe operation level of lithium batteries.

[0014] Secondly, this application also discloses a lithium battery short-circuit fault intelligent detection system for performing intelligent detection of lithium battery short-circuit faults, including:

[0015] The battery parameter acquisition module is used to acquire the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery.

[0016] The feature record update module is used to establish and continuously update the corresponding personalized operating feature record for each battery cell based on operating parameters and environmental parameters.

[0017] The normal parameter prediction module is used to predict the normal behavior parameters of each battery cell under the current operating conditions based on personalized operating characteristics, operating parameters, and environmental parameters.

[0018] The abnormal pattern recognition module is used to compare the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and to perform cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell.

[0019] The early warning execution module is used to trigger tiered early warnings based on abnormal deviation patterns and execute measures corresponding to the tiered early warnings.

[0020] Through this technical solution, this application can provide a complete system that integrates data acquisition, characteristic recording, parameter prediction, anomaly identification and early warning execution, thereby realizing intelligent, full-process detection and management of lithium battery short-circuit faults, and effectively improving the safety and reliability of the battery system.

[0021] Beneficial Effects: The intelligent detection method for short-circuit faults in lithium batteries disclosed in this application acquires the operating parameters and environmental parameters of lithium battery cells, and establishes and continuously updates personalized operating characteristic records for each battery cell, fully considering individual differences in batteries and the impact of dynamic operating conditions. Based on this, the method can predict the normal behavior parameters of each battery cell under the current operating conditions according to the personalized operating characteristic records, operating parameters, and environmental parameters, thus establishing a dynamic and accurate benchmark. By comparing the actual operating parameters of the battery cell with the predicted normal behavior parameters, the real-time deviation value is obtained, and the real-time deviation value is accumulated and analyzed. This application can effectively identify abnormal deviation patterns of battery cells, avoiding the false alarms and missed alarms caused by the single fixed voltage threshold judgment in traditional methods. Finally, based on the identified abnormal deviation patterns, a graded early warning is triggered and corresponding measures are executed, achieving early, accurate early warning and effective intervention for short-circuit faults.

[0022] Compared to existing technologies, this application overcomes the challenge of traditional methods confusing voltage fluctuations caused by driving behavior, battery aging, and external noise under complex and variable operating conditions with early short-circuit fault signals. By establishing personalized operating characteristic records and dynamically predicting normal behavior parameters, this application can effectively distinguish between normal aging deviations of the battery, transient operating condition fluctuations, and initial weak short-circuit fault signals, significantly improving the sensitivity and accuracy of fault detection. Furthermore, the tiered early warning mechanism and corresponding measures provide more refined management based on the severity and development trend of the fault, avoiding excessive intervention or insufficient response. This ensures the safe operation of the lithium battery while minimizing the impact on vehicle performance and driving experience. Therefore, this application effectively solves the blind spot problem in short-circuit fault detection in existing technologies, significantly improving the safety and reliability of lithium battery systems. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for intelligent detection of short-circuit faults in lithium batteries according to one embodiment of the present invention;

[0024] Figure 2 This is a flowchart of a method for intelligent detection of short-circuit faults in lithium batteries according to another embodiment of the present invention;

[0025] Figure 3 This is a system block diagram of an intelligent detection system for short-circuit faults in lithium batteries according to another embodiment of the present invention;

[0026] Explanation of reference numerals in the attached figures:

[0027] 1. Intelligent detection system for short circuit faults in lithium batteries; 11. Battery parameter acquisition module; 12. Characteristic record update module; 13. Normal parameter prediction module; 14. Abnormal pattern recognition module; 15. Early warning measure execution module. Detailed Implementation

[0028] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0029] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] This application proposes an intelligent detection method for short-circuit faults in lithium batteries, combining... Figure 1 As shown, it includes:

[0031] S1, obtain the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery;

[0032] S2, based on operating parameters and environmental parameters, establishes and continuously updates corresponding personalized operating characteristic records for each battery cell;

[0033] S3 predicts the normal behavior parameters of each battery cell under the current operating conditions based on personalized operating characteristic records, operating parameters, and environmental parameters.

[0034] S4. The actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. The real-time deviation value is then accumulated and analyzed to identify the abnormal deviation pattern of the battery cell.

[0035] S5, based on the abnormal deviation pattern, triggers a graded warning and executes the corresponding measures.

[0036] To facilitate understanding of the technical solutions proposed in this application, the key terms and their technical meanings are explained below.

[0037] "Operating parameters" refer to the electrical and thermal parameters that are collected in real time during the operation of the lithium battery and reflect the working status of the battery cell. These parameters include, but are not limited to, the battery cell's terminal voltage, current, temperature, state of charge / discharge, state of charge, and health status. These parameters directly characterize the actual operating behavior of the battery cell at the current moment. "Environmental parameters" refer to the external operating conditions of the battery cell and their dynamic changes, including ambient temperature, humidity, altitude, vehicle speed, acceleration, vibration intensity, and external electromagnetic interference. These parameters have a significant impact on the battery cell's performance, aging rate, and fault evolution process.

[0038] "Personalized operating characteristic records" refer to a historical characteristic data set for each individual battery cell, formed based on operating and environmental parameters collected during its long-term operation. This data set reflects the individual differences and performance evolution patterns of the battery cell, and is used to characterize individual operating features such as internal resistance variation trends, capacity decay behavior, self-discharge characteristics, and temperature sensitivity under different operating conditions. "Normal behavior parameters" refer to the predicted operating parameter performance of a healthy battery cell, based on the personalized operating characteristic records of the corresponding battery cell under current operating and environmental conditions. "Real-time deviation value" refers to the difference between the actual operating parameters of the battery cell and the normal behavior parameters, used to quantify the degree of deviation of the battery cell's current operating state from its expected normal state.

[0039] "Abnormal deviation pattern" refers to a deviation behavior pattern identified after cumulative analysis of the real-time deviation value over a time dimension, exhibiting specific temporal evolution characteristics, amplitude variation characteristics, or statistical distribution characteristics. This type of pattern corresponds to the early or developmental characteristics of fault types such as internal short circuits in lithium batteries. "Graded early warning" refers to classifying early warning information into levels based on the severity, evolution speed, and potential risk level of the identified abnormal deviation pattern, in order to achieve differentiated responses. "Measures" refers to the safety response strategies automatically executed or prompted by the system for different early warning levels.

[0040] In practical implementation, the intelligent detection method for short-circuit faults in lithium batteries proposed in this application may include the following steps.

[0041] First, the operating and environmental parameters corresponding to the battery cells are acquired. Specifically, voltage, current, and temperature sensors can be installed on each battery cell to collect electrical and thermal data in real time. Simultaneously, vehicle operating status information can be obtained through the vehicle bus system, combined with external sensor data such as ambient temperature and humidity, as environmental parameter inputs. The acquired data is then transmitted to the battery management system or central processing unit for subsequent analysis and processing.

[0042] Subsequently, based on the aforementioned operating and environmental parameters, a personalized operating characteristic record is established and continuously updated for each battery cell. This process can be achieved by modeling the historical operating and environmental data of the battery cell, for example, based on time series analysis, state evolution models, or regularized parameter update mechanisms, to dynamically characterize the battery cell's internal resistance changes, capacity decay, and operating condition response characteristics. By continuously introducing new operating data, the personalized operating characteristic record can be continuously corrected as the battery is used, thereby reflecting the true health status and behavioral characteristics of the battery cell.

[0043] Based on this, and according to the personalized operating characteristic records, combined with the current operating parameters and environmental parameters, the normal behavior parameters of the battery cell under the current operating conditions can be predicted. Specifically, the historical response relationships contained in the personalized operating characteristic records can be used to estimate the voltage, internal resistance, or temperature rise levels that the battery cell should exhibit in a fault-free state, given the current current, temperature, and environmental conditions.

[0044] Next, the actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the corresponding real-time deviation values, and cumulative analysis is performed on these real-time deviation values. Since single sampling may be affected by noise, instantaneous load changes, or measurement errors, the cumulative analysis is used to highlight persistent and trend-based deviation behaviors, thereby suppressing the interference of short-term fluctuations on the judgment results. By analyzing the changing trend, amplitude characteristics, and statistical properties of the cumulative deviation values, abnormal deviation patterns matching internal short-circuit faults in lithium batteries can be identified.

[0045] Finally, based on the identified abnormal deviation patterns, a tiered warning system is triggered, and measures corresponding to the warning level are implemented. Specifically, when the abnormal deviation pattern is small in magnitude and develops slowly, a lower-level warning can be triggered, and measures such as risk alerts or operational restrictions can be taken. When the abnormal deviation pattern shows a continuous amplification or rapid deterioration trend, a higher-level warning can be triggered, and safety control measures including power limiting, functional degradation, or battery cell isolation can be implemented, thereby reducing safety risks before the fault develops further.

[0046] In one embodiment, personalized operating characteristic records are established using individual battery cells as an index to describe the baseline response characteristics of individual battery cells under different operating conditions. The personalized operating characteristic records include at least: a set of baseline response parameters under voltage, current, and temperature-related operating conditions; an equivalent parameter set (e.g., a parameter set characterizing the ohmic internal resistance component and polarization component); and a set of deviation statistics formed within a fault-free sample segment (e.g., the statistical mean and statistical dispersion of voltage and internal resistance deviations). Optionally, the personalized operating characteristic records are updated using a sliding window method, i.e., the update window is composed of the most recent W sampling periods, and historical statistics are recursively updated using a forgetting factor λ, allowing the records to be gradually corrected as the battery state evolves, while reducing the impact of instantaneous disturbances on the records; wherein W and λ can be preset or adaptively adjusted according to the coverage of vehicle operating conditions.

[0047] In one embodiment, normal behavior parameters are obtained using a "baseline model output and operating condition correction" method. Specifically, firstly, based on the equivalent parameter set stored in the personalized operating characteristic record, the predicted terminal voltage and predicted equivalent internal resistance are calculated under the current SOC, temperature, and current conditions. Secondly, corrections related to environmental disturbances are introduced to obtain the set of normal behavior parameters. Optionally, the corrections are implemented using a lookup table method, that is, a correspondence between "disturbance level - allowable fluctuation range" is pre-established in the personalized operating characteristic record. During real-time prediction, the corresponding allowable bandwidth is selected according to the disturbance level, and the reference range of the predicted terminal voltage and predicted equivalent internal resistance is boundary-corrected accordingly, thereby forming a dynamic normal reference interval.

[0048] In one embodiment, the real-time deviation value includes at least the voltage deviation ΔV(t) and the internal resistance deviation ΔR(t), and a normalized deviation index η(t) is constructed. η(t) is calculated based on the deviation and the corresponding statistical dispersion to achieve comparability between different battery cells; to avoid division by zero, a very small positive term can be introduced. The cumulative analysis of the deviation can be implemented using exponential weighted moving average or cumulative sum statistics: for example, a recursive statistic E(t) = α·η(t) + (1-α)·E(t-1) is constructed. When E(t) meets the preset judgment condition within K consecutive sampling periods, it is determined that there is a continuous abnormal deviation; at the same time, feature quantities (such as change slope, duration, recovery speed, etc.) for characterizing evolution behavior can be optionally extracted from ΔV(t) and ΔR(t) to form a feature vector of abnormal deviation pattern for subsequent pattern matching and hierarchical indication. α is an exponential weighting coefficient used to determine the weighting of "current deviation information η(t)" and "historical statistics E(t-1)" in E(t). α preferably satisfies 0 < α < 1 and is a dimensionless parameter. When α is large, E(t) responds faster and is more sensitive to the current deviation η(t), but its ability to suppress transient noise is relatively reduced. When α is small, E(t) is smoother and more robust, and its false alarm suppression is better, but its early response speed to persistent abnormal deviations is relatively reduced. Therefore, α can be engineered according to the noise level, sampling period, and the speed of target anomaly evolution (e.g., appropriately reducing α when noise is amplified, and appropriately increasing α when it is necessary to improve the speed of capturing abnormal trends) to achieve a controllable trade-off between "response speed and noise robustness".

[0049] Optional, combined Figure 2 As shown, the process of comparing the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and then performing cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell, can be further refined into the following sub-steps:

[0050] A1 compares the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell.

[0051] A2 decomposes the real-time deviation value to obtain several deviation components;

[0052] A3. Match the deviation components according to the preset internal short-circuit fault feature template to obtain the matching result;

[0053] A4. Based on the matching results, determine the abnormal type of the battery cell to identify the abnormal deviation pattern of the battery cell.

[0054] The comparison between the actual operating parameters of the battery cell and the predicted normal behavior parameters involves real-time monitoring of the battery cell's voltage, current, temperature, internal resistance, and other actual operating parameters, and comparing them with the normal behavior parameters predicted based on personalized operating characteristic records and current operating conditions. This quantifies the difference between the battery cell's current state and its normal state, thus obtaining the battery cell's real-time deviation value. This real-time deviation value reflects the degree of abnormality of the battery cell at a specific moment.

[0055] Furthermore, the real-time deviation value is decomposed into several deviation components. This can be understood as separating a single real-time deviation signal into sub-signals with different frequency, amplitude, and duration characteristics through signal processing techniques (such as wavelet decomposition, Fourier transform, or empirical mode decomposition). These sub-signals, i.e., deviation components, may represent different types of anomalies or noise, such as slowly changing trend deviations, transient spike deviations, or periodic fluctuation deviations. The aim is to deconstruct the complex deviation signal into independent components that are easier to analyze and identify.

[0056] In practical applications, deviation components are matched against a pre-defined internal short-circuit fault feature template to obtain matching results. This template, built upon extensive historical data and expert knowledge, encompasses typical deviation patterns of different types and severity of internal short-circuit faults in parameters such as voltage, current, temperature, and internal resistance. Examples include a slow decrease in voltage, a continuous increase in internal resistance, and an abnormal increase in local temperature. By performing similarity calculations or pattern recognition on the decomposed deviation components and these templates, the degree of agreement between the current deviation component and known internal short-circuit fault patterns can be assessed, thus yielding the matching results.

[0057] Therefore, based on the matching results, the anomaly type of the battery cell is determined to identify its abnormal deviation pattern. When the matching results show that a certain deviation component is highly similar to a specific internal short-circuit fault feature template, it can be determined that the battery cell may have a corresponding internal short-circuit fault. For example, if the deviation component highly matches the template of "micro-short circuit causing slow voltage drop," then the battery cell can be identified as having a micro-short circuit abnormal deviation pattern. This determination process makes anomaly identification more specific and accurate.

[0058] In one embodiment, when decomposing the real-time deviation value, the decomposition output is defined as a set of mutually distinguishable deviation components {C_trend, C_cycle, C_impulse}, where C_trend represents low-frequency trend components, used to characterize slowly developing persistent deviations; C_cycle represents mid-frequency periodic components, used to characterize repetitive deviations related to operating cycle; and C_impulse represents high-frequency transient components, used to characterize intermittent abnormal channels or sudden disturbances. The fault feature templates are stored in the form of a template library, and each template includes at least: trend template parameters (e.g., target slope range and duration range), periodic template parameters (e.g., target frequency band range and energy distribution range), and transient template parameters (e.g., peak density and peak amplitude distribution range). During matching, the matching degree between each deviation component and the corresponding template parameters can be optionally calculated separately, and synthesized into a comprehensive matching index according to a preset judgment order; when the comprehensive matching index meets the preset judgment conditions and remains consistent within multiple consecutive windows, the abnormal type corresponding to the template is output as the abnormal deviation pattern recognition result.

[0059] Optionally, the steps for triggering a tiered early warning based on the abnormal deviation pattern and executing the corresponding measures include:

[0060] Obtain vehicle operating status information and road environment information;

[0061] Based on operational status information and road environment information, determine the driver's driving intention;

[0062] Obtain pre-defined tiered early warnings and corresponding measures;

[0063] Based on the abnormal deviation pattern, assess the potential impact of the measures corresponding to the graded early warning on vehicle driving safety and driving experience;

[0064] By considering the severity of the abnormal deviation pattern, the risk development trend, the vehicle operating environment, the driver's intention, and the potential impact, the implementation strategy and timing of the corresponding measures are adjusted to obtain the adjusted measures information;

[0065] The human-computer interaction interface conveys information about the adjusted measures, the reasons for the adjustments, and suggested driving behaviors.

[0066] Specifically, acquiring vehicle operating status and road environment information refers to collecting real-time vehicle operating parameters such as speed, acceleration, steering angle, braking status, and gear information, as well as environmental data such as current road type, traffic flow, weather conditions, and obstacles ahead, through onboard sensors, vehicle bus data, and navigation systems. This information is used to construct a comprehensive vehicle operating context.

[0067] In this context, determining the driver's driving intention based on the aforementioned operational status and road environment information can be understood as analyzing the vehicle's dynamic behavior (e.g., rapid acceleration, rapid deceleration, frequent lane changes, steering angle change rate, etc.) in conjunction with the road environment (e.g., highways, congested urban areas, curves, etc.) using machine learning models or expert systems to infer whether the driver is currently engaged in normal cruising, emergency avoidance, overtaking, or parking. The aim is to enable the system to more intelligently understand the driver's current needs and potential actions, thereby avoiding inappropriate interventions at critical moments.

[0068] In practical applications, obtaining preset graded early warnings and corresponding measures means that the system internally stores a series of early warning schemes (e.g., Level 1 mild warning, Level 2 moderate warning, Level 3 severe warning) for different degrees of anomaly severity and risk level, as well as corresponding countermeasures (e.g., only providing a notification, limiting partial power output, forced shutdown, disconnecting the battery circuit, etc.). These preset schemes are formulated based on a large amount of fault data and safety specifications.

[0069] Furthermore, based on the aforementioned abnormal deviation patterns, assessing the potential impact of tiered warning measures on vehicle safety and driving experience means that, after determining the abnormal deviation pattern of the battery cell, the system simulates or predicts the possible consequences of different warning measures in the current vehicle operating context. For example, if the abnormal deviation pattern indicates a minor short circuit, and the vehicle is traveling at high speed, the system will assess whether immediately limiting power output would increase the risk of vehicle loss of control or severely impact the driver's comfort. The aim is to ensure that the measures taken, while addressing battery malfunctions, do not introduce new safety hazards or significantly degrade the user experience.

[0070] Therefore, considering the severity of the aforementioned abnormal deviation patterns, risk development trends, vehicle operating environment, driving intentions, and the potential impacts, the system adjusts the execution strategy and timing of corresponding measures, resulting in adjusted measure information. This means the system will make multi-dimensional comprehensive decisions. For example, if the abnormal deviation pattern is severe but the vehicle is in an emergency braking state, the system may prioritize ensuring braking performance, temporarily delaying or weakening battery-related intervention measures; if the abnormal deviation pattern is minor and the driver is in normal cruising mode, the system may choose a milder warning method and suggest the driver check the vehicle at a safe time. The adjusted measure information includes the specific warning level, warning content, measure type, execution intensity, and optimal execution timing.

[0071] Ultimately, the adjusted measures, reasons for the adjustments, and suggested driving behaviors are conveyed through the human-machine interface. The human-machine interface can take various forms, including instrument panel displays, central control screen prompts, voice announcements, and seat vibrations. Communicating the reasons for the adjustments helps the driver understand the rationale behind the system's decisions, while suggested driving behaviors (such as "Please stop at the next service area to check" or "Please slow down and pay attention to the instrument panel") provide clear guidance for the driver to collaboratively address potential risks.

[0072] In one embodiment, the "potential impact" is represented by a calculable impact index, which includes at least a driving safety-related index I_safe and a driving experience-related index I_exp. Optionally, I_safe can be calculated from a risk score formed by vehicle speed, road curvature, distance between vehicles, braking margin, etc.; I_exp can be calculated from an experience score formed by power response limitation, achievable acceleration reduction ratio, and warning frequency, etc. The system generates candidate execution schemes for preset measures and calculates the comprehensive cost J=ω1·I_safe+ω2·I_exp+ω3·I_fault, where I_fault is used to characterize the degree of suppression of abnormal risk evolution (e.g., the degree of power / current limitation coverage or the strength of suppression of abnormal channel expansion). When preset scenario conditions are met (e.g., high-speed driving, complex road conditions, or safe parking), the weights ω1, ω2, and ω3 can be optionally adjusted to determine the execution order and timing of the measures, so as to achieve coordination between safety and availability.

[0073] In some preferred embodiments, a specific example is given below. Suppose an electric vehicle is traveling at 100 km / h on a highway, at which point the battery management system detects a slight internal short circuit anomaly in a battery cell, the severity of which is judged to be moderate, but the risk development trend is slow.

[0074] If a traditional method is used, the system might immediately trigger a level 2 moderate warning, such as limiting the vehicle's power output by 20% and displaying a warning on the dashboard: "Battery malfunction, please stop immediately for inspection." However, at high speeds, a sudden power limitation could cause a sharp drop in vehicle speed, affecting the safety of vehicles behind and potentially leading to rear-end collisions. It would also cause significant psychological stress and a poor driving experience for the driver.

[0075] The solution proposed in this application will be processed as follows:

[0076] First, the system acquires vehicle operating status information (high-speed driving, stable cruising) and road environment information (highway, moderate traffic flow).

[0077] Secondly, based on this information, the system determines that the driver's driving intention is normal cruising.

[0078] Next, the system assesses the potential impact of the preset moderate warning measures (limiting power output by 20%) on vehicle driving safety and driving experience in the current high-speed driving scenario, and believes that it may lead to safety risks and a poor experience.

[0079] Finally, the system comprehensively considers the moderate severity of the abnormal deviation pattern, the slow risk development trend, the vehicle operating environment on the highway, the driver's normal cruising intentions, and the potential impact to adjust the implementation strategy and timing of measures. For example, the system may not immediately limit power, but instead adjust the warning level to a Level 1 mild warning, conveying the adjusted measures through the human-machine interface (e.g., displayed on the central control screen): "The battery system has a slight anomaly and has been optimized. Please stop at the nearest service area to check," and providing the reason for the adjustment and suggested driving behavior. This approach not only promptly informs the driver of potential risks but also avoids taking measures that may endanger safety at inappropriate times, thereby maximizing the driving experience while ensuring safety.

[0080] Optionally, the step of decomposing the real-time deviation value into several deviation components includes:

[0081] Obtain the real-time deviation value of the battery cell;

[0082] Multi-time-window frequency component and energy distribution analysis was performed on the real-time deviation value to obtain information on frequency component and duration changes.

[0083] Based on the frequency components and duration change information, the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process are adjusted to obtain the adjusted decomposition process parameters.

[0084] Based on the individual operating characteristics of the battery cells and the historical background noise level, the thresholds for distinguishing fault signals from background noise are calculated and updated in real time, and the thresholds for distinguishing fault signals from normal aging deviation components are calculated and updated in real time, resulting in updated reference thresholds.

[0085] Based on the adjusted decomposition process parameters and updated reference thresholds, the real-time deviation value is decomposed to distinguish between slowly changing, persistent deviation components and transient, intermittent deviation components.

[0086] Specifically, obtaining the real-time deviation value of a battery cell refers to acquiring deviation data by comparing the actual operating parameters of the battery cell with the predicted normal behavior parameters. This involves performing multi-time-window frequency component and energy distribution analysis on the real-time deviation value to obtain information on frequency component and duration changes. This can be understood as using various time windows of different lengths to perform time-frequency analysis on the real-time deviation value, such as Fourier transform, wavelet transform, or empirical mode decomposition, to capture the frequency characteristics and energy distribution of the deviation signal at different time scales. For example, short time windows can be used to capture transient, high-frequency deviations, while long time windows are suitable for capturing slowly changing, low-frequency deviations, with the aim of comprehensively acquiring the dynamic characteristics of the deviation signal.

[0087] The decomposition process involves adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components based on information about frequency and duration variations. Specifically, if the analysis shows that the deviation signal mainly contains high-frequency, short-duration components, the time window length can be shortened and the frequency cutoff point increased to more accurately separate transient components. Conversely, if the deviation signal mainly contains low-frequency, long-duration components, the time window length can be extended and the frequency cutoff point lowered to better capture persistent components. The aim is to enable the decomposition process to adaptively optimize to suit the actual characteristics of the deviation signal.

[0088] In practical applications, based on the individual operating characteristic records of battery cells and historical background noise levels, thresholds for distinguishing fault signals from background noise and for distinguishing fault signals from normal aging deviations are calculated and updated in real time, resulting in updated reference thresholds. The individual operating characteristic records contain the unique behavioral patterns and parameter fluctuation ranges of each battery cell in a healthy state, while historical background noise levels provide statistical characteristics of environmental noise. By combining this information, thresholds for distinguishing fault signals from background noise and fault signals from normal aging deviations can be dynamically set. For example, when the background noise level increases, the threshold for distinguishing fault signals can be increased accordingly to avoid false alarms; when a battery cell exhibits a specific aging trend, the threshold for distinguishing normal aging deviations can be adjusted to avoid misjudging normal aging as a fault. The aim is to improve the accuracy and robustness of detection.

[0089] Optionally, the step of matching the deviation components according to the preset internal short-circuit fault feature template to obtain the matching result includes:

[0090] Obtain the real-time voltage deviation and real-time internal resistance deviation of the battery cell;

[0091] Time-series characteristic analysis is performed on real-time voltage deviation and real-time internal resistance deviation to obtain the corresponding evolution trend;

[0092] Based on the evolution trend, the decrease magnitude, duration, recovery rate of voltage deviation, and the magnitude and duration of internal resistance increase in the preset internal short-circuit fault characteristic template are adjusted to obtain the adjusted internal short-circuit fault characteristic template.

[0093] Based on the adjusted internal short-circuit fault feature template, the similarity of the deviation components is calculated to obtain the degree of matching between the deviation components and the adjusted internal short-circuit fault feature template.

[0094] The type of abnormality of the battery cell is determined based on the degree of matching.

[0095] Specifically, real-time voltage deviation refers to the difference between the actual voltage of a battery cell and the predicted normal voltage, while real-time internal resistance deviation refers to the difference between the actual internal resistance of a battery cell and the predicted normal internal resistance. These parameters directly reflect the occurrence and development of internal short-circuit faults in the battery. Time-series characteristic analysis of real-time voltage and internal resistance deviations aims to capture the patterns of these deviation signals over time, such as the rate of voltage drop, duration, and recovery characteristics, as well as the magnitude and duration of internal resistance increases. These evolution trends are the unique fingerprints of internal short-circuit faults. The preset internal short-circuit fault characteristic template can be understood as a set of parameters describing typical short-circuit fault characteristics, such as the typical range of voltage drop magnitude, duration, and recovery rate, and the typical range of internal resistance increases. Based on the time-series evolution trends of real-time voltage and internal resistance deviations, the preset template can be dynamically adjusted. For example, if a small voltage drop magnitude but a long duration is observed, the parameter ranges for voltage drop magnitude and duration in the template can be adjusted accordingly to better match the currently observed fault characteristics. This adjustment makes the template more adaptable, thereby improving the accuracy of matching. In practical applications, after obtaining the adjusted internal short-circuit fault feature template, the deviation components of the battery cell (e.g., slowly changing persistent deviation components or transient intermittent deviation components obtained through decomposition) are compared with the adjusted template for similarity calculation. Various methods can be used for similarity calculation, such as Euclidean distance, cosine similarity, or dynamic time warping (DTW), to quantify the degree of agreement between the deviation components and the template features. Based on the calculated matching degree, the anomaly type of the battery cell can be determined, such as whether it is a minor short circuit, a moderate short circuit, or a severe short circuit, or it can be distinguished from other non-short-circuit anomalies.

[0096] Optionally, the steps of performing multi-time-window frequency component and energy distribution analysis on the real-time deviation values ​​to obtain information on frequency component and duration changes include:

[0097] Read the real-time deviation value of the corresponding battery cell;

[0098] Acquire vibration signals, electromagnetic environment signals, and sensor status signals during vehicle operation;

[0099] Multi-time-window frequency component and energy distribution analysis was performed on the real-time deviation value to obtain information on the frequency component and duration changes of the multi-time-window.

[0100] Frequency components and energy distribution of vibration signals, electromagnetic environment signals, and sensor status signals are analyzed to obtain the frequency components and energy distribution characteristics of external noise.

[0101] Based on the frequency components and energy distribution characteristics of external noise, adjust the frequency response characteristics and energy threshold when performing frequency component and energy distribution analysis in multiple time windows;

[0102] Based on the adjusted frequency response characteristics and energy threshold, the true frequency components and energy distribution caused by the internal short-circuit fault are separated from the frequency components and duration variation information of multiple time windows.

[0103] Specifically, reading the real-time deviation value of the corresponding battery cell refers to obtaining pre-processed data from the battery management system or data acquisition module that reflects the difference between the actual operating state and the predicted normal state of the battery cell. These real-time deviation values ​​may contain weak signals caused by internal short-circuit faults, or they may be mixed with various external interferences. Acquiring vibration signals, electromagnetic environment signals, and sensor status signals during vehicle operation aims to comprehensively capture external noise sources that may affect the real-time deviation value of the battery cell. Vibration signals can be acquired through accelerometers, reflecting the impact of mechanical vibrations such as uneven road surfaces and engine vibrations on the battery during vehicle operation; electromagnetic environment signals can be acquired through electromagnetic sensors or spectrum analyzers, reflecting interference from internal electronic devices and external electromagnetic fields on battery measurement signals; sensor status signals are used to monitor the health status of the battery cell's own sensors, such as whether there is drift or malfunction, to eliminate noise introduced by the sensors themselves. In practical applications, frequency component and energy distribution analysis of real-time deviation values ​​across multiple time windows is performed to obtain information on the frequency components and duration changes within each time window. Signal processing techniques such as Fourier transform and wavelet transform can be employed to analyze real-time deviation values ​​at different time scales and frequency resolutions, revealing their inherent frequency structure and energy distribution characteristics. Furthermore, frequency component and energy distribution analysis is performed on vibration signals, electromagnetic environment signals, and sensor status signals to obtain the frequency components and energy distribution characteristics of external noise. The aim is to establish a "fingerprint" or "template" for external noise. By independently analyzing these external noise sources, their energy intensity and duration characteristics in different frequency ranges can be accurately identified. Based on this, the frequency response characteristics and energy threshold for multi-time-window frequency component and energy distribution analysis are adjusted according to the frequency components and energy distribution characteristics of external noise. This means that when analyzing real-time deviation values, the parameters of the signal processing algorithm are dynamically adjusted based on the identified external noise characteristics. For example, if it is known that the signal in a certain frequency range is mainly caused by vehicle vibration, the frequency response characteristics of that frequency range can be reduced or the energy threshold increased to suppress the influence of vibration noise. Therefore, based on the adjusted frequency response characteristics and energy threshold, the true frequency components and energy distribution caused by internal short-circuit faults are separated from the frequency component and duration variation information across multiple time windows. This method effectively filters out external noise interference, allowing subsequent fault identification to focus more intently on the true characteristics of internal short-circuit faults within the battery.

[0104] Optionally, the steps of adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process based on the frequency component and duration variation information to obtain the adjusted decomposition process parameters include:

[0105] Obtain vehicle operating status information;

[0106] Based on the operational status information, determine whether the vehicle is currently in a specific driving scenario;

[0107] When it is determined that the vehicle is in a specific driving scenario, the frequency components and energy distribution characteristics of transient or continuous noise corresponding to the specific driving scenario are obtained from the preset scenario noise feature library.

[0108] The frequency components and duration variation information are compared with the frequency components and energy distribution characteristics of the corresponding scene noise to identify the part of the deviation signal whose similarity to the corresponding scene noise exceeds a preset threshold.

[0109] When adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process, the parts with noise similarity to the corresponding scene exceeding a preset threshold are suppressed or corrected to obtain the adjusted decomposition process parameters.

[0110] Specifically, acquiring vehicle operating status information can include, but is not limited to, vehicle speed, acceleration, braking status, steering angle, road condition information (such as the degree of bumps and gradient obtained through onboard sensors or navigation systems), engine speed, and motor torque. This information can be acquired in real time through interfaces such as the vehicle's Controller Area Network (CAN bus) or On-Board Diagnostics (OBD) system.

[0111] Determining whether a vehicle is currently in a specific driving scenario based on its operational status information involves analyzing this information to identify the specific operating conditions the vehicle is experiencing, such as rapid acceleration, rapid deceleration, high-speed cruising, low-speed driving, driving on bumpy roads, and turning. For example, when the vehicle's acceleration or deceleration exceeds a preset threshold, it can be identified as a rapid acceleration or rapid deceleration scenario; when the vehicle's speed changes drastically within a short period of time and is accompanied by abnormal vibration sensor signals, it can be identified as a bumpy road scenario.

[0112] In practical applications, a pre-defined scenario noise feature library can be understood as a data set storing the frequency components, energy distribution, duration, and other characteristics of transient or continuous noise caused by non-battery fault factors (such as mechanical vibration, electromagnetic interference, road impact, etc.) under different specific driving scenarios. This feature library can be established by conducting extensive testing and data collection on fault-free battery systems under different driving scenarios, and then processing the collected noise signals through spectrum analysis, energy analysis, and other methods.

[0113] Furthermore, the frequency component and duration variation information is compared with the frequency component and energy distribution characteristics of the corresponding scene noise to quantify the similarity between the deviation signal and the known scene noise. This can be achieved by calculating the cross-correlation coefficient, Euclidean distance, cosine similarity, or by using machine learning algorithms (such as support vector machines and neural networks) for pattern matching. The portion of the similarity exceeding a preset threshold is considered a signal component affected by specific scene noise. The preset threshold can be adjusted according to the actual application scenario and the requirements for false alarm and false negative rates.

[0114] Based on this, when adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process, parts with similarity to the corresponding scene noise exceeding a preset threshold are suppressed or corrected. Suppression can be achieved by using digital filtering techniques (such as notch filters and band-stop filters) to remove or significantly attenuate these noise components from the off-target signal; correction can be achieved by compensatory adjustments to the decomposition parameters based on the characteristics of the noise. For example, if noise within a specific frequency range is identified, the signal within that frequency range is treated as background noise rather than a fault signal during decomposition, thus avoiding misjudging it as a transient or persistent component of the fault. This results in more accurate adjusted decomposition process parameters.

[0115] Optionally, based on the individual operating characteristics records of the battery cells and historical background noise levels, the steps of calculating and updating in real time the thresholds used to distinguish fault signals from background noise, and calculating and updating in real time the thresholds used to distinguish fault signals from normal aging deviation components, to obtain the updated reference thresholds include:

[0116] Continuous monitoring of the frequency components and energy distribution of the real-time deviation value is performed to obtain information on the changes in the frequency components, energy distribution, and duration of the deviation signal.

[0117] Based on the individual operating characteristics records of the corresponding battery cells, the baseline deviation distribution of the battery cells under fault-free conditions is obtained, and the frequency range and energy level of the normal aging deviation components are determined based on the baseline deviation distribution.

[0118] Based on historical background noise levels, obtain the corresponding frequency range and energy level of the background noise;

[0119] Based on the information on the frequency components, energy distribution and duration of the deviation signal, the frequency range and energy level of the normal aging deviation components, and the frequency range and energy level of the corresponding background noise, the overlapping area and degree of overlap between the initial small short-circuit fault signal and the background noise in terms of frequency, amplitude or duration are identified.

[0120] Based on the overlapping area and degree of overlap, the threshold used to distinguish fault signals from background noise is adjusted, and the threshold used to distinguish fault signals from normal aging deviation components is adjusted to obtain the updated reference threshold.

[0121] Based on the updated reference threshold, distinguish fault signals from background noise, or distinguish fault signals from normal aging deviation components.

[0122] Specifically, continuous monitoring of the frequency components and energy distribution of real-time deviation values ​​refers to the continuous analysis of the real-time deviation values ​​of battery cells using Fourier transform, wavelet analysis, or other time-frequency analysis methods. This captures the energy distribution and dynamic characteristics over time within different frequency ranges, thereby obtaining information on the changes in the frequency components, energy distribution, and duration of the deviation signal. The aim is to comprehensively understand the intrinsic structure and evolution of the deviation signal.

[0123] Specifically, by recording the individual operational characteristics of each battery cell, the baseline deviation distribution of the battery cell under fault-free conditions is obtained. This can be understood as using long-term accumulated data collected during the normal operation of the battery cell to construct a deviation behavior model unique to that battery cell, reflecting its health status and normal aging process. Based on this baseline deviation distribution, the typical frequency range and energy level of normal aging deviation components can be determined. For example, aging typically exhibits a slowly changing trend of deviation, with its frequency components concentrated in the low-frequency region and its energy level gradually increasing with the degree of aging. The purpose is to provide a personalized reference benchmark for distinguishing fault signals from normal aging deviation components.

[0124] In practical applications, historical background noise levels specifically refer to the deviation signal characteristics caused by non-fault factors such as inherent system noise, sensor noise, and external environmental interference, obtained through long-term monitoring and data statistics under different operating conditions and environmental environments. By analyzing this historical data, the frequency range and energy level of the corresponding background noise can be obtained. For example, vehicle vibration may generate high-energy noise at specific frequencies, and electromagnetic interference may manifest as transient high-frequency spikes. The purpose is to provide a reliable reference for distinguishing fault signals from background noise.

[0125] Furthermore, identifying the overlapping regions and degrees of overlap between the initial minor short-circuit fault signal and background noise, as well as with normal aging deviation components, in terms of frequency, amplitude, or duration, involves using cross-correlation analysis, pattern recognition algorithms, or machine learning models to meticulously compare the changes in the frequency components, energy distribution, and duration of the deviation signal with the frequency range and energy level of the normal aging deviation components, and the frequency range and energy level of the background noise. For example, a minor short-circuit fault might produce a weak but sustained energy increase in a specific frequency band, while background noise might exhibit intermittent or random fluctuations in that frequency band. By quantifying these overlapping regions and degrees of overlap, the source of the signal can be assessed more accurately. The purpose is to provide a quantitative basis for subsequent threshold adjustments.

[0126] Therefore, based on the overlapping area and degree of overlap, the thresholds used to distinguish fault signals from background noise and to distinguish fault signals from normal aging deviation components are adjusted to obtain updated reference thresholds. For example, if an initial minor short-circuit fault signal and background noise are found to have a high degree of overlap in a certain frequency band, the noise threshold for that frequency band can be dynamically increased to avoid false alarms; conversely, if the overlap is low, the threshold can be appropriately decreased to improve the sensitivity to early faults. For overlap with normal aging deviation components, a dynamic fault identification threshold that varies with the degree of aging can be set based on the aging model, according to the degree of overlap. The purpose is to ensure the adaptability and accuracy of the thresholds.

[0127] Finally, based on the updated reference threshold, the real-time deviation value is judged to distinguish fault signals from background noise, or fault signals from normal aging deviation components. This means that only when the characteristics of the deviation signal (such as frequency, amplitude, and duration) exceed the finely adjusted threshold that takes into account the effects of background noise and normal aging will it be judged as a potential fault signal, thereby achieving more accurate fault identification.

[0128] Optionally, the steps of continuously monitoring the frequency components and energy distribution of the real-time deviation value to obtain information on changes in the frequency components, energy distribution, and duration of the deviation signal include:

[0129] The real-time deviation value is decomposed into several time scales and frequency ranges to obtain the frequency components and energy distribution at different scales;

[0130] Based on the frequency components and energy distribution at each scale, the nonlinear trend and irregular fluctuation characteristics that change with time are analyzed.

[0131] Based on the material properties and electrolyte state change patterns of the corresponding battery cells, a reference signal characteristic caused by non-short-circuit factors is established.

[0132] By comparing the nonlinear trends and irregular fluctuation characteristics with the signal characteristics caused by non-short-circuit factors, the unique micro-signal characteristics of short-circuit faults can be identified.

[0133] Based on the identified microscopic signal characteristics, information on the changes in frequency components, energy distribution, and duration of the corresponding deviation signal is obtained.

[0134] Specifically, decomposing real-time deviation values ​​into several time scales and frequency ranges can be achieved using multi-scale analysis methods, such as wavelet transform, empirical mode decomposition (EMD), or variational mode decomposition (VMD). These methods decompose the original real-time deviation signal into components with different time and frequency resolutions, thereby capturing transient and persistent features at different scales in the signal. For example, rapidly changing transient signals can be identified at smaller time scales, while slowly changing trend signals can be observed at larger time scales.

[0135] Analyzing the nonlinear trends and irregular fluctuations over time based on the frequency components and energy distribution at each scale involves using nonlinear dynamics theory and statistical analysis methods to conduct in-depth analysis of the decomposed signal components. For example, nonlinear characteristic parameters such as the Lyapunov exponent, fractal dimension, and entropy can be calculated to reveal potential chaotic behavior, self-similarity, or complexity within the signal. These nonlinear characteristics often more sensitively reflect the weak and complex signal changes caused by initial short-circuit faults, changes that might be ignored in traditional linear analysis.

[0136] In practical applications, establishing a reference model for signal characteristics caused by non-short-circuit factors, based on the material properties and electrolyte state changes of the corresponding battery cells, involves learning and modeling historical data from a large number of fault-free battery cells under different operating conditions (such as temperature, charge-discharge cycles, and aging levels). This results in a database or model capable of characterizing signal deviations caused by non-short-circuit factors such as normal aging, ambient temperature fluctuations, and electrochemical reactions during charge-discharge processes. This reference model can include typical frequency components, energy distributions, nonlinear trends, and fluctuation characteristics of these non-short-circuit factors across different time scales and frequency ranges.

[0137] Furthermore, comparing the nonlinear trends and irregular fluctuation characteristics with the signal feature reference caused by non-short-circuit factors to identify the microscopic signal characteristics unique to short-circuit faults involves comparing the nonlinear characteristics of the currently monitored real-time deviation signal with the established signal feature reference caused by non-short-circuit factors. Through techniques such as similarity calculation, pattern recognition, or machine learning classification, normal fluctuations caused by non-short-circuit factors can be effectively distinguished from abnormal microscopic signal characteristics caused by initial short-circuit faults. For example, if the fractal dimension or Lyapunov exponent of the real-time deviation signal significantly deviates from the range predicted by the non-short-circuit factor reference model within a specific frequency range, and this deviation pattern matches known short-circuit fault characteristics, it can be identified as a microscopic signal characteristic unique to short-circuit faults.

[0138] Therefore, by acquiring information on the frequency components, energy distribution, and duration of the corresponding deviation signal based on the identified microscopic signal characteristics, it is possible to accurately extract the frequency components, energy distribution, and their duration of change over time once the unique microscopic signal characteristics of a short-circuit fault are identified. This information will serve as more precise input for distinguishing the fault signal from background noise or normal aging deviation components, thereby improving the accuracy and sensitivity of fault detection.

[0139] This application also proposes an intelligent detection system for lithium battery short-circuit faults, used to perform intelligent detection of lithium battery short-circuit faults, combined with... Figure 3 As shown, the intelligent detection system 1 for lithium battery short circuit faults includes:

[0140] Battery parameter acquisition module 11 is used to acquire the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery;

[0141] The feature record update module 12 is used to establish and continuously update the corresponding personalized operating feature record for each battery cell based on operating parameters and environmental parameters.

[0142] The normal parameter prediction module 13 is used to predict the normal behavior parameters of each battery cell under the current operating conditions based on the personalized operating characteristics record, operating parameters and environmental parameters.

[0143] The abnormal pattern recognition module 14 is used to compare the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and to perform cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell.

[0144] The early warning execution module 15 is used to trigger graded early warnings based on the abnormal deviation pattern and execute the corresponding measures for the graded early warning.

[0145] To better understand the technical solutions proposed in this application, some key terms involved will be explained below.

[0146] "Operating parameters" refer to various electrical and thermal data monitored in real time during the operation of a lithium battery, such as the battery cell's voltage, current, temperature, state of charge / discharge, state of charge (SOC), and state of health (SOH). These parameters directly reflect the battery cell's immediate operating status.

[0147] "Environmental parameters" refer to the external environmental conditions in which the battery cell operates, such as ambient temperature, humidity, altitude, vehicle speed, acceleration, vibration intensity, and external electromagnetic interference. These parameters affect the performance and aging process of the battery cell.

[0148] "Personalized operating characteristic records" refer to a set of data established and continuously updated for each individual battery cell, based on its historical operating data and environmental data, that reflects the unique performance degradation patterns and behavioral characteristics of that battery cell. This includes, but is not limited to, its internal resistance variation trend, capacity decay curve, self-discharge rate, and charge / discharge efficiency at different temperatures.

[0149] "Normal behavior parameters" refer to the expected operating parameters of a healthy, fault-free battery cell under current operating and environmental conditions. These parameters are derived from the battery cell's individual operating characteristics and real-time operating condition predictions.

[0150] "Real-time deviation" refers to the difference between the actual operating parameters of a battery cell and the predicted normal behavior parameters. This value reflects the degree to which the current behavior of the battery cell deviates from the expected normal behavior.

[0151] "Abnormal deviation patterns" refer to deviation patterns with specific time series characteristics, amplitude characteristics, or frequency characteristics identified by cumulative analysis of real-time deviation values. These patterns correspond to known types of internal short-circuit faults.

[0152] "Graded early warning" refers to classifying early warning information into different levels, such as mild warning, moderate warning, and severe warning, based on the severity, development trend, and potential risks of abnormal deviation patterns.

[0153] "Measures" refer to the response strategies that the system automatically or recommends for different levels of warnings, such as reducing charging and discharging power, limiting vehicle performance, prompting the driver to check, or even automatically disconnecting the faulty battery cell.

[0154] The intelligent detection system for short-circuit faults in lithium batteries proposed in this application may include the following modules and their functions in a specific implementation:

[0155] The battery parameter acquisition module is used to acquire the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery.

[0156] Specifically, this module can consist of a series of sensors, data acquisition units, and communication interfaces. For example, voltage, current, and temperature sensors can be directly deployed on each battery cell to collect its electrical and thermal parameters in real time. The data acquisition unit is responsible for digitizing and initially processing these analog signals. Communication interfaces, such as CAN bus interfaces or Ethernet interfaces, are used to transmit the collected data to the central processing unit or battery management system. Environmental parameters can be acquired through independent external environmental sensors (such as ambient temperature and humidity sensors) or by obtaining vehicle operating status information (such as vehicle speed and acceleration) from the vehicle bus system.

[0157] The feature record update module is used to establish and continuously update the corresponding personalized operating feature record for each battery cell based on operating parameters and environmental parameters.

[0158] This module can be a software module running within a battery management system (BMS) or a standalone computing unit. Its implementation can include: using statistical analysis algorithms based on historical data to model key parameters such as internal resistance and capacity degradation for each battery cell; or utilizing machine learning algorithms, such as support vector machines (SVM) or neural networks, to learn and predict the performance of battery cells under different operating conditions. By periodically receiving new operating and environmental parameters, this module can dynamically adjust and optimize its internal model, ensuring the accuracy and timeliness of personalized operating characteristic records.

[0159] The normal parameter prediction module is used to predict the normal behavior parameters of each battery cell under the current operating conditions based on personalized operating characteristic records, operating parameters, and environmental parameters.

[0160] This module can also be implemented as a software module, with its core being a prediction algorithm. For example, a method combining physical models and data-driven models can be used. By utilizing the personalized operating characteristic records provided by the characteristic record update module, and combining real-time operating parameters such as current, temperature, and state of charge, as well as environmental parameters, the voltage, internal resistance, or temperature parameters that the battery cell should have under the current fault-free state can be calculated through Gaussian process regression, Kalman filtering, or deep learning prediction models.

[0161] The abnormal pattern recognition module is used to compare the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and to perform cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell.

[0162] This module can consist of a data processing unit and a pattern recognition algorithm. The data processing unit is responsible for receiving actual operating parameters and predicted normal behavior parameters, and calculating the real-time deviation between the two. The pattern recognition algorithm can employ statistical analysis methods, such as the Cumulative Sum (CUSUM) algorithm or the Exponential Weighted Moving Average (EWMA) algorithm, to perform cumulative analysis on the real-time deviation values ​​to filter out instantaneous noise and highlight persistent deviation trends. Furthermore, it can be combined with feature extraction and classification algorithms, such as Support Vector Machines (SVM) or decision trees, to match the cumulative deviation values ​​with preset internal short-circuit fault feature templates, thereby identifying specific abnormal deviation patterns.

[0163] The early warning execution module is used to trigger tiered early warnings based on abnormal deviation patterns and execute measures corresponding to the tiered early warnings.

[0164] This module can consist of a decision logic unit, a human-machine interface, and an actuator control interface. The decision logic unit triggers different levels of warnings based on the severity and trend of the abnormal deviation patterns output by the abnormal pattern recognition module, combined with preset warning strategies. The human-machine interface can include a display screen, indicator lights, or a buzzer to convey warning information to the driver. The actuator control interface can send commands to the vehicle's power management system, such as limiting charging and discharging power, disconnecting faulty battery cells, or initiating emergency braking procedures to execute corresponding safety measures.

[0165] The intelligent detection system for short-circuit faults in lithium batteries proposed in this application works by constructing a dynamic and adaptive battery health monitoring and fault diagnosis framework, which overcomes the problem of insufficient detection accuracy of traditional methods under complex working conditions and battery aging backgrounds.

[0166] Specifically, the system first continuously acquires the operating and environmental parameters of each battery cell through a battery parameter acquisition module, laying the data foundation for subsequent personalized analysis. These parameters include not only the battery's electrical and thermal states but also the impact of the external environment on battery performance.

[0167] Next, based on this rich operational and environmental data, the characteristic recording and updating module establishes and continuously updates a personalized operational characteristic record for each battery cell. This step is one of the core innovations of this application. Traditional methods often use a uniform battery model or fixed thresholds, ignoring the differences between individual batteries and the non-uniformity of the aging process. In contrast, this application uses machine learning and other technologies to build a unique "health profile" for each battery cell, recording the changes in key parameters such as internal resistance, capacity, and self-discharge over time, temperature, and cycle count. This personalized recording enables the system to accurately grasp the "normal" behavioral baseline of each battery cell, distinguishing between normal fluctuations and abnormal deviations even under aging or complex operating conditions.

[0168] Based on this, the normal parameter prediction module predicts the normal behavior parameters of each battery cell under current operating conditions, according to personalized operating characteristic records, current operating parameters, and environmental parameters. This means that the system no longer relies on static, preset thresholds, but can dynamically predict the ideal performance of the battery under specific conditions. For example, during high-speed driving or rapid acceleration, battery voltage fluctuations will increase, but the system can predict the specific value of such fluctuations within the normal range based on personalized records, thereby avoiding misjudging normal voltage drops as faults.

[0169] Subsequently, the anomaly pattern recognition module compares the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. This deviation value represents the difference between the actual performance of the battery and the personalized predicted performance, and can more sensitively reflect minor anomalies within the battery. More importantly, this module performs cumulative analysis on the real-time deviation value to identify the abnormal deviation patterns of the battery cell. Cumulative analysis can effectively filter out instantaneous noise and random fluctuations, highlighting persistent, weak signals related to short-circuit faults. Through pattern recognition technology, the system can match these cumulative deviation values ​​with preset internal short-circuit fault feature templates, thereby identifying different types of abnormal deviation patterns, such as slowly developing micro-short circuits and intermittent short circuits.

[0170] Finally, the early warning execution module triggers tiered warnings based on the identified abnormal deviation patterns and executes corresponding measures. This tiered warning mechanism provides differentiated response strategies based on the severity, development trend, and potential risks of the fault. For example, for a minor initial short-circuit risk, the system may only issue a suggestive warning and recommend checking at the next maintenance; while for a serious short-circuit risk, it may immediately take emergency measures such as limiting power or even cutting off the power supply. This intelligent early warning and action execution maximizes driving safety and provides drivers and maintenance personnel with timely and accurate decision-making information.

[0171] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent detection of short-circuit faults in lithium batteries, characterized in that, include: Obtain the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery; Based on the operating parameters and the environmental parameters, a corresponding personalized operating characteristic record is established and continuously updated for each battery cell; Based on the personalized operating characteristic records, the operating parameters, and the environmental parameters, predict the normal behavior parameters of each battery cell under the current operating conditions; The actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. The real-time deviation value is then accumulated and analyzed to identify the abnormal deviation pattern of the battery cell. Based on the abnormal deviation pattern, a tiered warning is triggered, and measures corresponding to the tiered warning are executed. The steps of comparing the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and performing cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell include: The actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. The real-time deviation value is decomposed to obtain several deviation components; The deviation components are matched according to the preset internal short-circuit fault feature template to obtain the matching result; Based on the matching results, the abnormality type of the battery cell is determined to identify the abnormal deviation pattern of the battery cell. The step of decomposing the real-time deviation value to obtain several deviation components includes: Obtain the real-time deviation value of the battery cell; The frequency components and energy distribution of the real-time deviation value are analyzed over multiple time windows to obtain information on the changes in frequency components and duration. Based on the frequency component and duration change information, the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process are adjusted to obtain the adjusted decomposition process parameters. Based on the individual operating characteristics of the battery cells and the historical background noise level, the thresholds for distinguishing fault signals from background noise are calculated and updated in real time, and the thresholds for distinguishing fault signals from normal aging deviation components are calculated and updated in real time, resulting in updated reference thresholds. Based on the adjusted decomposition process parameters and the updated reference thresholds, the real-time deviation value is decomposed to distinguish between slowly changing, persistent deviation components and transient, intermittent deviation components.

2. The intelligent detection method for short-circuit faults in lithium batteries according to claim 1, characterized in that, The steps of triggering a graded early warning based on the abnormal deviation pattern and executing measures corresponding to the graded early warning include: Obtain vehicle operating status information and road environment information; Based on the operating status information and the road environment information, determine the driver's driving intention; Obtain pre-defined tiered early warnings and corresponding measures; Based on the abnormal deviation pattern, assess the potential impact of the measures corresponding to the graded early warning on vehicle driving safety and driving experience; Based on the severity of the abnormal deviation pattern, the risk development trend, the vehicle operating environment, the driving intention, and the potential impact, the execution strategy and timing of the corresponding measures are adjusted to obtain the adjusted measures information; The human-computer interaction interface conveys information about the adjusted measures, the reasons for the adjustments, and suggested driving behaviors.

3. The intelligent detection method for short-circuit faults in lithium batteries according to claim 1, characterized in that, The step of matching the deviation components according to the preset internal short-circuit fault feature template to obtain the matching result includes: Obtain the real-time voltage deviation and real-time internal resistance deviation of the battery cell; Time-series characteristic analysis is performed on the real-time voltage deviation and the real-time internal resistance deviation to obtain the corresponding evolution trend; Based on the evolution trend, the decrease magnitude, duration, recovery rate of voltage deviation, and the magnitude and duration of internal resistance increase in the preset internal short-circuit fault feature template are adjusted to obtain the adjusted internal short-circuit fault feature template. Based on the adjusted internal short-circuit fault feature template, the similarity of the deviation components is calculated to obtain the degree of matching between the deviation components and the adjusted internal short-circuit fault feature template. Based on the degree of matching, the abnormality type of the battery cell is determined.

4. The intelligent detection method for short-circuit faults in lithium batteries according to claim 1, characterized in that, The step of performing multi-time-window frequency component and energy distribution analysis on the real-time deviation value to obtain frequency component and duration change information includes: Read the real-time deviation value of the corresponding battery cell; Acquire vibration signals, electromagnetic environment signals, and sensor status signals during vehicle operation; The frequency components and energy distribution of the real-time deviation value are analyzed over multiple time windows to obtain information on the frequency components and duration changes over multiple time windows. Frequency component and energy distribution analysis is performed on the vibration signal, the electromagnetic environment signal, and the sensor status signal to obtain the frequency component and energy distribution characteristics of the external noise. Based on the frequency components and energy distribution characteristics of external noise, adjust the frequency response characteristics and energy threshold when performing frequency component and energy distribution analysis in multiple time windows; Based on the adjusted frequency response characteristics and energy threshold, the true frequency components and energy distribution caused by the internal short-circuit fault are separated from the frequency components and duration variation information of multiple time windows.

5. The intelligent detection method for short-circuit faults in lithium batteries according to claim 1, characterized in that, The step of adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process based on the frequency component and duration change information, to obtain the adjusted decomposition process parameters, includes: Obtain vehicle operating status information; Based on the operating status information, determine whether the vehicle is currently in a specific driving scenario; When it is determined that the vehicle is in a specific driving scenario, the frequency components and energy distribution characteristics of transient or continuous noise corresponding to the specific driving scenario are obtained from a preset scenario noise feature library. The frequency components and duration change information are compared with the frequency components and energy distribution characteristics of the corresponding scene noise to identify the part of the deviation signal whose similarity to the corresponding scene noise exceeds a preset threshold. When adjusting the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process, the parts with noise similarity to the corresponding scene exceeding a preset threshold are suppressed or corrected to obtain the adjusted decomposition process parameters.

6. The intelligent detection method for short-circuit faults in lithium batteries according to claim 1, characterized in that, The steps of calculating and updating the thresholds for distinguishing fault signals from background noise in real time based on the individual operating characteristics of the battery cells and historical background noise levels, and calculating and updating the thresholds for distinguishing fault signals from normal aging deviation components in real time, to obtain the updated reference thresholds, include: The frequency components and energy distribution of the real-time deviation value are continuously monitored to obtain information on the changes in the frequency components, energy distribution, and duration of the deviation signal. Based on the individual operating characteristics record of the corresponding battery cell, the baseline deviation distribution of the battery cell under fault-free condition is obtained, and based on the baseline deviation distribution, the frequency range and energy level of the normal aging deviation component are determined. Based on historical background noise levels, obtain the corresponding frequency range and energy level of the background noise; Based on the information on the frequency components, energy distribution and duration of the deviation signal, the frequency range and energy level of the normal aging deviation components, and the frequency range and energy level of the corresponding background noise, the overlapping area and degree of overlap between the initial small short-circuit fault signal and the background noise in terms of frequency, amplitude or duration are identified. Based on the overlapping region and the degree of overlap, the threshold used to distinguish fault signals from background noise is adjusted, and the threshold used to distinguish fault signals from normal aging deviation components is adjusted to obtain an updated reference threshold. Based on the updated reference threshold, distinguish fault signals from background noise, or distinguish fault signals from normal aging deviation components.

7. The intelligent detection method for short-circuit faults in lithium batteries according to claim 6, characterized in that, The step of continuously monitoring the frequency components and energy distribution of the real-time deviation value to obtain information on the changes in the frequency components, energy distribution, and duration of the deviation signal includes: The real-time deviation value is decomposed into several time scales and frequency ranges to obtain the frequency components and energy distribution at different scales; Based on the frequency components and energy distribution at each scale, the nonlinear trend and irregular fluctuation characteristics that change with time are analyzed. Based on the material properties and electrolyte state change patterns of the corresponding battery cells, a reference signal characteristic caused by non-short-circuit factors is established. By comparing the nonlinear trend and irregular fluctuation characteristics with the signal characteristic reference caused by non-short-circuit factors, the unique micro-signal characteristics of short-circuit faults are identified. Based on the identified microscopic signal characteristics, information on the changes in frequency components, energy distribution, and duration of the corresponding deviation signal is obtained.

8. A lithium battery short-circuit fault intelligent detection system, used to perform intelligent detection of lithium battery short-circuit faults, characterized in that, include: The battery parameter acquisition module is used to acquire the operating parameters and environmental parameters corresponding to the battery cells of the lithium battery. The feature record update module is used to establish and continuously update a corresponding personalized operating feature record for each battery cell based on the operating parameters and the environmental parameters. The normal parameter prediction module is used to predict the normal behavior parameters of each battery cell under the current operating conditions based on the personalized operating characteristic record, the operating parameters and the environmental parameters. An abnormal pattern recognition module is used to compare the actual operating parameters of the battery cell with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell, and to perform cumulative analysis on the real-time deviation value to identify the abnormal deviation pattern of the battery cell. The early warning execution module is used to trigger a graded early warning based on the abnormal deviation pattern and execute the measures corresponding to the graded early warning. The abnormal pattern recognition module is also used for: The actual operating parameters of the battery cell are compared with the predicted normal behavior parameters to obtain the real-time deviation value of the battery cell. The real-time deviation value is decomposed to obtain several deviation components; The deviation components are matched according to the preset internal short-circuit fault feature template to obtain the matching result; Based on the matching results, the abnormality type of the battery cell is determined to identify the abnormal deviation pattern of the battery cell. The decomposition of the real-time deviation value yields several deviation components, including: Obtain the real-time deviation value of the battery cell; The frequency components and energy distribution of the real-time deviation value are analyzed over multiple time windows to obtain information on the changes in frequency components and duration. Based on the frequency component and duration change information, the time window length and frequency cutoff point used to distinguish transient and persistent components during the decomposition process are adjusted to obtain the adjusted decomposition process parameters. Based on the individual operating characteristics of the battery cells and the historical background noise level, the thresholds for distinguishing fault signals from background noise are calculated and updated in real time, and the thresholds for distinguishing fault signals from normal aging deviation components are calculated and updated in real time, resulting in updated reference thresholds. Based on the adjusted decomposition process parameters and the updated reference thresholds, the real-time deviation value is decomposed to distinguish between slowly changing, persistent deviation components and transient, intermittent deviation components.

Images