Battery maintenance methods and systems for abnormal use during dynamic operation

Abstract

This invention discloses a method and system for abnormal maintenance of batteries during dynamic use. The invention relates to the technical field of batteries, and determines the abnormal characteristics of the battery based on multiple abnormal data and the corresponding abnormal regions. It also determines the abnormal dynamic curve of the battery during operation, which presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes. Multiple abnormal data are introduced, and each abnormal node and its corresponding associated data triggers the filtering of each abnormal node to eliminate false abnormal nodes and determine multiple final abnormal nodes. Based on multiple abnormal nodes and the current working content of the battery, corresponding abnormal events are determined, improving the accuracy of abnormal events. Simultaneously, a multi-level maintenance system for the battery is constructed based on multiple abnormal maintenance measures and the current working content of the battery, realizing a holistic consideration of multiple abnormal nodes and the current working content of the battery, further improving the accuracy of abnormal events.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and more particularly to a method and system for abnormal maintenance of batteries during dynamic use. Background Technology

[0002] With the rapid development of new energy vehicles and energy storage technologies, the safety and reliability of lithium-ion batteries, as core energy storage units, are receiving increasing attention. To ensure the stable operation of battery packs under complex operating conditions, existing technologies typically employ battery management systems (BMS) to monitor key parameters such as voltage, current, and temperature in real time. When parameters exceed preset thresholds, corresponding protection mechanisms (such as circuit disconnection or fan activation) are triggered. However, existing technologies often focus on single-dimensional data analysis or simple threshold comparisons when monitoring battery status, without controlling data such as voltage, current, temperature, and vibration. This affects the accuracy of abnormal dynamic curves during battery operation and fails to guarantee the accuracy of abnormal battery events and the accuracy of the battery's multi-level maintenance system. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for abnormal maintenance of batteries during dynamic use.

[0004] This invention provides a method for abnormal maintenance of a battery during dynamic use, including:

[0005] Multiple battery data points are aligned over time, and multiple data combinations are identified, each encompassing voltage, current, temperature, and vibration data. Based on the identification of these multiple data combinations, corresponding anomalous data points are determined.

[0006] Based on multiple abnormal data and the abnormal regions corresponding to the battery, the abnormal characteristics of the battery are determined, and the abnormal dynamic curve of the battery during operation is determined. The abnormal dynamic curve shows the abnormal changes of the battery at different times and marks the corresponding abnormal nodes.

[0007] Each abnormal node and its corresponding associated data is used to trigger the filtering of each abnormal node to eliminate false abnormal nodes and identify multiple final abnormal nodes; the corresponding abnormal events are determined based on multiple abnormal nodes and the current working content of the battery.

[0008] Based on the various abnormal items of the abnormal events, their corresponding abnormal priorities, and multiple temperature data of the battery, a multi-factor abnormal area is determined. Based on the mapping relationship between the multi-factor abnormal area and the battery's abnormal maintenance, multiple abnormal maintenance measures are determined. Based on the multiple abnormal maintenance measures and the battery's current working content, a multi-level maintenance system for the battery is constructed.

[0009] This invention provides a battery anomaly maintenance system for dynamic use, which is applied to the aforementioned battery anomaly maintenance method for dynamic use; the battery anomaly maintenance system for dynamic use includes:

[0010] The abnormal data module is used to align multiple data points of the battery over time and identify multiple data combinations, each of which includes voltage data, current data, temperature data, and vibration data; based on the identification of multiple data combinations, corresponding abnormal data are determined.

[0011] The abnormal dynamic curve module is used to determine the abnormal characteristics of the battery based on multiple abnormal data and the abnormal area corresponding to the battery, and to determine the abnormal dynamic curve of the battery during operation. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes.

[0012] The abnormal event module is used to trigger the filtering of each abnormal node based on each abnormal node and its corresponding associated data, so as to eliminate false abnormal nodes and identify multiple final abnormal nodes; and to determine the corresponding abnormal events based on multiple abnormal nodes and the current working content of the battery.

[0013] The multi-level maintenance system module is used to determine multi-factor abnormal areas based on various abnormal items of abnormal events, corresponding abnormal priorities, and multiple temperature data of the battery. Based on the abnormal maintenance mapping relationship between the multi-factor abnormal areas and the battery, multiple abnormal maintenance measures are determined. The multi-level maintenance system of the battery is constructed according to the multiple abnormal maintenance measures and the current working content of the battery.

[0014] Compared with the prior art, the beneficial effects of the present invention are:

[0015] (1) Align multiple data points of the battery in time and determine multiple data combinations, each data combination covering voltage data, current data, temperature data and vibration data; determine multiple abnormal data based on the identification of multiple data combinations; determine the abnormal characteristics of the battery based on multiple abnormal data and the abnormal area corresponding to the battery, and determine the abnormal dynamic curve of the battery during operation. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes. Multiple abnormal data are introduced to control the abnormal characteristics of the battery and improve the accuracy of the abnormal dynamic curve of the battery during operation.

[0016] (2) Based on each abnormal node and its corresponding associated data, the filtering of each abnormal node is triggered to eliminate false abnormal nodes and determine multiple final abnormal nodes; based on the current working content of multiple abnormal nodes and the battery, the corresponding abnormal events are determined, realizing the overall consideration of the current working content of multiple abnormal nodes and the battery, and improving the accuracy of abnormal events.

[0017] (3) Based on the abnormal items of the abnormal events, the corresponding abnormal priorities and multiple temperature data of the battery, the multi-factor abnormal area is determined, and multiple abnormal maintenance measures are determined based on the mapping relationship between the multi-factor abnormal area and the battery's abnormal maintenance. Based on the multiple abnormal maintenance measures and the current working content of the battery, a multi-level maintenance system for the battery is constructed to control the multi-factor abnormal area, realize the overall consideration of multiple abnormal maintenance measures and the current working content of the battery, and improve the accuracy of the multi-level maintenance system for the battery. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the abnormal maintenance method for a battery during dynamic use in an embodiment of the present invention.

[0019] Figure 2 This is a flowchart illustrating step S11 of the abnormal maintenance method for a battery during dynamic use in an embodiment of the present invention.

[0020] Figure 3 This is a flowchart illustrating step S12 of the abnormal maintenance method for a battery during dynamic use in an embodiment of the present invention.

[0021] Figure 4 This is a flowchart illustrating step S13 of the abnormal maintenance method for a battery during dynamic use in an embodiment of the present invention.

[0022] Figure 5 This is a flowchart illustrating step S14 of the abnormal maintenance method for a battery during dynamic use in an embodiment of the present invention.

[0023] Figure 6 This is a schematic diagram of the structural composition of the abnormal maintenance system for the battery during dynamic use in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0025] Please see Figures 1 to 6 An abnormal maintenance method for batteries under dynamic use, applied to battery scenarios; the abnormal maintenance method for batteries under dynamic use includes:

[0026] Step S11: Align multiple data points of the battery in time and determine multiple data combinations, each data combination covering voltage data, current data, temperature data and vibration data; determine multiple corresponding abnormal data based on the identification of multiple data combinations;

[0027] Step S12: Based on multiple abnormal data and the abnormal area corresponding to the battery, determine the abnormal characteristics of the battery and determine the abnormal dynamic curve of the battery during operation. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes.

[0028] Step S13: Based on each abnormal node and its corresponding associated data, trigger the filtering of each abnormal node to eliminate false abnormal nodes and identify multiple final abnormal nodes; determine the corresponding abnormal events based on multiple abnormal nodes and the current working content of the battery.

[0029] Step S14: Based on the various abnormal items of the abnormal event, the corresponding abnormal priority, and multiple temperature data of the battery, determine the multi-factor abnormal area, and determine multiple abnormal maintenance measures based on the abnormal maintenance mapping relationship between the multi-factor abnormal area and the battery. Construct a multi-level maintenance system for the battery based on the multiple abnormal maintenance measures and the current working content of the battery.

[0030] refer to Figure 2 In step S11, the specific steps are as follows:

[0031] S111: The battery is installed in the electric vehicle, the battery management system of the battery is marked, and multiple data of the battery are determined based on the monitoring of the battery management system. The multiple data are aligned on the same time axis, and the coupling relationship of each data is combined to determine the data combination of multiple features. Each data combination covers voltage data, current data, temperature data and vibration data.

[0032] S112: In each data combination, voltage data, current data, temperature data and vibration data are correlated and scanned, and features are extracted along multiple frequency domain dimensions to determine the corresponding feature vectors, which reflect the internal response of the battery; the feature vectors are semantically mapped with the spatial topology and operating conditions of the battery, and multiple abnormal data are identified, which cover electrochemical parameters, thermodynamic parameters and mechanical parameters.

[0033] In the embodiments of this application, the battery is installed in an electric vehicle, the battery management system of the battery is marked, and multiple data of the battery are determined based on the monitoring of the battery management system. The multiple data are aligned on the same time axis, and the coupling relationship of each data is combined to determine a data combination of multiple features. Each data combination covers voltage data, current data, temperature data and vibration data, which is compatible with the overall consideration of the monitoring of the battery management system and ensures the accuracy of multiple data of the battery.

[0034] At this point, the battery is installed in the electric vehicle. Since the sampling frequencies of different sensors in the Battery Management System (BMS) are often different (for example, voltage and current are usually 10Hz-50Hz, while temperature is only 1Hz-5Hz, and vibration acceleration sensors are as high as 1kHz or more), direct fusion will lead to data misalignment. Therefore, a unified high-precision time reference (such as the system clock of the BMS main control chip) should be established. For low-frequency data (such as temperature), linear interpolation or Newton interpolation is used to "upsample" its time series to the same time density as high-frequency data. For high-frequency data (such as vibration), bandpass filtering and downsampling are used to extract the amplitude or root mean square value of specific characteristic frequencies and map them onto the low-frequency control cycle. All data streams are mapped to the same timestamp sequence T={t1,t2,...,tn} to form a spatiotemporally synchronized data matrix.

[0035] The system defines a variety of data combination strategies. For example, the "voltage-current-temperature" combination is used to analyze the coupling of electrochemical and thermal behavior, and the "current-vibration" combination is used to analyze the impact of mechanical shock on the transient response of current. For each time point, the system not only reads the raw values, but also calculates derived features (such as the first derivative of voltage and the rate of change of current). These features are encapsulated into feature vectors, and each data combination represents a specific physical dimension observation perspective.

[0036] Specifically, an electric vehicle equipped with a battery is accelerating rapidly on a bumpy unpaved road; the vehicle frequently experiences severe vertical vibrations and horizontal impacts, while the battery outputs high power; the BMS's high-voltage sampling unit records the violent fluctuations in current at a frequency of 50Hz, while the NTC thermistor records temperature changes at a frequency of 2Hz, and the vibration sensor captures road impacts at a frequency of 200Hz.

[0037] The system extends the 2Hz temperature data to a 50Hz time axis through cubic spline interpolation, while windowing the 200Hz vibration data to extract the peak vibration acceleration (Peakg) and vibration intensity (VibrationSeverity) within each 20ms window, and then down-maps them to a 50Hz control cycle.

[0038] At timestamp t=10.00s, the system simultaneously obtained: current=150A, voltage=380.5V, temperature (interpolation)=42.3℃, vibration peak=2.5g; and introduced data combination 1 (electrochemical + thermodynamic), data combination 2 (mechanical + electrochemical) and data combination 3 (thermodynamic).

[0039] Data Combination 1 (Electrochemical + Thermodynamic): System analysis of the "voltage-current-temperature" combination at t=10.00s; Under a high current of 150A, the normal battery voltage drop should be 15V (i.e., the voltage drops to 385V); however, the actual monitored voltage was only 380.5V, with a voltage drop as high as 19.5V; combined with the fact that the temperature at this time (42.3℃) is not in the low temperature range, the system identified through the coupling relationship that the voltage drop exceeded the normal threshold composed of ohmic internal resistance and polarization internal resistance; and initially extracted the abnormal electrochemical parameters (abnormally increased internal resistance).

[0040] Data Combination 2 (Mechanical Engineering + Electrochemistry): The system analyzes the "vibration-current" combination at the same moment; at this time, the vibration peak reaches 2.5g, and the frequency is mainly concentrated in the low frequency range (corresponding to road bumps); the system further analyzes the high frequency ripple of the current and finds that the current signal is superimposed with a weak ripple with the same frequency as the vibration; through correlation scanning, the system identifies that mechanical vibration causes a slight change in contact resistance in the physical connection components inside the battery (such as the tabs or busbars). This change in contact resistance that fluctuates with vibration is the root cause of the abnormal voltage drop; the system marks it as abnormal mechanical parameters (poor dynamic contact).

[0041] Data combination 3 (thermodynamics): Monitor the rate of change of temperature data; although the current temperature is 42.3℃, the rate of temperature rise is found to be abnormally higher than the standard curve of discharge at the same rate; determine the abnormal thermodynamic parameters (abnormal heating).

[0042] At time t=10.00s, the system successfully constructed a feature fusion data combination based on aligned data and coupling relationships, and identified abnormal data covering three dimensions—electrochemical (increased internal resistance), thermodynamic (rapid temperature rise), and mechanical (vibration-induced contact loosening)—from the simple phenomenon of "low voltage".

[0043] Furthermore, in each data combination, voltage data, current data, temperature data, and vibration data are correlated and scanned, and features are extracted along multiple frequency domain dimensions to determine the corresponding feature vectors, which reflect the internal response of the battery. The feature vectors are semantically mapped with the spatial topology and operating conditions of the battery, and multiple anomalous data are identified. These anomalous data cover electrochemical parameters, thermodynamic parameters, and mechanical parameters.

[0044] At this point, short-time Fourier transform (STFT) or wavelet transform is used to process the voltage, current, and vibration data within the sliding time window; the time-domain signal is converted into a time-varying spectrum, thus obtaining the three-dimensional characteristic distribution of frequency-amplitude-time; electrochemical dimension: electrochemical impedance spectroscopy (EIS) characteristics are calculated to extract the real and imaginary impedances at specific frequencies (reflecting charge transfer impedance and diffusion coefficient); mechanical dimension: the power spectral density (PSD) of the vibration signal is extracted, the main vibration frequency and its harmonic components are identified, and the distribution of vibration energy in different frequency bands is analyzed; simultaneously, whether the voltage ripple has a peak at the main vibration frequency is detected to determine whether mechanical disturbance has triggered an electrical response.

[0045] Using the CAD model of the battery pack or the BMS topology configuration diagram, establish the correspondence between feature vectors and physical locations; for example, locate the feature vector of voltage anomaly to the specific "Nth string Mth cell" or "Xth module bus" through the topology structure; map the operating condition data provided by the vehicle CAN bus (such as accelerator pedal opening, vehicle speed, suspension travel) into semantic tags (such as "rapid acceleration", "high frequency impact", "steady-state cruise").

[0046] Based on the mapping results, the feature vectors are classified into different physical parameter anomalies: electrochemical parameter anomalies, such as SEI film thickening, lithium deposition, and internal resistance mismatch; thermodynamic parameter anomalies, such as local hot spots, blocked heat dissipation channels, and precursors of thermal runaway; and mechanical parameter anomalies, such as tab fatigue fracture, loose module fasteners, and buffer pad failure.

[0047] Specifically, an electric vehicle equipped with a battery accelerates rapidly on an unpaved road (time point t=10.05s); at this time, the vehicle passes through a deep pothole, and the chassis is subjected to a severe impact; the system extracts a time window from t=10.00s to t=10.10s (including the moment of impact); wavelet transform is performed on the vibration data, and abnormally high energy peaks are found at 25Hz and 80Hz; 25Hz corresponds to the natural frequency of the suspension system (road impact), while 80Hz corresponds to the natural frequency of the battery module itself; at the same time, frequency domain analysis is performed on the current and voltage; a weak 80Hz AC ripple component is superimposed on the steady-state current (DC component).

[0048] The system calculates the cross-correlation coefficient between the vibration signal and the voltage ripple and finds that the 80Hz voltage ripple and the 80Hz mechanical vibration are highly synchronized in phase. This is not random electromagnetic interference (EMI), but a physical contact problem caused by mechanical resonance. This synchronized mechanical-electrical characteristic constitutes the feature vector F=[Vripple,ωmech,Coherence] that reflects the internal response of the battery.

[0049] The system queries the battery's BMS topology and finds that the voltage sampling point Vpack_min is located in the third module (Module3) at the bottom rear of the battery pack. Combining this with the feature vector, the system determines that the source of the anomaly is not the entire battery pack, but the specific "Module3" area. Based on the vehicle status data, the current operating condition is marked as "high current output + vertical impact".

[0050] Abnormal mechanical parameters: Based on the 80Hz resonance characteristics and topology positioning, the system determined the abnormal data as "the internal busbar of Module3 has increased dynamic contact resistance (slight loosening) under vibration conditions". This loosening is normal under static conditions, but it will break instantly or have high impedance at a specific resonance frequency.

[0051] Abnormal electrochemical parameters: Due to mechanical loosening, the current path in this area is blocked, resulting in uneven local electrochemical reactions. The system identifies an abnormal electrochemical feature of "surge in local polarization resistance".

[0052] Abnormal thermodynamic parameters: Although the current temperature rise is not obvious, according to the Joule heating model generated by high-resistivity contact, the system predicts that "point-like heat accumulation" will occur at the contact point, which is marked as a precursor to thermodynamic anomaly.

[0053] The system did not stop at the surface phenomenon of "voltage fluctuations". Instead, it used frequency domain analysis to capture the deep coupling relationship between mechanical vibration (80Hz) and electrical response (ripple). Through semantic mapping, it accurately located the fault to "dynamic contact failure of Module3 bus" and generated a precise anomaly dataset covering three dimensions: mechanical, electrochemical and thermodynamic.

[0054] refer to Figure 3 In step S12, the specific steps are as follows:

[0055] S121: Acquire the three-dimensional spatial model of the battery, map multiple abnormal data to the three-dimensional spatial model in real time, and trigger the traversal of the battery region along the response position of multiple abnormal data to determine the corresponding abnormal region; each abnormal region presents the characteristics of the corresponding functional component.

[0056] S122: Extract corresponding local feature parameters from the abnormal data for each abnormal region; perform evolution trajectory analysis on a continuous time axis to determine the corresponding data deviation; generate an abnormal dynamic curve based on each data deviation and the corresponding data node, which presents the abnormal changes of the battery at different times; at this time, when the data deviation exceeds the preset deviation threshold, the corresponding data node is regarded as an abnormal node.

[0057] In the embodiments of this application, a three-dimensional spatial model of the battery is acquired, and multiple abnormal data are mapped to the three-dimensional spatial model in real time. The battery region is traversed along the response positions of the multiple abnormal data to determine the corresponding abnormal regions. Each abnormal region exhibits the characteristics of the corresponding functional components, thus introducing the characteristic of each abnormal region to exhibit the characteristics of the corresponding functional components.

[0058] At this point, the system pre-builds or calls up a high-precision 3D digital model of the battery pack. This model not only includes the geometric shape, but also contains detailed internal topology, such as the cell arrangement matrix, busbar routing, sampling board position, sensor installation coordinates, and electrical connection topology between modules.

[0059] The system acquires abnormal data with sensor IDs or topology path identifiers; using coordinate transformation algorithms, it maps these abstract data labels to specific meshes or vertices in the 3D model; for example, it maps "abnormal voltage of the third string" to the specific cell coordinates of the third module in the model; and it maps "abnormal vibration of the Z-axis" to a specific area of ​​the bottom structural beam.

[0060] Data mapping typically involves discrete points. This step aims to "expand" these discrete points into physically meaningful functional regions through spatial neighborhood analysis. The system initiates a 3D spatial traversal algorithm (such as flooding or radius search) centered on the mapped point. The algorithm examines the data state within the neighborhood of the center point, searching for neighboring data points with the same or related anomalous trends. The boundary of the anomalous region is determined by calculating the spatial gradient descent of the anomalous intensity. The traversal stops when the data characteristics of neighboring points revert to the normal baseline range, thus defining a closed spatial geometry as the "anomalous region." The defined anomalous region is overlaid and compared with the metadata in the 3D model. The functional components to which this region belongs in the physical design (such as "tab welding area," "module fixing bolt," "liquid cooling channel interface," etc.) are identified, thereby assigning specific engineering attributes to the region.

[0061] Specifically, the system has identified a coupling relationship between voltage ripple and 80Hz vibration through S112; the system has loaded a detailed digital twin model of the battery; the model shows that the battery pack consists of 4 modules connected in series, each module is connected through a copper-aluminum composite bus; according to the BMS sampling topology, the abnormal voltage drop signal was detected from "sampling channel A3", which corresponds to the 3rd module at the bottom rear of the battery pack; the triaxial accelerometer that detected the 80Hz resonance component is located at the crossbeam position between the 3rd and 4th modules.

[0062] Using the connection point between the 3rd and 4th modules as the geometric center, the system triggers a spatial scanning ray; the scanning range covers the positive busbar, negative busbar, and the insulating mounting plate connecting the two modules; during the traversal, the system checks the data of other auxiliary sensors within the spatial cluster (such as nearby temperature probes); it finds that although the absolute temperature value does not exceed the limit, the temperature probe data near the positive output terminal of the 3rd module shows a slight jitter synchronized with the vibration frequency (hot spot jitter).

[0063] The system spatially merges the "voltage anomaly point," "vibration resonance source," and "temperature fluctuation point." Since these three points highly overlap with the positive output terminal of the third module, the system stops traversing and delineates this area as the current "anomaly region." The system retrieves the defined attributes of this region in the model and determines that the core functional component corresponding to this region is the "inter-module positive busbar connection assembly." The characteristics of this region are marked as: a high current-carrying rigid connection component, which includes threaded fasteners and a multi-layer metal lamination structure, and is sensitive to mechanical stress. The system successfully physically locks the electrochemical-mechanical coupling characteristics captured in S112 onto the positive busbar connection assembly of the third battery module. The system no longer reports "low battery voltage" but accurately reports an anomaly in this specific functional component region.

[0064] Furthermore, for each abnormal region, corresponding local feature parameters are extracted from the abnormal data; these local feature parameters are placed on a continuous time axis for evolution trajectory analysis to determine the corresponding data deviation. Based on each data deviation and the corresponding data node, an abnormal dynamic curve is generated, which presents the abnormal changes of the battery at different times. At this point, when the data deviation exceeds the preset deviation threshold, the corresponding data node is designated as an abnormal node, and multiple abnormal data are introduced to control the abnormal characteristics of the battery, thereby improving the accuracy of the abnormal dynamic curve of the battery during operation.

[0065] At this point, based on the boundary of the abnormal region determined by S121 (such as the geometry of a specific cell string or bus), the sensor channels associated with this region (such as a specific voltage sampling line, a nearby NTC thermistor, or a local accelerometer) are locked; characteristic parameters that can characterize the health status of the component are extracted from the original time-series signal; common parameters include: local DC internal resistance (DCIR), local AC impedance amplitude in a specific frequency band, temperature gradient, voltage variance, and vibration kurtosis, etc., which constitute the "characteristic information" of this region.

[0066] The system maintains an adaptive baseline model that records the standard parameter range of the region under similar operating conditions (such as equal SOC, equal ambient temperature, and equal load power). The baseline is not a fixed value but a surface that is dynamically adjusted according to the operating conditions. The distance between the real-time extracted local feature parameters and the dynamic baseline is calculated. Commonly used algorithms include Euclidean distance, Mahalanobis distance, or statistical standard deviation multiples. The data deviation reflects the degree of deviation of the current data point from the normal model and is a dimensionless normalized index. The calculated deviations are arranged in chronological order to form an evolution trajectory, which is used to observe whether the anomaly converges (disappears) or diverges (deteriorates).

[0067] An abnormal dynamic curve is plotted with time on the horizontal axis and data deviation on the vertical axis. The slope of the curve reflects the speed of fault development, and the amplitude of the curve reflects the severity of the fault. Multiple threshold levels are set (such as warning threshold I, alarm threshold II, and severe threshold III). These thresholds are trained based on a historical fault database and take into account measurement noise and operating condition disturbances. When a data point on the dynamic curve crosses a preset threshold and meets certain persistence conditions (to prevent accidental triggering of instantaneous spikes), the time point ti is officially marked as an "abnormal node". The node includes a timestamp, deviation value, parameter type, and corresponding operating condition snapshot.

[0068] Specifically, the abnormal area was identified as the "positive busbar connection component of module 3". Now, we proceed to step S122 to perform in-depth quantitative analysis of this area. The system focuses on the positive busbar area of ​​module 3 and extracts two key parameters: local dynamic contact impedance: calculated based on the ratio of the weak voltage drop across the busbar in this area (after deducting the voltage drop of the battery cell itself) to the current flowing through it; local vibration synchronization rate: the cross-correlation coefficient between the voltage ripple component and the vibration acceleration waveform in this area. High-frequency electromagnetic interference noise in the data is removed, and low-frequency fluctuation components that reflect the physical contact state are retained.

[0069] The system calls the battery health model. Under the current current (150A) and temperature (42℃) conditions, the standard contact impedance baseline should be 0.5mΩ, and the vibration synchronization rate baseline should be close to 0 (no correlation). Trajectory point A (t=10.00s): The vehicle passes over a gravel road surface, and the local dynamic contact impedance instantly rises to 0.8mΩ. Deviation calculation: |0.8-0.5| / 0.5=0.6 (i.e., 60% deviation). At this time, it is judged as a slight disturbance affected by the road surface.

[0070] Trajectory point B (t=10.05s): The vehicle passes through a deep pit and is subjected to a huge impact; the local dynamic contact impedance soars to 2.6mΩ; deviation calculation: |2.6-0.5| / 0.5=4.2 (i.e. 420% deviation).

[0071] Trajectory point C (t=10.10s): The impact ended, but the local dynamic contact resistance dropped to 1.2mΩ and did not fall back; the deviation remained at 1.4 (140% deviation), which indicates that the contact parts underwent plastic deformation (such as a permanent gap caused by loose screws).

[0072] The system plotted a "contact impedance deviation - time" curve; the curve showed an extremely high peak at t=10.05s, and after the peak, it did not return to the zero axis, but remained at a high level; the system set the deviation threshold for "loose mechanical connection" to 1.0 (i.e., 100% deviation); judgment: although the 60% deviation at t=10.00s did not meet the standard, at t=10.05s, the deviation reached 420%, far exceeding the threshold; result: t=10.05s was marked as a first-level abnormal node; from t=10.06s to t=10.10s, the deviation remained above 140%, triggering the "continuous abnormality" logic, and the system generated a new second-level abnormal node sequence, confirming that the fault had changed from "transient impact" to "permanent damage".

[0073] The system successfully transformed the physical "busbar loosening" into the mathematical "impedance trajectory jump" by using the quantitative indicator of "data deviation". In particular, it accurately marked the critical point of the anomaly at t=10.05s and captured the feature that the deviation could not fall back after the impact, thus determining the permanent damage of the functional component.

[0074] refer to Figure 4 In step S13, the specific steps are as follows:

[0075] S131: Mark each abnormal node and perform data filtering in the deep filtering mechanism of multidimensional related data along each abnormal node to output the corresponding related data. Determine the corresponding pseudo-abnormal content based on the matching of each abnormal node and the corresponding related data. The pseudo-abnormal content presents incorrect abnormal information.

[0076] S132: For each pseudo-abnormal content, trigger the screening of each abnormal node and propose abnormal nodes with pseudo-abnormal content to retain multiple final abnormal nodes. Combine the spatial distribution density of multiple final abnormal nodes, the duration on the time axis, and the current working content of the battery to perform multi-dimensional logical weighted decision-making to generate the corresponding abnormal event.

[0077] In the embodiments of this application, each abnormal node is marked, and data is filtered along each abnormal node using a deep filtering mechanism of multidimensional related data to output the corresponding related data. Based on the matching of each abnormal node and the corresponding related data, the corresponding pseudo-abnormal content is determined. The pseudo-abnormal content presents incorrect abnormal information, which is compatible with the overall consideration of matching each abnormal node and the corresponding related data, and ensures the accuracy of the corresponding pseudo-abnormal content.

[0078] At this point, the system marks all deviations exceeding the threshold in S122 as "abnormal nodes." These nodes manifest as spikes, steps, or discontinuous fluctuations on the time axis. For each abnormal node, the system does not only look at a single parameter but initiates a full-stack data retrieval to retrieve multi-dimensional correlation data before and after that moment, including: electrical correlations: current change rate, total voltage fluctuation, and voltage status of other strings; mechanical correlations: triaxial acceleration, suspension displacement, and wheel speed pulses; environmental / control correlations: CAN bus messages, relay status, and fan / pump operating status. Using a time window sliding algorithm, the system extracts data slices synchronized with the abnormal nodes from the multi-dimensional data stream; for example, it extracts the vibration acceleration curve within 100ms before and after the voltage anomaly point.

[0079] The system checks whether abnormal signals conform to physical laws; for example, according to Ohm's law ΔV=I×R, abnormal voltage fluctuations should be strongly correlated with changes in current; if the voltage jumps drastically but the current remains constant and the system does not experience a sudden load change, then electrical consistency is violated.

[0080] If the abnormal signal is completely synchronized in the time domain with a specific interference source (such as high-frequency vibration or electromagnetic radiation burst), and the interference source is known to cause sensor measurement deviation, then a causal relationship is determined between the two. When the matching result shows that the abnormality is caused by external interference or measurement error, rather than a change in the internal physicochemical properties of the battery, the system defines the abnormal node as "pseudo-abnormal content". Its characteristic is "erroneous abnormal information", that is, although it appears to exceed the limit in the data, it is false at the physical level.

[0081] Specifically, in step S122, a voltage anomaly node is marked at t=10.02s, showing that the voltage of the third module drops by 50mV instantaneously; the system locks t=10.02s on the time axis and marks it as "Node-A, anomaly to be identified", with the characteristic attribute "instantaneous voltage drop"; the system extracts a time window [10.00s, 10.04s] centered on t=10.02s; retrieves three-axis acceleration data from the vehicle chassis domain controller; retrieves total circuit current data from the battery manager (BMS); and retrieves the status of the high-voltage relay from the high-voltage system.

[0082] Current data comparison: At t=10.02s, the total battery circuit current stabilized at 145A, without any sudden load drop or increase; according to ΔV=I×ΔR, if the voltage drops, the impedance must increase dramatically instantaneously; however, physically, the impedance of structural components cannot increase dramatically in milliseconds and then recover instantly; Vibration data comparison: The data shows that at t=10.02s, the chassis Z-axis acceleration showed a sharp peak of 8g, corresponding to the wheel hitting the protruding rock; the time point of the voltage drop completely overlapped with the high-g vibration impact (time difference <1ms).

[0083] The system determined that the voltage drop was not due to abnormal electrochemical reactions inside the battery or a real change in impedance, but rather to measurement noise caused by momentary micro-movement of the voltage sampling line (Kelvin line) due to severe mechanical vibration, or momentary disconnection of the circuit board connector. This information was marked as "pseudo-abnormal content" and described as: "Sampling link transient interference caused by mechanical vibration".

[0084] The system successfully identified the pseudo-anomaly content corresponding to the abnormal node at t=10.02s. Without this screening step, the system would have misjudged it as an internal short circuit or open circuit in the battery, thus triggering unnecessary shutdown protection. By eliminating this pseudo-anomaly, the system avoided causing unnecessary fright and misoperation to the driver.

[0085] Furthermore, for each pseudo-abnormal content, the filtering of each abnormal node is triggered, and abnormal nodes with pseudo-abnormal content are proposed to retain multiple final abnormal nodes. The spatial distribution density and duration on the time axis of multiple final abnormal nodes are combined with the current working content of the battery to perform multi-dimensional logical weighted decision-making to generate corresponding abnormal events. This introduces the generation of corresponding abnormal events, and at the same time, realizes the overall consideration of multiple abnormal nodes and the current working content of the battery, improving the accuracy of abnormal events.

[0086] At this point, the system receives the "pseudo-abnormal content" list output by S131 and establishes a filter mask; all abnormal nodes marked as "vibration interference", "electromagnetic pulse (EMC) interference" or "sensor drift" are marked as Invalid; the nodes that are not eliminated are marked as final abnormal nodes. These nodes must meet the "physical authenticity" principle, that is, their characteristic changes conform to the basic laws of electrochemistry or thermodynamics inside the battery and cannot be explained by external interference; a complete operating condition snapshot (such as SOC, temperature, power) is bound to each final abnormal node at that moment to provide a data basis for subsequent weighted analysis.

[0087] In the three-dimensional spatial model of the battery, the degree of spatial clustering of the final abnormal nodes is calculated; the system uses a spatial clustering algorithm (such as DBSCAN) to analyze whether the nodes are concentrated in a specific module or bus area (high density, indicating component failure) or randomly scattered throughout the battery pack (low density, indicating system-level noise or uniform aging).

[0088] Calculate the span of the final abnormal node in the time series; statistically analyze whether the abnormal characteristics are single pulses (transient), periodic occurrences (intermittent), or continuous existence (steady state); the longer the duration, the stronger the physical irreversibility of the fault (such as continuous fretting wear after bolt loosening).

[0089] The system reads the battery's current operating status, including operating conditions (rapid acceleration / steady speed / charging), environmental conditions (high and low temperatures), and load rate. Under "high-power charging" conditions, the weight of voltage fluctuation judgment is appropriately reduced (because the increase in polarization internal resistance is a normal phenomenon). Under "resting" conditions, any tiny voltage fluctuation will be given extremely high weight. If the SOC is high (>80%), the activity is strong and the risk of thermal runaway is high. Therefore, temperature-related abnormal nodes are given higher weight, and a weighted comprehensive risk score is calculated. When the score exceeds the event trigger threshold, the system packages all related final abnormal nodes to generate an "abnormal event". This event includes event ID, type (electrical / thermal / mechanical), level, source location coordinates, and confidence level.

[0090] Specifically, after screening by S131, the interference node at t=10.02s has been eliminated; the system now focuses on analyzing a series of abnormal nodes at t=10.05s and thereafter; the system executes the elimination instruction; the node at t=10.02s (pure vibration interference) identified in S131 is removed;

[0091] The system retains the node at t=10.05s: characterized by "contact impedance step + voltage drop", and this drop and current change conform to Ohm's law characteristics; the system retains the continuous nodes from t=10.06s to t=10.15s: characterized by "impedance oscillating at a high level", accompanied by a continuous vibration signal. This group of nodes is identified as "final anomalous nodes" because they reflect the actual change in the physical connection state (plastic deformation / loosening).

[0092] The system maps these nodes in the 3D model of the battery; it finds that all the retained nodes are highly overlapping in spatial coordinates and are densely distributed in the "positive busbar of the third module" area; showing extremely high spatial density, ruling out the possibility of random system error, and locking in a local physical fault.

[0093] Analysis shows that starting from t=10.05s, the abnormal characteristics did not disappear after the vibration subsided, but persisted for more than 10 seconds and showed a divergent trend (with a slight increase in impedance). This is a "persistent anomaly" rather than a "transient disturbance", indicating that the fault has irreversible physical properties.

[0094] The system identifies the current working condition as "medium speed driving (unpaved road surface)" with a SOC of 65%. Although road bumps (work content) usually mask faults, given the extremely high confidence levels of the two characteristic indicators "high spatial density" and "temporal continuity", the system gives these two parameters a very high weight (Weight=0.9) to overcome the uncertainty caused by environmental interference.

[0095] The comprehensive risk score is calculated as Score = (spatial density score × W1) + (duration score × W2) + (operating condition correction score × W3); the result far exceeds the fault confirmation threshold (Threshold = 85); the system officially generates "abnormal event EVT-202X-A3-001"; the event description is: "Dynamic contact failure occurred in the positive busbar of module 3 (P0 level)".

[0096] refer to Figure 5 In step S14, the specific steps are as follows:

[0097] S141: In this abnormal event, multiple abnormal items are identified based on the identification of the abnormal event, and the item content of each abnormal item is marked. The corresponding abnormal priority is determined according to the item content of each abnormal item and the corresponding abnormal area.

[0098] S142: Collect multiple temperature data of the battery, comprehensively consider the project content of the abnormal project, the corresponding abnormal priority, and multiple temperature data of the battery, and define the core range affected by the abnormality and its diffusion path through three-dimensional heat map gradient analysis, so as to lock the multi-factor abnormal area.

[0099] S143: Call the preset "abnormal-maintenance" mapping database and determine the abnormal maintenance mapping relationship of the battery. Dynamically match the abnormal maintenance mapping relationship of the multi-factor abnormal area and the battery, and determine multiple abnormal maintenance measures during the matching process. Based on the multiple abnormal maintenance measures and the current working content of the battery, dynamically coordinate and schedule them to dynamically construct a multi-level maintenance system for the battery. The multi-level maintenance system of the battery includes immediate emergency blocking content, short-term trend suppression content, and long-term performance repair content.

[0100] In the embodiments of this application, in the abnormal event, multiple abnormal items are identified based on the identification of the abnormal event, and the item content of each abnormal item is marked. The corresponding abnormal priority is determined according to the item content of each abnormal item and the corresponding abnormal area, which takes into account the overall consideration of the item content of each abnormal item and the corresponding abnormal area, and ensures the accuracy of the corresponding abnormal priority.

[0101] At this point, the system takes the "abnormal event" as input and deconstructs it level by level based on the fault tree analysis (FTA) model. A physical event is often accompanied by failures in multiple technical dimensions. The system identifies and establishes specific abnormal items, which typically cover the following dimensions: Electrical dimension items: involving impedance changes, voltage / current distortion, insulation integrity, etc.; Thermal dimension items: involving abnormal heat generation, uneven heat distribution, thermal runaway risk, etc.; Mechanical / structural dimension items: involving loose connections, seal failure, mechanical vibration, etc.

[0102] The system traces back the data streams of S12 and S13 to extract key physical quantities related to each project; the extracted values ​​are then filled into the project attributes to form a structured content description; for example, for electrical projects, “contact impedance increment (ΔR)” and “voltage drop amplitude (ΔV)” are marked; for thermal projects, “virtual hot spot temperature” and “thermal power density” are marked; additional time-varying characteristics of the data, such as “step change”, “continuous divergence”, and “periodic oscillation”, are used to characterize the dynamic characteristics of the fault.

[0103] Assess the location weight of abnormal regions within the battery system topology; for example, the high-voltage main circuit, busbars, and positive / negative electrode connections have the highest weight (Critical), while sampling circuits and auxiliary low-voltage circuits have lower weights; assess the severity based on the quantitative values ​​of the project content; for example, projects involving thermal runaway thresholds and insulation breakdown risks are assigned the highest severity; calculate the priority value using the weighted risk function P=f(Warea,Scontent); map the calculation results to standard levels, such as P0 (Critical / Urgent), P1 (Severe / Important), and P2 (General / Alert).

[0104] Specifically, S13 has generated the abnormal event: "EVT-A3-001: Dynamic contact failure of the positive busbar of the 3rd module"; S13 has generated the abnormal event: "EVT-A3-001: Dynamic contact failure of the positive busbar of the 3rd module"; now proceed to step S141 to deconstruct and prioritize this event; the system deconstructs the following three abnormal items: Abnormal item A: Dynamic high impedance abnormality (belonging to the electrical dimension); Abnormal item B: Local Joule heat accumulation (belonging to the thermal dimension); Abnormal item C: Mechanical connection loosening (belonging to the mechanical dimension).

[0105] Project A Content Label: The system label content is: "The contact resistance undergoes a step change at t=10.05s, with an increment ΔR≈1.4mΩ; under dynamic operating conditions, the voltage ripple (RippleVoltage) amplitude increases by 300% in the 80Hz frequency band."

[0106] Project B content labeling: The system combines thermal model inversion, and the labeling content is: "Local thermal power loss at the fault point Ploss≈I2ΔR≈31.5W; Local virtual temperature Tvirtual is estimated to be 105℃, and shows a continuous upward trend."

[0107] Project C Content Marking: The system is based on vibration analysis, and the marking content is: "A continuous mechanical resonance characteristic appears in the 80Hz±5Hz frequency band, which is synchronized with the electrical impedance fluctuation, indicating that the physical connection pair (bolt / busbar lug) has lost its preload and fretting wear has occurred."

[0108] The abnormal area is located in the "positive busbar of module 3", which is the main current channel of the series battery pack. If this area is completely disconnected, it will cause the high voltage of the whole vehicle to drop. If arcing occurs, it is very easy to ignite the surrounding modules. The system determines that the criticality weight of this area is Level 1 (the highest level).

[0109] Project B (Local Joule Heat Accumulation): The virtual temperature has reached 105°C, approaching the melting point (approximately 130°C) of the surrounding insulating materials (such as PET film); if not controlled, it is extremely easy to ignite; Priority: P0 (highest priority - catastrophic risk).

[0110] Project A (Dynamic High Impedance Anomaly): Affects power output efficiency, causes voltage drop, triggers undervoltage protection, and causes the vehicle to "break down," but usually does not directly lead to an instantaneous fire; Priority: P1 (High Priority - Critical Fault).

[0111] Item C (mechanical connection loosening): This is a physical damage source, but it develops relatively slowly (unless subjected to another severe impact) and will not immediately lead to a safety accident; Priority: P2 (medium priority - functional degradation).

[0112] The system successfully quantified the general description of "busbar looseness" into three abnormal items with specific parameters. More importantly, through risk assessment, it clarified the maintenance logic of "first put out the fire (suppress the thermal risk of P0), then protect the power (handle the electrical fault of P1), and then repair the vehicle (solve the mechanical problem of P2)," providing clear weight guidance for the heat map analysis of S142.

[0113] Furthermore, multiple temperature data points of the battery were collected. Taking into account the content of the abnormal project, the corresponding abnormality priority, and the multiple temperature data of the battery, the core range affected by the abnormality and its diffusion path were defined by three-dimensional heat map gradient analysis, so as to lock the multi-factor abnormal area and introduce the method of locking the multi-factor abnormal area.

[0114] At this time, the data of the NTC thermistor array deployed in the battery pack is read in real time to obtain the physical temperature of key monitoring points; combined with the abnormal items marked in S141 (such as "Joule heat power loss 31.5W"), the virtual temperature of the abnormal source point is calculated using the thermoelectric coupling model; this inversion data is crucial for blind areas that cannot be directly covered by sensors (such as busbar connections); at the same time, boundary conditions such as coolant flow rate and ambient temperature are collected as input parameters for thermal map reconstruction.

[0115] Based on the finite difference method or radial basis function (RBF) interpolation algorithm, a continuous, gridded transient temperature field is reconstructed on the 3D CAD model of the battery pack by fusing physical temperature and virtual temperature points. This fills the data gaps between sensors; the temperature gradient vector is then calculated on the 3D grid. The magnitude of the vector represents the degree of temperature change (thermal steepness), and the direction of the vector represents the main path of heat conduction (thermal flow line).

[0116] Search temperature gradient magnitude Mesh sets exceeding a set threshold and with absolute temperatures exceeding the material's tolerance limit are defined as the "core region affected by the anomaly," typically an irregular sphere or ellipsoid centered at the fault source; along the gradient vector The streamline integral is performed in the positive direction to trace the path of heat transfer to adjacent modules or structural components; at the same time, combined with the anomaly priority in S141 (such as P0 level), the path length is extended and predicted (considering thermal inertia) to identify potential areas affected by thermal radiation or heat conduction; the "electrical fault coordinates", "thermal core range" and "thermal diffusion path" are performed in Boolean union operation in the spatial topology to finally lock a multi-factor anomaly area, which not only contains the current fault point, but also the associated area that is highly likely to be damaged in the future.

[0117] Specifically, S141 has identified a P0-level "local Joule heat accumulation" item; the system collected the NTC temperature of the 3rd module body at 45℃ and the NTC temperature of the 4th module at 42℃. These surface data seem normal and have not reached the alarm threshold; however, the system also retrieved the item content in S141: there is an abnormal heat power injection of 31.5W at the busbar; based on the heat capacity of the busbar and the ambient heat transfer coefficient, the system inversely calculated that the real virtual temperature of the fault point (connector) has reached 105℃. This key data will be used as the "heat source singularity" for thermal map reconstruction.

[0118] In the digital twin model of the battery, the system uses a virtual heat source point of 105℃ as the center, with the busbar coordinates (x, y, z) as the center, for interpolation calculations. The generated three-dimensional thermal map shows that the high-temperature area is concentrated in a narrow space above the busbar and decreases exponentially outwards. The system calculations found that there is a huge temperature gradient (up to 20℃ / cm) between the air and the insulating medium area above the busbar. Gradient vector analysis shows that the heat flow has two main directions: the main path: conduction along the copper busbar to the positive electrode of the 4th module; the secondary path: radiation upwards to the battery pack cover plate through thermal convection.

[0119] The system defines the isosurface with a temperature exceeding 80°C (the critical temperature for softening insulation material) as the core range. This range includes the faulty bus body, its fastening bolts, and the insulating flame-retardant sleeve wrapped above it. The system tracks the main path gradient and finds that heat has been conducted to the series cable between the 3rd and 4th modules. Tracking the secondary path gradient, it finds that the temperature of the inner wall of the battery pack cover plate is rising rapidly.

[0120] Considering that S141 determined the project to be at level P0 (highest priority), the system introduced a thermal diffusion safety factor of 1.5 times when locking the space; the finally locked multi-factor anomaly area is a complex geometry that covers: "the positive busbar assembly of the 3rd module + the connecting harness between the 3rd and 4th modules + the inner liner of the corresponding area's upper cover plate". This area is marked as a high-risk area, and any cooling strategy for this area (S143) must cover all the above components.

[0121] Therefore, a pre-defined "anomaly-maintenance" mapping database is invoked, and the abnormal maintenance mapping relationship of the battery is determined. The abnormal maintenance mapping relationship of the multi-factor anomaly area and the battery is dynamically matched, and multiple abnormal maintenance measures are determined during the matching process. Based on the multiple abnormal maintenance measures and the current working content of the battery, dynamic collaborative scheduling is carried out to dynamically construct a multi-level maintenance system for the battery. This multi-level maintenance system for the battery includes immediate emergency blocking, short-term trend suppression, and long-term performance repair. At the same time, the multi-factor anomaly area is controlled, realizing the overall consideration of multiple abnormal maintenance measures and the current working content of the battery, thus improving the accuracy of the multi-level maintenance system for the battery.

[0122] At this point, the system calls the preset "anomaly-maintenance" mapping database (KnowledgeGraph), which stores the correspondence between various fault feature vectors and maintenance strategy templates; it performs fuzzy matching of the attributes (such as location, temperature, fault type P0 / P1) of the currently locked "multi-factor anomaly area" with the templates in the database; once a match is successful, the system establishes the corresponding anomaly maintenance mapping relationship, which usually includes a set of basic maintenance actions and logical constraints between each action (such as mutual exclusion relationships and preconditions).

[0123] Input the spatial coordinates and thermal gradient data of the "multi-factor abnormal area" locked in S142 into the matching algorithm; determine the execution parameters of specific measures based on the temperature level of the core area and the direction of the diffusion path; generate a set of specific abnormal maintenance measures, such as: electrical side: power limiting (SOP limit), contactor logic adjustment; thermal side: cooling pump speed adjustment, liquid cooling valve on / off control, fan duty cycle adjustment; data side: log encryption recording, fault code freezing.

[0124] The system acquires the battery's current operating context, including vehicle speed, motor torque request, driving intent (rapid acceleration / cruising), and ambient temperature. It dynamically adjusts maintenance measures based on real-time operating conditions. If operating conditions conflict with maintenance measures (e.g., rapid acceleration is required but power must be limited), arbitration is performed based on safety weights. The final determined measures are constructed into a three-tiered system according to time and depth of effect: Immediate emergency interruption: millisecond-level response, designed to break the chain of danger and prevent catastrophic consequences; Short-term trend suppression: second- to minute-level response, designed to maintain the fault state at a controllable level by adjusting system boundaries; Long-term performance repair: storage and delayed execution, designed to eliminate the root cause of the fault or plan maintenance, restoring the system's health throughout its entire lifecycle.

[0125] Specifically, S142 has identified the multi-factor anomaly area as the "heat-affected zone of the third module busbar" (including the busbar, insulating bushing, and adjacent connecting harnesses). Based on the feature vector of "dynamic contact failure at high-voltage connection + local P0 level overheating", the system queries the database and matches the strategy template "Strategy-HighImpedance-Thermal-Control". This template defines the basic handling logic for this type of fault: the heat source must be suppressed first, the current stress must be limited, and the fault must be recorded.

[0126] The system inputs the "bus coordinates" and "core temperature 105℃" output by S142 into the matching algorithm; the system parses and determines the following specific measures from the template: Measure M1: Immediately limit the maximum discharge current (reduce heat source power); Measure M2: Request the maximum cooling flow for the liquid cooling circuit where the 3rd module is located; Measure M3: Increase the insulation monitoring frequency of the bus area (spark prevention monitoring); Measure M4: Freeze the fault data and upload it to the cloud.

[0127] The system detected that the vehicle's current speed was 60 km / h, it was on an unpaved road, and the driver had not requested to stop. If the power was cut off directly, the vehicle would lose power steering and vacuum brake assist, posing a significant safety risk. The system decided not to execute the extreme measure of "cutting off the contactor," but instead to retain a minimum power output to maintain basic maneuverability; it would fully execute M1 and M2; ultimately constructing a multi-level maintenance system.

[0128] Immediate emergency shutdown (millisecond-level execution): Instead of directly cutting off power, the current change rate is limited to within -50A / s to -50A / s to prevent arcing caused by instantaneous current interruption; the allowable power of the BMS is instantly reduced from the current 200kW to 50kW; this measure will significantly reduce the current flowing through the busbar, thereby physically cutting off the source of Joule heating and preventing P0-level thermal runaway;

[0129] Short-term trend suppression (execution at the second to minute level): The system instructs the thermal management controller (TMS) to increase the water pump speed to 110% of the rated value and adjust the three-way valve to prioritize the cooling branch where the 3rd module is located; the coolant is forced to carry away the heat accumulated in the "multi-factor abnormal area" locked by S142, and the manifold temperature gradient is gradually flattened from 20℃ / cm to suppress the diffusion path of heat to the 4th module and the top cover plate;

[0130] Long-term performance repair content (storage and delayed execution): The voltage, current, vibration, temperature and SOC data within 10 seconds before and after the fault occurred are encrypted and written to the BMS non-volatile memory with a high sampling rate, and an indelible "permanent fault code" is generated; a work order for "the high voltage connector of the third module needs to be repaired" is pushed to the fleet management backend through the Telematics module, and the super fast charging function of the vehicle is remotely locked to prevent the next high current charging from aggravating the damage until the repair is completed.

[0131] Please see Figure 6 , Figure 6 This is a schematic diagram of the structural composition of the battery abnormality maintenance system during dynamic use in an embodiment of the present invention; the battery abnormality maintenance system during dynamic use is applied to the above-mentioned battery abnormality maintenance method during dynamic use; the battery abnormality maintenance system during dynamic use includes:

[0132] The abnormal data module 21 is used to align multiple data points of the battery over time and identify multiple data combinations, each data combination covering voltage data, current data, temperature data, and vibration data; based on the identification of multiple data combinations, corresponding abnormal data are identified.

[0133] The abnormal dynamic curve module 22 is used to determine the abnormal characteristics of the battery based on multiple abnormal data and the abnormal area corresponding to the battery, and to determine the abnormal dynamic curve of the battery during operation. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes.

[0134] The abnormal event module 23 is used to trigger the filtering of each abnormal node based on each abnormal node and its corresponding associated data, so as to exclude false abnormal nodes and identify multiple final abnormal nodes; and to determine the corresponding abnormal events based on multiple abnormal nodes and the current working content of the battery.

[0135] The multi-level maintenance system module 24 is used to determine multi-factor abnormal areas based on various abnormal items of abnormal events, corresponding abnormal priorities, and multiple temperature data of the battery, and to determine multiple abnormal maintenance measures based on the abnormal maintenance mapping relationship between the multi-factor abnormal areas and the battery, and to construct a multi-level maintenance system for the battery based on the multiple abnormal maintenance measures and the current working content of the battery.

[0136] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for abnormal maintenance of a battery during dynamic use, characterized in that, include: Multiple battery data points are aligned over time, and multiple data combinations are determined, each encompassing voltage, current, temperature, and vibration data. Multiple abnormal data are identified based on the identification of multiple data combinations, including: the battery is installed in the electric vehicle, the battery management system of the battery is marked, and multiple data of the battery are determined based on the monitoring of the battery management system. The multiple data are aligned on the same time axis, and the coupling relationship of each data is combined to determine multiple feature fusion data combinations. Each data combination covers voltage data, current data, temperature data and vibration data. Based on multiple abnormal data and the abnormal regions corresponding to the battery, the abnormal characteristics of the battery are determined, and the abnormal dynamic curve of the battery during operation is determined. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes. This includes: acquiring a three-dimensional spatial model of the battery, mapping multiple abnormal data to the three-dimensional spatial model in real time, and triggering the traversal of the battery region along the response position of multiple abnormal data to determine the corresponding abnormal region; each abnormal region presents the characteristics of the corresponding functional component. The process involves filtering each anomalous node based on its associated data to eliminate false anomalous nodes and identify multiple final anomalous nodes. The corresponding anomalous events are then determined based on these multiple anomalous nodes and the battery's current operational status, including: Each abnormal node is marked, and data is filtered along each abnormal node using a deep filtering mechanism of multidimensional related data to output the corresponding related data. Based on the matching of each abnormal node and the corresponding related data, the corresponding pseudo-abnormal content is determined, which presents incorrect abnormal information. Based on the various abnormal items of the abnormal events, their corresponding abnormal priorities, and multiple temperature data of the battery, a multi-factor abnormal area is determined. Based on the mapping relationship between the multi-factor abnormal area and the battery's abnormal maintenance, multiple abnormal maintenance measures are determined. Based on the multiple abnormal maintenance measures and the battery's current working content, a multi-level maintenance system for the battery is constructed.

2. The method for abnormal maintenance of a battery during dynamic use according to claim 1, characterized in that, The battery's multiple data points are aligned in time, and multiple data combinations are determined, each data combination encompassing voltage data, current data, temperature data, and vibration data. Identifying multiple anomalous data based on multiple data combinations also includes: In each data combination, voltage data, current data, temperature data and vibration data are correlated and scanned, and features are extracted along multiple frequency domain dimensions to determine the corresponding feature vector, which reflects the internal response of the battery. The feature vectors are semantically mapped to the spatial topology and operating conditions of the battery, and multiple anomalous data are identified, covering electrochemical parameters, thermodynamic parameters and mechanical parameters.

3. The method for abnormal maintenance of a battery during dynamic use according to claim 1, characterized in that, The abnormal characteristics of the battery are determined based on multiple abnormal data and the abnormal regions corresponding to the battery, and an abnormal dynamic curve of the battery during operation is determined. This abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes, including: For each abnormal region, corresponding local feature parameters are extracted from the abnormal data; the local feature parameters are placed on a continuous time axis for evolution trajectory analysis to determine the corresponding data deviation degree; based on each data deviation degree and the corresponding data node, an abnormal dynamic curve is generated, which presents the abnormal changes of the battery at different times; at this time, when the data deviation degree exceeds the preset deviation degree threshold, the corresponding data node is regarded as an abnormal node.

4. The method for abnormal maintenance of a battery during dynamic use according to claim 1, characterized in that, The process involves triggering the filtering of each abnormal node based on its corresponding associated data to exclude false abnormal nodes and thus identify multiple final abnormal nodes. The corresponding abnormal events are determined based on the current operating status of multiple abnormal nodes and batteries, including: For each pseudo-abnormal content, the various abnormal nodes are selected and abnormal nodes with pseudo-abnormal content are proposed to retain multiple final abnormal nodes. The spatial distribution density of multiple final abnormal nodes, the duration on the time axis, and the current working content of the battery are combined to perform multi-dimensional logical weighted decision-making to generate the corresponding abnormal event.

5. The method for abnormal maintenance of a battery during dynamic use according to claim 1, characterized in that, The system determines multi-factor anomaly regions based on various anomaly items, corresponding anomaly priorities, and multiple temperature data of the battery. Based on the mapping relationship between these multi-factor anomaly regions and the battery's anomaly maintenance, it determines multiple anomaly maintenance measures. Finally, based on these multiple anomaly maintenance measures and the battery's current operating conditions, it constructs a multi-level battery maintenance system, including: In this abnormal event, multiple abnormal items are identified based on the identification of the abnormal event, and the content of each abnormal item is marked. The corresponding abnormal priority is determined according to the content of each abnormal item and the corresponding abnormal area. Multiple temperature data points of the battery are collected. Taking into account the content of the abnormal project, the corresponding abnormality priority, and the multiple temperature data of the battery, the core range affected by the abnormality and its diffusion path are defined by three-dimensional heat map gradient analysis, so as to lock the multi-factor abnormal area.

6. The method for abnormal maintenance of a battery during dynamic use according to claim 5, characterized in that, The process involves determining multi-factor anomaly regions based on various anomaly items of abnormal events, their corresponding anomaly priorities, and multiple temperature data of the battery. Based on this multi-factor anomaly region and the battery's anomaly maintenance mapping relationship, multiple anomaly maintenance measures are determined. Furthermore, a multi-level battery maintenance system is constructed based on these multiple anomaly maintenance measures and the battery's current operating conditions. This also includes: The system calls the preset "abnormality-maintenance" mapping database and determines the abnormal maintenance mapping relationship of the battery. It then dynamically matches the abnormal maintenance mapping relationship between the multi-factor abnormal area and the battery, and determines multiple abnormal maintenance measures during the matching process. Based on the multiple abnormal maintenance measures and the current working content of the battery, it dynamically coordinates and schedules them to build a multi-level maintenance system for the battery. This multi-level maintenance system includes immediate emergency blocking, short-term trend suppression, and long-term performance repair.

7. A battery malfunction maintenance system during dynamic use, characterized in that, The abnormal maintenance system for batteries under dynamic use is applied to the abnormal maintenance method for batteries under dynamic use as described in any one of claims 1-6; The battery's abnormal maintenance system during dynamic use includes: The abnormal data module is used to align multiple data points of the battery over time and identify multiple data combinations, each of which includes voltage data, current data, temperature data, and vibration data; based on the identification of multiple data combinations, corresponding abnormal data are determined. The abnormal dynamic curve module is used to determine the abnormal characteristics of the battery based on multiple abnormal data and the abnormal area corresponding to the battery, and to determine the abnormal dynamic curve of the battery during operation. The abnormal dynamic curve presents the abnormal changes of the battery at different times and marks the corresponding abnormal nodes. The abnormal event module is used to trigger the filtering of each abnormal node based on each abnormal node and its corresponding associated data, so as to eliminate false abnormal nodes and identify multiple final abnormal nodes; and to determine the corresponding abnormal events based on multiple abnormal nodes and the current working content of the battery. The multi-level maintenance system module is used to determine multi-factor abnormal areas based on various abnormal items of abnormal events, corresponding abnormal priorities, and multiple temperature data of the battery. Based on the abnormal maintenance mapping relationship between the multi-factor abnormal areas and the battery, multiple abnormal maintenance measures are determined. The multi-level maintenance system of the battery is constructed according to the multiple abnormal maintenance measures and the current working content of the battery.

Images