Battery thermal runaway fault cell parameter tracing method and system

Abstract

The invention provides a battery thermal runaway fault cell parameter tracing method and system, and relates to the technical field of battery safety monitoring and fault diagnosis. The method comprises the following steps: presetting a trigger condition, and starting a traceability algorithm when the condition is met; constructing and reading basic binding data; collecting two-dimensional data; data preprocessing and abnormal feature extraction; reversely positioning the initial battery cell based on the binding relationship; performing multi-parameter cross validation; comparing and verifying adjacent battery cells; and outputting a positioning result and generating a traceability report. The system comprises a trigger module, a basic data construction and reading module, a two-dimensional data acquisition module, a data processing and feature extraction module, an initial positioning module, a cross verification module, a comparison verification module and a result output and report generation module. By means of multi-parameter fusion, precise binding and cross validation, second-level precise positioning of the thermal runaway fault cell is achieved, the positioning precision is larger than or equal to 99%, the misjudgment rate is smaller than or equal to 1%, and the maintenance efficiency is improved by 50% or above.

Description

Technical Field

[0001] This invention relates to the field of battery safety monitoring and fault diagnosis technology, specifically to a method and system for tracing the parameters of battery cells in thermal runaway faults. Background Technology

[0002] In the field of thermal runaway early warning for power batteries and energy storage batteries, existing technologies generally suffer from low accuracy in locating faulty cells, high false alarm rates, and poor troubleshooting efficiency. Specifically, these problems manifest as follows: Single parameter dependence: Most systems rely on a single parameter such as temperature or voltage for alarms, which cannot distinguish between "single cell fault" and "module-level fault", and can easily lead to misjudgment of the entire battery pack.

[0003] Insufficient positioning granularity: Existing solutions can only locate at the "module" level, and cannot be accurate to the specific battery cell, resulting in a large inspection scope and high maintenance costs.

[0004] Lack of cross-validation: Without combining adjacent cell parameters and historical baseline data for multi-dimensional verification, non-cell faults such as "uneven heat dissipation" and "electromagnetic interference" are easily misjudged as precursors to thermal runaway. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for tracing the parameters of battery cells with thermal runaway faults. By using multi-parameter fusion, precise binding, and cross-validation, it can achieve second-level accurate positioning of thermal runaway faulty cells, reduce the false judgment rate, and improve the investigation efficiency.

[0006] To achieve the above-mentioned objectives, this invention provides a method for tracing the parameters of battery cells in thermal runaway faults, comprising the following steps: Step 1: Preset the algorithm start trigger conditions. When the sensors in the battery management system's perception layer detect that the core parameters meet the trigger conditions, the tracing algorithm is started. Step 2: Construct and read basic binding data, which includes binding relationship data between battery cells and sensors, and battery pack topology data; Step 3: Simultaneously collect real-time anomaly data and historical baseline data to form a two-dimensional data support; Step 4: Preprocess the two-dimensional data and extract abnormal features; Step 5: Based on the binding relationship data between the battery cell and the sensor, the initial battery cell is located in reverse through the abnormal sensor ID, and the location process adopts a dual verification mechanism; Step 6: Perform multi-parameter cross-verification on the initially located battery cells to determine whether they are suspected faulty battery cells; Step 7: Compare and verify the parameters of the suspected faulty cell with those of adjacent cells to distinguish between single-cell faults and module-level faults, and determine the range of faulty cells. Step 8: Output the location results of the faulty battery cell and generate a traceability report.

[0007] Preferably, the triggering condition in step 1 adopts a dual logic of threshold triggering and trend prediction; if the instantaneous value of a single core parameter exceeds the preset safety range, or if multiple consecutive acquisition cycles show an abnormal change trend, the algorithm is activated; the core parameters include temperature, impedance, and voltage.

[0008] Preferably, the binding relationship data between the battery cell and the sensor in step 2 is constructed in the following way: Step 2.1, Production Binding Stage: When the battery cell enters the module, each battery cell is equipped with a dedicated temperature sensor and impedance sensor, and an independent voltage sampling chip channel is allocated. The cell ID and sensor ID are collected and uploaded to the battery production MES system, and the system verifies the uniqueness of the binding. Step 2.2, System Solidification Stage: After the module is installed into the battery pack, the BMS reads the sensor ID information through the SPI interface, writes the four-level association relationship of "battery pack number - module number - cell ID - sensor ID" into the EEPROM storage module, and synchronizes it to the host computer system at the same time; The battery pack topology data includes the cell series and parallel connection method, physical arrangement, electrical connection relationship, number of cells in each string, physical location coordinates of each cell, electrical connection relationship between adjacent cells, and correspondence between the module and the battery pack interface.

[0009] Preferably, the real-time abnormal data in step 3 includes temperature parameters, impedance parameters, state of charge (SOC), and ion and electrical parameters; the temperature parameters include the instantaneous value of the cell surface temperature and the rate of temperature rise; the impedance parameters include the real and imaginary parts of the AC impedance, the impedance change amplitude, and the duration of the change; the state of charge (SOC) includes the real-time value and the SOC jump amplitude; the ion and electrical parameters include the lithium-ion concentration characteristic signal, the amplitude of sudden rises and falls in charging and discharging current, and the terminal voltage fluctuation value. The real-time abnormal data collection frequency is as follows: temperature and voltage are collected once every 100ms~150ms, impedance is collected once every 500ms~600ms, and SOC and lithium ion concentration are collected once every 1s~2s. The historical baseline data mentioned in step 3 refers to the parameter range of normal cells of the same batch and model under the same operating conditions, and meets the requirements of operating condition matching, parameter integrity, and dynamic updating; the historical baseline data is automatically updated every 100 to 150 normal charge and discharge cycles.

[0010] Preferably, the preprocessing in step 4 includes: Step 4.1, Filtering and Denoising: The Kalman filter algorithm is used to process the high-frequency fluctuation signals of temperature and voltage to eliminate instantaneous interference; Step 4.2, outlier correction: When sensor data exceeds the physical reasonable range, the average value of the parameters of the previous 5 to 6 collection cycles of the cell is automatically called to fill the data and marked as "data abnormal - to be verified"; Step 4.3, Data Alignment: Align parameters from different acquisition frequencies to a 100ms time axis and supplement missing data points using linear interpolation. The abnormal feature extraction in step 4 adopts a judgment method that combines "quantitative threshold + qualitative trend". If any two or more features are satisfied, it is judged as an "abnormal cell". The abnormal features include temperature features, impedance features, SOC features, lithium ion features, current features, and voltage features.

[0011] Preferably, the dual verification mechanism in step 5 includes: Step 5.1, Level 1 matching: Quickly query the corresponding cell ID in the binding table using the sensor ID corresponding to the abnormal parameters to achieve preliminary positioning; Step 5.2, Secondary Verification: Retrieve data from other sensors bound to the battery cell to confirm whether there is a multi-sensor synchronization anomaly. If only a single sensor is abnormal, mark it as "suspected sensor failure". If multiple sensors are abnormal, proceed to the next verification step.

[0012] Preferably, the multi-parameter cross-validation in step 6 includes: Step 6.1, Feature Statistics: Count the number and specific values ​​of abnormal features that the battery cell meets in the preprocessed data; Step 6.2, weight determination: Temperature, voltage, and impedance are the core features, with a weight of 30%; SOC, lithium-ion concentration, and current are the auxiliary features, with a weight of 20%. If the total weight score is ≥60%, it is determined as a "suspected faulty cell". Step 6.3, Result Processing: If the weighted score requirement is met, proceed to the adjacent cell comparison and verification stage; if not, it is determined to be a "sensor false alarm", the abnormal mark is cleared and the data is reread. If the sensor false alarms accumulate to 3 times, a sensor fault alarm will be triggered.

[0013] Preferably, the adjacent cells in step 7 include physically adjacent and electrically adjacent cells; physically adjacent refers to the 2-4 cells in the same module that are closest to the suspected cell, and electrically adjacent refers to adjacent cells that are in the same series branch or parallel branch as the suspected cell; The acquisition time window for the comparative verification is 3 seconds before and after the suspected cell abnormality start time, and the acquisition frequency is consistent with that of the suspected faulty cell. Single cell fault determination: If only the parameters of the suspected cell meet the abnormal characteristics, and the parameters of all adjacent cells are within the historical baseline range, and the parameter fluctuation range is ≤50% of the baseline range, then it is confirmed as a "single cell fault". Module-level fault determination: If a suspected cell and at least two adjacent cells exhibit abnormal characteristics at the same time, or if the parameters of adjacent cells do not exceed the baseline range but the fluctuation range is ≥80% of the baseline range, then it is determined to be a "module-level fault". Boundary case handling: If only one adjacent cell shows a slight abnormality with a fluctuation range of 60%-80% of the baseline range, the acquisition time window is extended to 10 seconds, and the parameter change trend is re-compared. If the suspected cell abnormality continues to worsen and the abnormality of adjacent cells does not spread, it is still determined to be a "single cell fault".

[0014] Preferably, the location result in step 8 includes the absolute location of the faulty cell, details of abnormal parameters, and a predicted fault type; the absolute location includes the battery pack number, module number, series and parallel connection number, and physical coordinates; the details of abnormal parameters include the type, specific value, and start time of the abnormal characteristic parameters. The location results are presented visually in three ways: local terminal display, host computer display, and alarm push notification. The traceability report includes data snapshots, curve comparison charts, verification process records, and recommended handling plans.

[0015] On the other hand, the present invention provides a battery thermal runaway fault cell parameter tracing system, the system being based on the battery thermal runaway fault cell parameter tracing method, comprising: Trigger module: Used to preset the algorithm start trigger conditions. When the core parameters are detected to meet the trigger conditions, the tracing algorithm is started. Basic data construction and reading module: used to construct and read basic binding data, which includes data on the binding relationship between battery cells and sensors and data on the topology of battery pack; Dual-dimensional data acquisition module: used to simultaneously collect real-time abnormal data and historical baseline data; Data processing and feature extraction module: used to preprocess the two-dimensional data and extract abnormal features; Initial positioning module: used to reverse locate the initial battery cell based on the binding relationship data between the battery cell and the sensor, and the positioning process adopts a dual verification mechanism; Cross-validation module: Used to perform multi-parameter cross-validation on the initially located battery cells to determine whether they are suspected faulty battery cells; Comparison and verification module: Used to compare and verify the parameters of suspected faulty cells with those of adjacent cells, distinguish between single-cell faults and module-level faults, and determine the range of faulty cells; The results output and report generation module is used to output the location results of the faulty battery cell and generate a traceability report.

[0016] The present invention has the following beneficial effects: 1. High positioning accuracy: Through "sensor binding + multi-parameter cross-verification + comparison of adjacent cells", the positioning accuracy of faulty cells is ≥99%, which can be accurate to "specific cells".

[0017] 2. Low false alarm rate: The multi-dimensional verification logic effectively eliminates interference such as "sensor false alarms" and "uneven heat dissipation", with a false alarm rate of ≤1%.

[0018] 3. Improved efficiency: Faulty cells can be located in seconds, greatly reducing the scope of troubleshooting and improving battery repair efficiency by more than 50%.

[0019] 4. Strong data support: The traceability report provides accurate data for battery design optimization and thermal runaway mechanism research, helping to iterate on battery safety. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments. The accompanying drawings, which constitute a part of this application, are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0021] Figure 1 This is a flowchart illustrating the overall execution process of the fault cell location algorithm of this invention.

[0022] This diagram is used to fully present the entire execution process of the faulty battery cell location algorithm, forming a closed loop from parameter anomaly trigger to result output. Key nodes are marked with core thresholds (comprehensive score ≥0.7, confidence level ≥90%, etc.), clearly demonstrating the technical logic of 'multi-parameter fusion, cross-validation, and iterative optimization', ensuring second-level accurate location of faulty battery cells, and providing intuitive support for the algorithm's feasibility.

[0023] Figure 2 This is a schematic diagram of the three-dimensional topological positioning of the battery pack of the present invention.

[0024] This diagram shows the physical layout of the battery pack, modules, and cells in a three-dimensional topological format.

[0025] Table 1 shows the logic of the multi-parameter correlation anomaly mode for thermal runaway faults in this invention.

[0026] This table defines three strongly correlated abnormal modes of thermal runaway faults. Each module clarifies the correspondence between "parameter combination conditions and fault causes". By using multi-parameter collaborative judgment to eliminate misjudgment of a single parameter, it achieves accurate positioning of the core cause of the fault, which is the core embodiment of the "multi-parameter fusion" technology innovation. Detailed Implementation

[0027] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0028] Example 1: Step 1. Preset the algorithm start trigger conditions. When the sensors in the battery management system's perception layer detect that the core parameters meet the trigger conditions, the tracing algorithm is started. When various sensors in the Battery Management System (BMS) sensing layer detect abnormal fluctuations in core parameters such as temperature, impedance, and voltage, the parameter tracing algorithm is automatically triggered. The triggering conditions employ a dual logic of "threshold triggering + trend prediction": if the instantaneous value of a single core parameter exceeds a preset safety range, or if it exhibits an abnormal trend (temperature continuously rising, impedance steadily increasing) for three consecutive acquisition cycles, the algorithm is activated. The core parameters include temperature, impedance, and voltage. This dual-logic triggering mechanism ensures that no precursory signals of thermal runaway are missed, avoiding false alarms caused by a single triggering method.

[0029] Step 2. Construct and read basic binding data, which includes data on the binding relationship between the battery cell and the sensor, and data on the topology of the battery pack; Basic binding data is the core prerequisite for achieving accurate traceability. It includes two key types of information: the "cell-sensor" binding relationship and the battery pack topology. Through solidification in the production process and system storage, full life cycle traceability is achieved.

[0030] Cell-to-sensor binding system: A unique binding relationship is pre-established between each cell and its corresponding monitoring sensor, forming an immutable "cell ID-sensor ID" binding table. The establishment of this binding relationship involves two core stages: Production binding stage: When the battery cells are put into the module, each battery cell is equipped with a dedicated temperature sensor and impedance sensor through automated equipment, and an independent voltage sampling chip channel is allocated. At the same time, an industrial barcode scanner is used to collect the cell ID and sensor ID, and the associated data is uploaded to the battery production MES system in real time. The system automatically verifies the uniqueness of the binding. If duplicate binding or missing binding occurs, the production line alarm is immediately triggered, and manual investigation and correction are required.

[0031] System solidification phase: After the module is installed into the battery pack, the BMS reads the ID information of the sensors in each module through the SPI interface and writes the four-level association relationship of "battery pack number - module number - cell ID - sensor ID" into the built-in EEPROM storage module to form a permanent binding data table to ensure that the data is not lost after power failure. At the same time, it is synchronized to the battery management host computer system for real-time call by the traceability algorithm.

[0032] Battery pack topology definition: Clearly define the series and parallel connection methods, physical arrangement and electrical connection relationships of the cells in the battery pack to form structured topology data. It is necessary to record in detail the number of cells in each string, the physical location coordinates of each cell, the electrical connection relationship of adjacent cells, and the correspondence between the module and the battery pack interface, so as to provide a dual basis of physical space and electrical logic for subsequent fault location.

[0033] The battery pack topology data includes the series and parallel connection method of the cells, their physical arrangement, electrical connections, the number of cells in each string, the physical coordinates of each cell, the electrical connections between adjacent cells, and the correspondence between the module and the battery pack interface. This clear topology data provides both physical and electrical logic basis for fault location, resulting in more accurate fault identification.

[0034] Step 3. Synchronously collect real-time anomaly data and historical baseline data to form a two-dimensional data support; Once the algorithm is started, it simultaneously acquires real-time anomaly data and historical baseline data, forming a two-dimensional data support of "real-time monitoring - historical comparison" to ensure the accuracy of anomaly judgment.

[0035] Real-time anomaly data acquisition: Collects multi-dimensional anomaly parameters output by the sensing layer, covering core indicators of the cell's operating status, specifically including: Temperature parameters: instantaneous surface temperature of the battery cell, rate of temperature increase (temperature change per unit time); Impedance parameters: real and imaginary parts of AC impedance, impedance change amplitude and duration; State of charge (SOC): real-time SOC value, SOC jump amplitude (the degree of matching with the state of charge and discharge). Ionic and electrical parameters: characteristic signals of lithium-ion concentration, amplitude of sudden rise and fall of charging and discharging current, and terminal voltage fluctuation value; Acquisition frequency: Temperature and voltage are acquired once every 100ms, impedance is acquired once every 500ms, and SOC and lithium ion concentration are acquired once every 1s to ensure the capture of instantaneous abnormal signals.

[0036] Historical baseline data retrieval: Retrieve parameter ranges for normal battery cells of the same batch and model under the same operating conditions from the system database to construct a personalized historical baseline. The baseline data must meet the following requirements: Operating condition matching: strictly corresponds to the current working state of the battery. In the fast charging scenario, the baseline data of the same batch of cells at 1.5C charging is retrieved, and in the low temperature scenario, the baseline data at -10℃ is retrieved. Complete parameters: Includes all indicators corresponding to real-time acquired parameters, such as temperature rise rate ≤2℃ / min, impedance fluctuation ≤10mΩ, voltage fluctuation ≤20mV, and SOC stability variation range ±5%. Dynamic updates: Every 100 normal charge-discharge cycles, the historical baseline data is automatically updated to incorporate the parameter change patterns during the cell aging process, avoiding misjudgments caused by cell degradation.

[0037] Step 4. Preprocess the two-dimensional data and extract abnormal features; The collected two-dimensional data needs to be preprocessed before feature extraction to ensure data validity, while clarifying the criteria for judging abnormal features and the extraction logic.

[0038] The preprocessing includes: Filtering and noise reduction: The Kalman filter algorithm is used to process high-frequency fluctuation signals of temperature and voltage, eliminating instantaneous interference (signals with instantaneous voltage jumps >50mV and durations <100ms) and preserving the true parameter change trend; the Kalman filter algorithm has a good filtering effect and can effectively remove interference signals and improve the accuracy of data.

[0039] Outlier correction: When sensor data exceeds the physical reasonable range (temperature > 85℃ or < -20℃, voltage > 4.5V or < 2.5V), the average value of the parameters from the previous 5 acquisition cycles of the cell is automatically used to fill the data and marked as "data abnormal - to be verified". Through outlier correction, data distortion caused by sensor failure or instantaneous anomalies can be avoided, ensuring the reliability of subsequent analysis.

[0040] Data alignment: Parameters from different acquisition frequencies are aligned to a 100ms time axis, and missing data points are supplemented using linear interpolation. Data alignment ensures consistency across the time dimension of multiple parameters, providing a guarantee for multi-parameter fusion analysis.

[0041] The abnormal feature extraction employs a judgment method combining "quantitative threshold + qualitative trend," where a cell is judged as "abnormal" if any two or more of the following characteristics are met: temperature characteristics, impedance characteristics, SOC characteristics, lithium-ion characteristics, current characteristics, and voltage characteristics. Specific feature criteria are as follows: Temperature characteristics: The rate of temperature increase is >5℃ / min, or the temperature continues to rise for 5 consecutive cycles with a cumulative increase of >8℃; Impedance characteristics: Impedance change of more than 50mΩ within 10s, or impedance value that is more than 20% higher than the upper limit of the historical baseline or less than 30% lower than the lower limit of the baseline for 3 consecutive acquisition cycles; SOC characteristics: A jump of ≥10% that is mismatched with the state of charge or discharge, or a SOC change rate that exceeds twice the baseline range for two consecutive cycles; Lithium-ion characteristics: The fluctuation range of lithium-ion concentration characteristic signal is >30%, or obvious intercalation / deintercalation anomaly signal appears; Current characteristics: sudden rise or fall of charging and discharging current >20% (relative to rated current), or continuous exceedance of normal operating current range for 2 cycles; Voltage characteristics: a voltage drop of more than 50mV within 100ms, or a voltage below 3.0V for 5 seconds, or a voltage fluctuation exceeding 3 times the historical baseline.

[0042] By adopting a "quantitative threshold + qualitative trend" judgment method, abnormal cells can be identified more comprehensively and accurately, avoiding the limitations of a single judgment method.

[0043] Step 5. Based on the binding relationship data between the battery cell and the sensor, the initial battery cell is located in reverse through the abnormal sensor ID, and the location process adopts a dual verification mechanism; The dual verification mechanism includes: 1. First-level matching: By using the sensor ID corresponding to the abnormal parameters, the corresponding cell ID is quickly queried in the binding table to achieve preliminary positioning; first-level matching can quickly lock the suspected cell and improve positioning efficiency.

[0044] 2. Secondary Verification: Retrieve data from other sensors bound to the battery cell to confirm whether there are any multi-sensor synchronization anomalies. If only a single sensor is faulty, mark it as "suspected sensor failure." If multiple sensors are faulty, proceed to the next verification step. Secondary verification can effectively eliminate false positioning caused by a single sensor failure, improving positioning accuracy.

[0045] Step 6. Perform multi-parameter cross-verification on the initially located battery cells to determine whether they are suspected faulty battery cells; Multi-parameter cross-validation is performed on the initially located battery cells. The core verification is whether the battery cell meets at least two abnormal characteristics to rule out false alarms from a single sensor.

[0046] The multi-parameter cross-validation includes: Feature statistics: Count the number and specific values ​​of abnormal features that the battery cell meets in the preprocessed data; Weighting determination: Temperature, voltage, and impedance are the core features, accounting for 30% of the weight; SOC, lithium-ion concentration, and current are the auxiliary features, accounting for 20% of the weight. A total weight score of ≥60% (i.e., at least 2 core features, or 1 core feature + 2 auxiliary features) is used to determine a "suspected faulty cell". By assigning different weights to different features, the importance of core parameters can be highlighted and the scientific nature of the determination can be improved.

[0047] Results Processing: If the weighted score requirement is met, the system proceeds to the adjacent cell comparison and verification stage; if not, it is determined to be a "sensor false alarm," the abnormal marker is cleared, and the data is reread. If the cumulative number of sensor false alarms reaches 3, a sensor fault alarm is triggered. This processing method can further eliminate false alarms and provide early warning of sensor faults, ensuring the reliability of the system.

[0048] Step 7. Compare and verify the parameters of the suspected faulty cell with those of adjacent cells to distinguish between single-cell faults and module-level faults, and determine the range of faulty cells; By comparing the parameter differences between suspected faulty cells and adjacent cells, the system can distinguish between "single cell fault" and "module-level fault," thereby further improving the positioning accuracy.

[0049] The adjacent cells include physically adjacent cells and electrically adjacent cells; physically adjacent refers to the 2-4 cells in the same module that are closest to the suspected cell, and electrically adjacent refers to adjacent cells that are in the same series branch or parallel branch as the suspected cell; The acquisition time window for the comparative verification is 3 seconds before and after the suspected cell abnormality start time, and the acquisition frequency is consistent with that of the suspected faulty cell. Single cell fault determination: If only the parameters of the suspected cell meet the abnormal characteristics, and the parameters of all adjacent cells are within the historical baseline range, and the parameter fluctuation range is ≤50% of the baseline range, then it is confirmed as a "single cell fault". Module-level fault determination: If the suspected cell and at least two adjacent cells show abnormal characteristics at the same time, or if the parameters of the adjacent cells do not exceed the baseline range but the fluctuation range is ≥80% of the baseline range, it is determined to be a "module-level fault". The root cause of the fault may be abnormality in the module's connection lines, heat dissipation system or management module. Boundary case handling: If only one adjacent cell shows a slight abnormality with a fluctuation range of 60%-80% of the baseline range, the acquisition time window is extended to 10 seconds, and the parameter change trend is re-compared. If the suspected cell abnormality continues to worsen and the abnormality of adjacent cells does not spread, it is still determined to be a "single cell fault".

[0050] Verification result processing and positioning range adjustment: Based on the results of comparative verification of adjacent battery cells, a differentiated processing strategy was adopted to optimize the scope of the investigation: Single cell fault handling: There is no need to expand the scope of investigation. The cell initially located is identified as the core fault object, and the process proceeds to the location result output stage. Module-level fault handling: Expand the investigation scope to the entire module where the suspected cell is located, and repeat the data collection, abnormal feature extraction, and cross-validation process for all cells in the module to check the faulty cells one by one. If more than 50% of the cells in the module are abnormal, the scope is further expanded to the entire battery pack, and a high-level safety alarm is triggered at the same time.

[0051] By comparing and verifying adjacent cells, it is possible to accurately distinguish between single cell faults and module-level faults, further narrowing the scope of troubleshooting and improving maintenance efficiency.

[0052] Step 8. Output the location results of the faulty battery cell and generate a traceability report.

[0053] It outputs precise location information of faulty battery cells and presents it in a multi-dimensional format to facilitate quick identification and handling by maintenance personnel.

[0054] Output core information: Absolute location: Clearly state the battery pack number, module number, and series / parallel connection number of the faulty cell, and provide physical coordinates; Details of abnormal parameters: Lists the type, specific value, and start time (accurate to the second) of all abnormal characteristic parameters. Fault type prediction: Predict the fault type based on the combination of abnormal characteristics.

[0055] Visual presentation format: Local display: The BMS display screen shows the location diagram of the faulty battery cell, abnormal parameter values ​​and confidence level (location accuracy) in real time; Upper computer display: Displays parameter curve comparison charts and historical baseline comparison trend charts of faulty cells and adjacent cells through a web interface, and marks the time nodes and values ​​of abnormal feature points; Alarm push notification: Push alarm information to the vehicle controller (VCU) or energy storage monitoring platform, including fault location, anomaly type, and risk level (Level 1 risk requires immediate shutdown, Level 2 risk requires power limitation).

[0056] Source tracing report generation: Automatically generates source tracing report segments to provide accurate data support for subsequent fault analysis and battery design optimization. The report includes the following core contents: Data snapshot: Records real-time parameter data for 5 seconds before and after a fault occurs, including instantaneous values ​​and rates of change of parameters such as temperature, impedance, voltage, and SOC; Curve Comparison Chart: An overlay of parameter curves during abnormal periods of embedded faulty cells and historical baseline curves, clearly marking abnormal characteristic points (specific time points and values ​​of sudden temperature rise and impedance change). Verification process record: Detailed record of the entire process data of sensor ID matching, multi-parameter cross-verification, and comparison of adjacent cells, including the scores and judgment criteria for each verification; Recommended solutions: Based on the predicted fault type, provide targeted solutions.

[0057] Example 2: This invention also provides a battery thermal runaway fault cell parameter tracing system, comprising: Trigger module: Used to preset the algorithm start trigger conditions. When the core parameters are detected to meet the trigger conditions, the tracing algorithm is started. Basic data construction and reading module: used to construct and read basic binding data, which includes data on the binding relationship between battery cells and sensors and data on the topology of battery pack; Dual-dimensional data acquisition module: used to simultaneously collect real-time abnormal data and historical baseline data; Data processing and feature extraction module: used to preprocess the two-dimensional data and extract abnormal features; Initial positioning module: used to reverse locate the initial battery cell based on the binding relationship data between the battery cell and the sensor, and the positioning process adopts a dual verification mechanism; Cross-validation module: Used to perform multi-parameter cross-validation on the initially located battery cells to determine whether they are suspected faulty battery cells; Comparison and verification module: Used to compare and verify the parameters of suspected faulty cells with those of adjacent cells, distinguish between single-cell faults and module-level faults, and determine the range of faulty cells; The results output and report generation module is used to output the location results of the faulty battery cell and generate a traceability report.

[0058] Furthermore, the triggering module adopts a dual logic of "threshold triggering + trend prediction": if the instantaneous value of a single core parameter exceeds the preset safety range, or if it shows an abnormal change trend for three consecutive acquisition cycles, the algorithm is activated; the core parameters include temperature, impedance, and voltage.

[0059] The basic data construction and reading module includes a binding relationship construction unit and a topology definition unit; The binding relationship construction unit is used to construct the binding relationship data between the battery cell and the sensor through two stages: production binding and system solidification. The topology definition unit is used to define the series and parallel connection methods, physical arrangement and electrical connection relationships of the cells in the battery pack, forming structured topology data.

[0060] The dual-dimensional data acquisition module includes a real-time data acquisition unit and a historical baseline retrieval unit; The real-time data acquisition unit is used to acquire real-time abnormal data such as temperature parameters, impedance parameters, state of charge, ion and electrical parameters; The historical baseline retrieval unit is used to retrieve the parameter range of normal battery cells of the same batch and model under the same operating conditions from the system database to construct a personalized historical baseline.

[0061] The data processing and feature extraction module includes a preprocessing unit and a feature extraction unit; The preprocessing unit is used to perform filtering and noise reduction, outlier correction, and data alignment on the collected data. The feature extraction unit is used to extract abnormal features from the preprocessed data using a combination of quantitative threshold and qualitative trend judgment method.

[0062] The initial positioning module includes a primary matching unit and a secondary verification unit; The primary matching unit is used to quickly query the corresponding cell ID in the binding table using the sensor ID corresponding to the abnormal parameter to achieve preliminary positioning. The secondary verification unit is used to retrieve data from other sensors bound to the battery cell to confirm whether there is a multi-sensor synchronization anomaly.

[0063] The cross-validation module includes a feature statistics unit, a weight determination unit, and a result processing unit; The feature statistics unit is used to count the number and specific values ​​of abnormal features that the battery cell meets in the preprocessed data; The weight determination unit is used to assign differentiated weights to different abnormal features and determine whether a cell is a suspected faulty cell based on the total weight score. The result processing unit is used to determine, based on the weighted judgment result, whether to proceed to the next step of verification or to determine it as a sensor false alarm.

[0064] The comparison and verification module includes an adjacent cell definition unit, a parameter acquisition unit, and a fault determination unit. The adjacent cell definition unit is used to define the scope of physically adjacent and electrically adjacent cells; The parameter acquisition unit is used to synchronously acquire all monitoring parameters of adjacent cells; The fault determination unit is used to distinguish between single-cell faults and module-level faults based on the parameters of adjacent cells, and to determine the range of faulty cells. The result output and report generation module includes a location information output unit, a visualization unit, and a tracing report generation unit, which are used to output the location results and visualize them in various forms, while generating a tracing report.

[0065] Example 3: The fast-charging scenario for new energy vehicle power batteries was selected as the implementation scenario. This scenario aligns with the core designs of this invention, such as "operating condition matching," "multi-parameter fusion," and "precise binding," and can fully reproduce the entire execution logic of the algorithm. The application target is a square ternary lithium battery pack installed in new energy vehicles, employing a "16 series 18 parallel" topology, with a total of 288 cells. Each cell has a rated voltage of 3.7V and a rated capacity of 200Ah. The operating conditions are 1.5C charging at a fast-charging station (charging current 300A) at an ambient temperature of 25℃. The corresponding historical baseline data consists of the parameter statistics of the same batch of cells under 1.5C fast charging and 25℃ conditions (temperature rise rate ≤ 2℃ / min, impedance fluctuation ≤ 10mΩ, voltage fluctuation ≤ 20mV, SOC stability variation range ±5%). The established relationships have been formed into a four-level association table through production binding and system solidification. For example, the cell ID code is BP-02-MD-03-CELL-12 (battery pack No. 2, module No. 3, cell No. 12), and the corresponding temperature sensor ID is BP-02-MD-03-CELL-12-T, the impedance sensor ID is BP-02-MD-03-CELL-12-Z, and the voltage sampling channel ID is BP-02-MD-03-CELL-12-V. The bound data is stored in the BMS's EEPROM module and synchronized to the host computer.

[0066] During charging, the temperature sensor (BP-02-MD-03-CELL-12-T) corresponding to the BP-02-MD-03-CELL-12 cell detected temperature data of 58℃, 62℃, and 67℃ for three consecutive sampling cycles (each cycle is 100ms). The temperature rise rate reached 6.2℃ / min, exceeding the historical baseline threshold (≤2℃ / min). At the same time, the voltage sensor detected that the terminal voltage dropped from 3.92V to 3.85V, fluctuating by 70mV within 100ms. This meets the dual logic of "trend prediction + threshold triggering", and the algorithm is automatically started.

[0067] The BMS retrieves the binding table stored in the EEPROM via the SPI interface to confirm that the sensor IDs (T / Z / V) corresponding to the abnormal parameters are all associated with cell BP-02-MD-03-CELL-12. At the same time, it reads the battery pack topology data to determine that the cell is located in the 6th row of the 2nd string of module 3. The physically adjacent cells are CELL-11 (left), CELL-13 (right), CELL-02 (top), and CELL-22 (bottom), and the electrically adjacent cells are CELL-11 (series pre-stage) and CELL-13 (series post-stage).

[0068] According to the present invention, real-time abnormal data are collected synchronously at set frequencies. Temperature is collected once every 100ms, and the instantaneous value and ΔT / Δt are recorded. Impedance is collected once every 500ms, and Rs=85mΩ and Rct=62mΩ (impedance change of 72mΩ within 10s) are recorded. Voltage is collected once every 100ms, and the instantaneous terminal voltage value and ΔV=65mV / 100ms are recorded. SOC is collected once every 1s, and the SOC value jumps from 82% to 71% (jump of 11%), and the lithium ion concentration signal fluctuation amplitude is 35%. All collected data are appended with the cell ID and sensor ID. At the same time, baseline data of the same batch of cells under 1.5C fast charging and 25℃ environment are retrieved from the system database, and it is determined that the impedance fluctuation is ≤10mΩ, the voltage fluctuation is ≤20mV, the SOC jump is ≤5%, and the lithium ion concentration fluctuation is ≤30%.

[0069] Kalman filtering (3-cycle window) was used to process the voltage signal, eliminating interference signals with a single instantaneous jump of 60mV lasting 80ms, while retaining the true voltage fluctuation trend. All acquired parameters were within a physically reasonable range (temperature 25℃~85℃, voltage 2.5V~4.5V), requiring no mean padding and without any "data anomaly - pending verification" markings. Impedance (2Hz) and SOC (1Hz) data were aligned to the 100ms time axis using linear interpolation to supplement missing data points and ensure consistency of temperature, voltage, impedance, and other parameters in the time dimension.

[0070] According to the "quantitative threshold + qualitative trend" standard of this invention, the abnormal characteristics of the battery cell are extracted as follows: temperature rise rate 6.2℃ / min > 5℃ / min, which meets the temperature abnormality; impedance change within 10s 72mΩ > 50mΩ, which meets the impedance abnormality; voltage drop within 100ms 65mV > 50mV, which meets the voltage abnormality; SOC jump 11% ≥ 10%, which meets the SOC abnormality; lithium ion concentration fluctuation 35% > 30%, which meets the lithium ion characteristic abnormality; each characteristic is marked with a corresponding code, and the abnormality start time is recorded as 2024 XX Month XX Day 15:48:32.156 (accurate to ms).

[0071] Using the T / Z / V sensor IDs corresponding to the abnormal parameters, the battery cell BP-02-MD-03-CELL-12 is quickly located in the binding table; all sensor data bound to this battery cell are retrieved, and it is confirmed that the temperature, impedance, voltage, SOC, and lithium-ion concentration sensors are all synchronously abnormal, thus eliminating the "single sensor failure" and proceeding to the cross-validation stage.

[0072] The cell met five abnormal characteristics, of which temperature, voltage, and impedance were the core characteristics (3 items), and SOC and lithium-ion concentration were the auxiliary characteristics (2 items). According to the weighting rules of this invention, the core characteristics have a weight of 30% per item, the auxiliary characteristics have a weight of 20% per item, and the total weight score = 3 × 30% + 2 × 20% = 130% ≥ 60%, so it was judged as a "suspected faulty cell" and entered the comparison and verification stage with adjacent cells.

[0073] Parameters of physically adjacent (CELL-11, CELL-13, CELL-02, CELL-22) and electrically adjacent (CELL-11, CELL-13) cells were collected simultaneously. The collection time window was 3 seconds before and after the abnormal start time, and the collection frequency was consistent with that of the suspected faulty cells. The temperature rise rate of all adjacent cells was ≤1.8℃ / min, impedance fluctuation was ≤8mΩ, voltage fluctuation was ≤15mV, and SOC jump was ≤3%, all of which were within the historical baseline range, and the fluctuation amplitude was ≤50% of the baseline range. "Module-level fault" was ruled out, and "single cell fault" was confirmed.

[0074] The absolute location of the faulty cell is clearly identified as battery pack 2, module 3, and cell 12 (physical coordinates X=120mm, Y=85mm, Z=40mm); abnormal parameters include temperature rise rate of 6.2℃ / min, impedance change of 72mΩ, and voltage drop of 65mV / 100ms, with the abnormal start time accurate to milliseconds; the predicted fault type is "internal short circuit" (conforming to the characteristics of Mode 1 of this invention). The BMS display shows a real-time 3D topology map of the battery pack, with the location of the faulty cell marked in red; the host computer web interface displays a comparison graph of the parameter curves of the faulty cell and adjacent cells, marking abnormal feature points; a level 1 risk alarm is pushed to the vehicle controller (VCU), prompting immediate cessation of charging.

[0075] Automatically generate a traceability report, including snapshots of instantaneous values ​​and rates of change of parameters such as temperature, impedance, voltage, and SOC 5 seconds before and after the fault; embed an overlay graph of the fault cell parameter curve and the historical baseline curve, and mark abnormal feature points such as the temperature rise rate of 6.2℃ / min and the impedance mutation of 72mΩ; record detailed data of the entire process of sensor ID matching, multi-parameter cross-validation (total weight score 130%), and comparison with adjacent cells; and provide handling recommendations: immediately stop using the module, disassemble and replace cell No. 12, and check for wear on the internal connection lines of the module.

[0076] In this implementation, the algorithm takes only 0.8 seconds from parameter anomaly trigger to fault location completion, achieving second-level location; the location result is accurate to the specific battery cell, and after disassembly and verification, it was confirmed that the internal short circuit of the battery cell caused the precursor to thermal runaway, which is consistent with the predicted fault type, and the location accuracy reaches 100%; the whole process eliminates interference factors such as sensor false alarms and uneven heat dissipation, with a false judgment rate of 0%; during maintenance, the faulty battery cell is directly locked, without the need to check the entire battery pack, and the maintenance efficiency is improved by 60% compared with the existing technology, which fully meets the beneficial effects described in this invention.

[0077] Table 1. Logic of Multi-parameter Correlation Anomaly Mode in Thermal Runaway Fault of the Invention

[0078] Table 1 is a logic diagram of multi-parameter correlation anomaly modes in thermal runaway faults, defining three types of strongly correlated anomaly modes for thermal runaway faults. Each mode clearly defines the correspondence between "parameter combination conditions and fault causes": Mode 1: Internal short circuit fault, triggered by temperature rise rate T_RATE>5℃ / s, voltage drop V_DROP>0.3V / 100ms, impedance change Z_CHANGE>20%, and the fault is caused by an internal short circuit in the cell. Mode 2: Polarization failure type fault, triggered by the following conditions: imaginary impedance Z_IMAG>50mΩ, low voltage duration V_LOW_DUR>5s (voltage<2.8V), and absolute temperature T_ABS>60℃. The fault is caused by cell polarization attenuation. Mode 3: Thermal propagation type fault, triggered by adjacent cell temperature rise rates >3℃ / s and abnormal time difference between adjacent cells <2s, with the fault cause being inter-cell thermal diffusion. Multi-parameter collaborative judgment eliminates false positives based on a single parameter, achieving precise localization of the core fault cause.

[0079] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the disclosure of this invention should fall within the protection scope of this invention.

Claims

1. A method for tracing the parameters of a battery cell in thermal runaway fault, characterized in that, Includes the following steps: Step 1: Preset the algorithm start trigger conditions. When the sensors in the battery management system's perception layer detect that the core parameters meet the trigger conditions, the tracing algorithm is started. Step 2: Construct and read basic binding data, which includes binding relationship data between battery cells and sensors, and battery pack topology data; Step 3: Simultaneously collect real-time anomaly data and historical baseline data to form a two-dimensional data support; Step 4: Preprocess the two-dimensional data and extract abnormal features; Step 5: Based on the binding relationship data between the battery cell and the sensor, the initial battery cell is located in reverse through the abnormal sensor ID, and the location process adopts a dual verification mechanism; Step 6: Perform multi-parameter cross-verification on the initially located battery cells to determine whether they are suspected faulty battery cells; Step 7: Compare and verify the parameters of the suspected faulty cell with those of adjacent cells to distinguish between single-cell faults and module-level faults, and determine the range of faulty cells. Step 8: Output the location results of the faulty battery cell and generate a traceability report.

2. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The triggering conditions described in step 1 employ a dual logic of threshold triggering and trend prediction. If the instantaneous value of a single core parameter exceeds the preset safety range, or if multiple consecutive acquisition cycles show an abnormal trend, the algorithm will be activated; the core parameters include temperature, impedance, and voltage.

3. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The cell-sensor bonding data mentioned in step 2 is constructed in the following way: Step 2.1, Production Binding Stage: When the battery cell enters the module, each battery cell is equipped with a dedicated temperature sensor and impedance sensor, and an independent voltage sampling chip channel is allocated. The cell ID and sensor ID are collected and uploaded to the battery production MES system, and the system verifies the uniqueness of the binding. Step 2.2, System Solidification Stage: After the module is installed into the battery pack, the BMS reads the sensor ID information through the SPI interface, writes the four-level association relationship of "battery pack number - module number - cell ID - sensor ID" into the EEPROM storage module, and synchronizes it to the host computer system at the same time; The battery pack topology data includes the cell series and parallel connection method, physical arrangement, electrical connection relationship, number of cells in each string, physical location coordinates of each cell, electrical connection relationship between adjacent cells, and correspondence between the module and the battery pack interface.

4. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The real-time abnormal data mentioned in step 3 includes temperature parameters, impedance parameters, state of charge (SOC), and ion and electrical parameters. The temperature parameters include the instantaneous value of the cell surface temperature and the rate of temperature rise. The impedance parameters include the real and imaginary parts of the AC impedance, the impedance change amplitude, and the duration of the change. The state of charge (SOC) includes the real-time value and the SOC jump amplitude. The ion and electrical parameters include the lithium-ion concentration characteristic signal, the amplitude of sudden rises and falls in charging and discharging current, and the terminal voltage fluctuation value. The real-time abnormal data collection frequency is as follows: temperature and voltage are collected once every 100ms~150ms, impedance is collected once every 500ms~600ms, and SOC and lithium ion concentration are collected once every 1s~2s. The historical baseline data mentioned in step 3 refers to the parameter range of normal cells of the same batch and model under the same operating conditions, and meets the requirements of operating condition matching, parameter integrity, and dynamic updating; the historical baseline data is automatically updated every 100 to 150 normal charge and discharge cycles.

5. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The preprocessing described in step 4 includes: Step 4.1, Filtering and Denoising: The Kalman filter algorithm is used to process the high-frequency fluctuation signals of temperature and voltage to eliminate instantaneous interference; Step 4.2, outlier correction: When sensor data exceeds the physical reasonable range, the average value of the parameters of the previous 5 to 6 collection cycles of the cell is automatically used to fill the data and marked as "data abnormal - to be verified"; Step 4.3, Data Alignment: Align parameters from different acquisition frequencies to a 100ms time axis and supplement missing data points using linear interpolation. The abnormal feature extraction in step 4 adopts a judgment method that combines "quantitative threshold + qualitative trend". If any two or more features are satisfied, it is judged as an "abnormal cell". The abnormal features include temperature features, impedance features, SOC features, lithium ion features, current features, and voltage features.

6. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The dual verification mechanism described in step 5 includes: Step 5.1, Level 1 matching: Quickly query the corresponding cell ID in the binding table using the sensor ID corresponding to the abnormal parameters to achieve preliminary positioning; Step 5.2, Secondary Verification: Retrieve data from other sensors bound to the battery cell to confirm whether there is a multi-sensor synchronization anomaly. If only a single sensor is abnormal, mark it as "suspected sensor failure". If multiple sensors are abnormal, proceed to the next verification step.

7. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The multi-parameter cross-validation mentioned in step 6 includes: Step 6.1, Feature Statistics: Count the number and specific values ​​of abnormal features that the battery cell meets in the preprocessed data; Step 6.2, weight determination: temperature, voltage, and impedance are the core features, with a weight of 30%; SOC, lithium-ion concentration, and current are the auxiliary features, with a weight of 20%. If the total weight score is ≥60%, it is determined as a "suspected faulty cell". Step 6.3, Result Processing: If the weighted score requirement is met, proceed to the adjacent cell comparison and verification stage; if not, it is determined to be a "sensor false alarm", the abnormal mark is cleared and the data is reread. If the sensor false alarms accumulate to 3 times, a sensor fault alarm will be triggered.

8. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The adjacent cells mentioned in step 7 include physically adjacent cells and electrically adjacent cells; physically adjacent refers to the 2-4 cells in the same module that are closest to the suspected cell, and electrically adjacent refers to adjacent cells that are in the same series branch or parallel branch as the suspected cell; The acquisition time window for the comparative verification is 3 seconds before and after the suspected cell abnormality start time, and the acquisition frequency is consistent with that of the suspected faulty cell. Single cell fault determination: If only the parameters of the suspected cell meet the abnormal characteristics, and the parameters of all adjacent cells are within the historical baseline range, and the parameter fluctuation range is ≤50% of the baseline range, then it is confirmed as "single cell fault". Module-level fault determination: If a suspected cell and at least two adjacent cells exhibit abnormal characteristics at the same time, or if the parameters of adjacent cells do not exceed the baseline range but the fluctuation range is ≥80% of the baseline range, then it is determined to be a "module-level fault". Boundary case handling: If only one adjacent cell shows a slight abnormality with a fluctuation range of 60%-80% of the baseline range, the acquisition time window is extended to 10 seconds, and the parameter change trend is re-compared. If the suspected cell abnormality continues to worsen and the abnormality of adjacent cells does not spread, it is still determined to be a "single cell fault".

9. The method for tracing the parameters of a battery cell in thermal runaway fault according to claim 1, characterized in that, The location results in step 8 include the absolute location of the faulty cell, details of abnormal parameters, and a predicted fault type. The absolute location includes the battery pack number, module number, series and parallel connection number, and physical coordinates. The details of abnormal parameters include the type, specific value, and start time of the abnormal characteristic parameters. The location results are presented visually in three ways: local terminal display, host computer display, and alarm push notification. The traceability report includes data snapshots, curve comparison charts, verification process records, and recommended handling plans.

10. A battery thermal runaway fault cell parameter tracing system, characterized in that, The system is based on the battery thermal runaway fault cell parameter tracing method according to any one of claims 1-9, including: Trigger module: Used to preset the algorithm start trigger conditions. When the core parameters are detected to meet the trigger conditions, the tracing algorithm is started. Basic data construction and reading module: used to construct and read basic binding data, which includes data on the binding relationship between battery cells and sensors and data on the topology of battery pack; Dual-dimensional data acquisition module: used to simultaneously collect real-time abnormal data and historical baseline data; Data processing and feature extraction module: used to preprocess the two-dimensional data and extract abnormal features; Initial positioning module: used to reverse locate the initial battery cell based on the binding relationship data between the battery cell and the sensor, and the positioning process adopts a dual verification mechanism; Cross-validation module: Used to perform multi-parameter cross-validation on the initially located battery cells to determine whether they are suspected faulty battery cells; Comparison and verification module: Used to compare and verify the parameters of suspected faulty cells with those of adjacent cells, distinguish between single-cell faults and module-level faults, and determine the range of faulty cells; The results output and report generation module is used to output the location results of the faulty battery cell and generate a traceability report.

Images