Battery testing method and system

Abstract

The invention relates to the technical field of electrical testing, in particular to a battery testing method and system, and the method comprises the steps: obtaining a voltage data sequence and historical log data of each cell in a to-be-tested battery pack; determining a dynamic instability score of each battery cell according to the parameter volatility index of each battery cell and the spatial correlation weight between each battery cell and the adjacent battery cell; on the basis of the historical log data, analyzing a repeated cumulative effect of all historical damage events corresponding to each battery cell, and determining a historical damage degree of each battery cell; performing deep fusion on the dynamic instability score and the historical damage degree through a nonlinear function to obtain a comprehensive risk index of each battery cell; and based on the extreme values and distribution of the comprehensive risk indexes of all the cells, determining a fault degree capable of representing the overall health risk of the battery pack, and determining a health test result of the battery pack according to the fault degree. According to the method, the accuracy and reliability of battery pack testing are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical testing technology. Specifically, it relates to a battery testing method and system. Background Technology

[0002] In the testing of new energy vehicles, conducting accurate and in-depth health assessments of the power battery pack is a core step in ensuring the long-term safety and performance of the vehicle.

[0003] Currently, portable diagnostic equipment is commonly used to communicate with the vehicle's battery management system (BMS) to perform a rapid test on the battery pack based on a snapshot mode. This test method captures various static parameters of the battery pack at a certain moment within a few seconds, such as cell voltage and temperature, and compares them with standard thresholds to determine whether the various static parameters of the battery pack are normal, thereby determining the health status of the battery pack.

[0004] However, during long-term vehicle use, battery packs often exhibit cumulative potential abnormalities that cannot be promptly identified by snapshot-based rapid testing. For example, repeated irregular charging or high-load discharging of the battery pack can continuously induce trace amounts of lithium deposition on the surface of the negative electrode of the cells. The changes in battery parameters caused by a single deposition are extremely minor. Over time, these deposited lithium metals accumulate and grow into dangerous lithium dendrites, gradually piercing the internal separator and forming hidden internal micro-short circuits. For this cumulative damage that progresses from quantitative to qualitative change, instantaneous snapshot-based testing methods cannot identify this gradual abnormal process until the micro-short circuit develops into a severe internal short circuit, at which point the instantaneous snapshot-based testing method can identify the abnormal state of the battery pack. In summary, traditional instantaneous snapshot-based testing methods are lagging in identifying the cumulative potential abnormalities of battery packs caused by long-term use, resulting in inaccurate and incomplete test results. Summary of the Invention

[0005] To address the problem that existing technologies using instantaneous snapshot testing modes cannot detect potential cumulative anomalies in battery packs in a timely manner, resulting in inaccurate and incomplete test results, this invention proposes a battery testing method and system.

[0006] In a first aspect, the present invention provides a battery testing method, comprising: Obtain the voltage data sequence of each cell in the battery pack under test and the historical log data of the battery pack; Based on the voltage data sequence of each cell, the parameter volatility index of each cell is calculated to characterize the degree of fluctuation in the voltage data of each cell. Based on the physical topology of the battery pack, all adjacent cells of each cell are identified. Using the parameter volatility index of each cell and all adjacent cells, the spatial correlation weight between each cell and all adjacent cells is calculated. Combining the parameter volatility index and spatial correlation weight of each cell, the dynamic instability score of each cell is calculated. Historical log data is analyzed to extract all historical damage events corresponding to each cell. By analyzing the repeated cumulative effect of all historical damage events, the historical damage level of each cell is determined. The dynamic instability score and historical damage level of each cell are integrated to generate a comprehensive risk index for each cell. By integrating the comprehensive risk index of all battery cells, the degree of failure of the battery pack is calculated, and the health status of the battery pack is determined based on the comparison result of the degree of failure with the preset failure degree threshold, so as to complete the battery test.

[0007] This technical solution first acquires high-frequency time-series dynamic data and historical log data, constructing a three-dimensional data foundation for diagnosis that includes both micro-dynamics and macro-history. Based on this solid foundation, the solution does not stop at simple numerical comparisons, but further refines the raw data into insights with clear physical meaning. On the one hand, through time-series volatility and spatial correlation analysis, it transforms the rapidly changing voltage signal into a dynamic instability score that can distinguish between isolated noise and systemic jitter, which is equivalent to accurately quantifying the current functional symptoms of the battery. On the other hand, through a decay accumulation model, it transforms scattered historical logs into historical damage measurements that measure the degree of cumulative damage, which is equivalent to quantifying the substantial damage to the battery. Through nonlinear fusion, it comprehensively evaluates the current health status in the context of its historical health status, generating a comprehensive risk index for individual cells. This simulates the physical law that damaged components are more prone to failure under disturbances. By elevating all local diagnostic information to the final judgment of the overall safety of the battery pack, it ensures that no single fatal defect or widespread problem can escape detection, thus completing an accurate diagnosis of the battery pack's health status within the limited time of production line testing.

[0008] Preferably, the voltage data sequence of each cell in the battery pack under test is obtained, including: A test time window and a sampling frequency are preset. Within the test time window, the voltage of each cell is continuously collected according to the sampling frequency to form a voltage data sequence of each cell.

[0009] As a preferred method, the parameter volatility index of each cell is determined as follows: For any cell, the voltage data sequence of the cell is subjected to first-order difference processing to obtain a first-order difference sequence; the absolute value of all data in the first-order difference sequence is first calculated, and then the average value is calculated, and the average value is used as the parameter volatility index of the cell.

[0010] As a preferred embodiment, the spatial association weight between each cell and all adjacent cells is obtained by the following calculation formula:

[0011] in, For the first Spatial association weights between each cell and all its adjacent cells For cell numbering, For the first The parameter volatility index of individual battery cells For the first The total number of all adjacent cells of a given cell. For the first The first cell The parameter volatility index of adjacent cells, A preset positive number used to prevent the denominator from being zero. It is the absolute value symbol.

[0012] This technical solution assesses the synchronicity of the parameter fluctuation behavior of a single cell with its physical neighboring cell group. When the parameter fluctuation behavior of a cell is highly consistent with that of its neighboring cells, it indicates that there is more likely to be a regional event around the cell. Conversely, the cell is more likely to be an isolated local anomaly. This can effectively distinguish whether the root cause of the fault is an isolated problem or a systemic risk.

[0013] As a preferred method, the dynamic instability score of each cell is obtained using the following formula:

[0014] in, For the first Dynamic instability score of individual battery cells For the first The parameter volatility index of individual battery cells For the first Spatial association weights between each cell and all its adjacent cells It is a sine function.

[0015] This technical solution aims to calculate a dynamic instability score that can intelligently distinguish risk types based on the fluctuation level of the battery cell itself and its linkage with neighboring battery cells. Through a smooth nonlinear function, it achieves a significant amplification of systemic failure risk while tolerating isolated noise fluctuations, so that the final score can more accurately reflect the true danger level of different types of fluctuations.

[0016] Preferably, the historical damage level of each battery cell is determined by analyzing the cumulative effect of repeated historical damage events. This includes: for each battery cell, selecting all historical damage events corresponding to that cell from historical log data; for each historical damage event, extracting the basic severity weight of that historical damage event, the total number of times that historical damage event has occurred among all historical damage events, and the time interval between the most recent occurrence of that historical damage event and the current testing time; using an exponential decay function, determining the time decay coefficient of that historical damage event based on the time interval; multiplying the basic severity weight, the total number of occurrences, and the time decay coefficient to obtain the weighted damage level of that historical damage event; and summing the weighted damage levels of all historical damage events to obtain the historical damage level of the battery cell.

[0017] This technical solution successfully transforms discrete historical damage log data in the BMS into a quantitative historical damage level that accurately reflects the cumulative damage level of the battery cell by constructing a decay accumulation model for a single battery cell that integrates event severity, occurrence frequency, and time decay effects. This provides a clearly physical input for subsequent risk fusion.

[0018] Preferably, the time decay coefficient of historical damage events is determined based on the following method: For the The first cell Historical damage events The time decay coefficient is , It is a natural exponential function. The preset time decay constant, It is the first The first cell The time interval between the most recent occurrence of a historical damage event and the time being tested.

[0019] As a preferred approach, the comprehensive risk index for each battery cell is generated by integrating the dynamic instability score and historical damage level of each cell, based on the following formula: ; in, For the first The comprehensive risk index of each battery cell For the first Dynamic instability score of individual battery cells This represents the maximum value of the dynamic instability score for all battery cells. It is the hyperbolic tangent function. It is a natural exponential function. For the first The extent of historical damage to each battery cell This represents the average historical damage level of all battery cells. This is a parameter used to prevent the denominator from being 0.

[0020] This technical solution utilizes the statistical characteristics of the current data of the battery pack for dynamic self-calibration, thereby improving the accuracy and robustness of the test. This design enables the evaluation system to adaptively distinguish between two completely different situations: one is the cell exhibiting abnormal indicators in a battery pack with good overall health, and the other is the cell exhibiting similar indicators in a battery pack with general aging. The logic of combining the individual state of a single cell with the background of the battery pack it belongs to make a comprehensive judgment can more accurately identify abnormal cells that are truly at risk.

[0021] Preferably, the failure level of the battery pack is determined as follows: the maximum value among the comprehensive risk indices of all cells is taken as the extreme risk index; the number of cells whose comprehensive risk index exceeds the preset comprehensive risk index threshold is counted, and the proportion of this number of cells in the total number of cells in the battery pack is non-linearly amplified to obtain a systematic risk index; the extreme risk index is multiplied by the systematic risk index to obtain the failure level of the battery pack.

[0022] This technical solution uses an aggregation model to elevate the local risk information of all battery cells into a fault degree that represents the overall health status of the battery pack. This model not only focuses on the weakest link (extreme risk index) determined by the barrel theory, but also non-linearly quantifies the systemic collapse risk brought about by the breadth of risk spread (the number of battery cells whose comprehensive risk index exceeds the preset comprehensive risk index threshold). This allows the test results to take into account both the severity of local problems and the universality of the problems, making them more comprehensive and accurate.

[0023] In a second aspect, the present invention also provides a battery testing system, the battery testing system including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of any of the battery testing methods described above.

[0024] The present invention has the following effects: This invention deeply mines the historical damage events of each cell by constructing a decay accumulation model. By nonlinearly fusing the quantified historical damage with the current dynamic instability analyzed by high-frequency sampling, a forward-looking comprehensive risk index is generated. This index can identify cumulative potential anomalies such as internal micro-short circuits caused by long-term complex operating conditions, thereby improving the accuracy and comprehensiveness of battery test results. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a diagram of the portable testing device of the present invention; Figure 3 This is a schematic diagram showing the distribution of the comprehensive risk index of each cell in the battery pack of the present invention; Figure 4 This is a schematic diagram of the health test results of the battery pack of the present invention; Figure 5 This is a schematic diagram showing the distribution of the failure level of the battery pack sample of the present invention. Detailed Implementation

[0026] 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.

[0027] Reference Figure 1 A battery testing method, specifically including the following steps: S1: Obtain the voltage data sequence and historical log data of each cell in the battery pack under test.

[0028] By communicating briefly with the battery pack under test, high-frequency time-series data for subsequent dynamic instability analysis and deep log data for historical damage event assessment are collected simultaneously, providing an accurate data foundation for building a comprehensive testing process for the battery pack.

[0029] Since a power battery pack is a complex system composed of hundreds or thousands of independent cells connected in series and parallel, its overall performance and safety status essentially depend on the consistency and health status of each individual cell. Therefore, it is crucial to delve into the cell level and obtain the most refined raw data. This process is performed at the battery testing station on the automotive assembly line, and is specifically implemented as follows: First, establish reliable communication links at the physical and logical levels, using methods such as... Figure 2 The portable testing equipment shown has a portable testing terminal with a standardized battery pack-specific interface. It connects to the diagnostic CAN (Controller Area Network) bus of the battery pack through a physical interface. The testing terminal then initiates a communication handshake on the bus to establish a diagnostic session with the BMS (Battery Management System) of the entire battery pack under test. The BMS is responsible for monitoring and managing each cell in the battery pack. Through communication with the BMS, the data of all underlying cells can be accessed.

[0030] Next, in the established diagnostic session, two types of data acquisition for each battery cell are performed in parallel: The first category is the acquisition of high-frequency dynamic voltage data sequences of each cell. This part of the data acquisition is limited to a preset test time window and a sampling frequency. In this embodiment, the test time window is set to 3 seconds and the sampling frequency is set to 500Hz. The BMS will continuously report the voltage values ​​of all individual cells it monitors through the CAN bus according to the test time window and the sampling frequency to obtain the voltage data sequence of each cell.

[0031] The test time window was set at 3 seconds because it is an engineering choice that achieves the best balance between production line cycle time and data validity. On the one hand, the operation time of each station on the automotive assembly line is usually strictly controlled within tens of seconds, and a 3-second test duration is sufficient to fit into this cycle time without creating a production bottleneck. On the other hand, at a sampling rate of 500Hz, 3 seconds is enough to collect 1500 data points for each cell. This sample size is sufficient for subsequent calculations of statistical characteristics such as volatility, ensuring the statistical significance and stability of the analysis results. The sampling frequency was set at 500Hz to identify millisecond-level intermittent faults occurring in a single cell or its connection point, caused by transient contact failures such as loose wiring harness connections or terminal micro-vibrations. The voltage jump duration at the cell level is usually in the range of 1-10 milliseconds. In order to accurately capture events as short as 2ms (equivalent to a signal frequency of 500Hz), a sampling rate of 500Hz ensures that at least one sampling point falls within the fault range, thus leaving an analyzable trace on the data sequence of the cell and avoiding missing key fault information due to insufficient sampling rate.

[0032] The second category is in-depth reading of historical log data related to each cell. Within the same time window for collecting high-frequency data, the test terminal will send a series of diagnostic commands to the BMS in parallel to read the historical information associated with specific cells that is stored in its internal non-volatile memory. This includes reading all historical fault records, each of which reflects a historical damage event that occurred in the cell. The specific cell number is located through the additional information of the fault code. Specific data identifiers are read, which point to records of extreme operating conditions experienced during the battery pack's life cycle, such as: the historical highest / lowest single cell voltage and its corresponding cell number, the historical highest / lowest cell temperature and its location, etc.

[0033] Through the above operations, a three-dimensional dataset containing the current micro-dynamics and past macro-experiences was successfully established for each cell in the battery pack during a brief production line test. This provides a reliable data foundation for all subsequent steps of risk quantification for individual cells and final assessment of the overall health status of the battery pack.

[0034] S2: Calculate the dynamic instability score of each cell based on the voltage data sequence of each cell.

[0035] This step aims to generate a score index reflecting the connection stability of each cell from high-frequency timing dynamic data through a series of calculations. Specifically, it includes: S21: Determine the parameter volatility index for each cell.

[0036] The goal is to transform a dynamically changing voltage time series containing thousands of data points into a single numerical index that can stably measure the severity of its fluctuations. Specifically, for any given cell, the voltage data series of the cell is subjected to first-order difference processing to obtain a first-order difference sequence. The absolute values ​​of all data in the first-order difference sequence are first calculated, and then the average of these absolute values ​​is calculated. The final average value is used as the parameter volatility index of the cell.

[0037] For example, collecting five voltage data points from a battery cell within a short period of time forms the following voltage data sequence (unit: volts): The voltage data sequence is then subjected to first-order differencing, which involves subtracting the previous data from the next data point to obtain the first-order differencing sequence. The absolute value of all data in the first-order difference sequence is calculated, and then the average value is calculated to obtain 0.02 as the parameter volatility index.

[0038] The parameter volatility index of each cell quantifies the overall instability of the voltage of each cell during the test. A cell with a stable physical connection and healthy chemical state should have a first-order differential sequence that is close to zero and a parameter volatility index that is extremely low. Conversely, any momentary poor contact caused by connector micro-motion, loose wiring harness connection, etc., will manifest as spike pulses on the differential sequence, thereby raising the parameter volatility index. Therefore, the magnitude of the parameter volatility index is directly inversely proportional to the stability of the cell connection and is the cornerstone of subsequent risk assessment.

[0039] S22: Analyze the spatial association weights between each cell and all adjacent cells.

[0040] After obtaining the volatility index of each cell, we further analyze whether the voltage fluctuation of the cell is a problem unique to that cell or a localized systemic problem.

[0041] Before analysis, it is necessary to identify all adjacent cells for each cell, because many systemic faults tend to cluster locally at the physical level. For example, if the data acquisition cable or high-voltage connector of a single battery module becomes loose, it will directly affect all the cells in that module; or if a busbar develops a micro-crack, it will affect the cells connected to it.

[0042] Therefore, by defining physically adjacent cells based on the physical topology of the battery pack (such as module division, wiring layout, and series-parallel topology), an analysis framework that matches potential fault modes can be constructed, enabling the algorithm to accurately capture the regional fault characteristics that occur in large areas.

[0043] Since all physically adjacent cells of each cell are pre-configured in the test system as prior knowledge, for each type of battery pack under test, the internal cell topology needs to be digitized in advance according to the design blueprint (such as CAD file). This is usually stored as a cell adjacency table (which can be a JSON, XML, CSV file or a database table). This table clearly maps each cell ID to a list of all physically adjacent cell IDs, thus obtaining all the adjacent cells of each cell.

[0044] For example, using JSON format, a small portion of the cell adjacency table is as follows: {"Cell-001":["Cell-002"],"Cell-002":["Cell-001","Cell-003"],"Cell-105":["Cell-104","Cell-106","Cell-115"]}; In this list, Cell is the prefix for cell number. From this list, we can see that all physically adjacent cells of Cell-001 are numbered Cell-002; all physically adjacent cells of Cell-002 are numbered Cell-001 and Cell-003; and all physically adjacent cells of Cell-105 are numbered Cell-106 and Cell-115.

[0045] Before the test begins, the test software loads an adjacency table that matches the model of the battery pack under test. When calculating the spatial association weight between any cell and all its adjacent cells, the software only needs to query this table to instantly obtain the set of cell numbers of all physically adjacent cells. This process is a one-time configuration task, which ensures the accuracy, efficiency and repeatability of subsequent calculations.

[0046] After determining the set of cell numbers of all physically adjacent cells for each cell, the spatial association weight between each cell and all adjacent cells is calculated using the following formula:

[0047] in, For the first Spatial association weights between each cell and all its adjacent cells For cell numbering, For the first The parameter volatility index of individual battery cells For the first The total number of all adjacent cells of a given cell. For the first The first cell The parameter volatility index of adjacent cells, To prevent the denominator from being zero, a preset positive number is set as follows: This is a very small positive number, close to 0, and is typically used for zero-prevention parameters. It is the absolute value symbol.

[0048] The numerator of this formula reflects the first The consistency of electrical characteristics between a cell and its neighboring cells is determined by the numerator. The closer the numerator is to zero, the more consistent the behavior of the cell is with its neighboring cells. Whether the group is stable or deteriorating, the better the consistency of electrical characteristics. Conversely, the larger the numerator is than zero, the more inconsistent the electrical characteristics are with the cell and its neighboring cells.

[0049] The denominator of the formula represents the overall parameter volatility index within the local microenvironment formed by the battery cell and adjacent battery cells. It provides a dynamic and adaptive evaluation scale for the numerator. Its core function is to eliminate the interference of absolute fluctuation magnitude and realize contextualized relative evaluation.

[0050] The core design logic of this formula lies in the fractional part, which is a normalized dimensionless index between 0 and 1. Based on this, the fractional part is transformed by subtracting the overall structure of the fraction from 1, and finally the spatial association weight is obtained. Through this transformation, an index that measures difference is reversed into an index that measures the degree of association more intuitively. The value of the spatial association weight is dynamically adjusted between 0 and 1. The larger the value of the spatial association weight, the better the consistency of the electrical characteristics of a single cell with its neighboring cells, and vice versa.

[0051] When a systemic event occurs, such as a loose connector in the entire module, the voltage of all cells in the module will fluctuate synchronously. At this time, the parameter volatility index of any cell is highly consistent with the average parameter volatility index of its neighboring cells, causing the numerator of the fraction to approach 0, thus making the fraction also approach 0. The greater the spatial correlation weight between a single cell and all its neighboring cells, the greater the spatial correlation weight. Conversely, when an isolated fault occurs, such as a poor solder joint on only a single cell, the parameter volatility index of that cell will be very high, while the parameter volatility index of its neighboring cells will remain stable at a low level. At this time, the numerator of the fraction is almost equal to the denominator, causing the fraction to approach 1, and the smaller the spatial correlation weight between a single cell and all its neighboring cells.

[0052] S23: Calculate the dynamic instability score for each cell.

[0053] The purpose of this step is to integrate the information from two dimensions—parameter volatility index and spatial correlation weight—for each cell into a final risk score. The magnitude of the risk depends not only on the severity of the fault but also on the type of fault. Systemic faults have a much greater impact on the battery pack than isolated faults and should be given a higher dynamic instability score.

[0054] The dynamic instability score for each cell is calculated using the following formula:

[0055] in, For the first Dynamic instability score of individual battery cells For the first The volatility index of an individual battery cell represents a fundamental risk factor, reflecting how drastic the voltage fluctuations of that cell are. The more drastic the fluctuations, the higher the risk level will be identified, ensuring that the risk of failure of a single battery cell is not overlooked. For the first Spatial association weights between each cell and all its adjacent cells It is a sine function.

[0056] In this formula, This part constitutes a nonlinear risk amplification factor, which dynamically changes according to the failure mode (by...). The penalty weight should be determined based on the reflection of the risk level. exist Within this range, the degree to which risk is amplified is directly controlled. Its function is to of The range of values ​​is precisely mapped to the most variable range of the sine function. Within the domain, the sine function is monotonically increasing, and its slope smoothly transitions from maximum to minimum, which matches the nonlinear characteristics required by this scheme. Within the interval, the output value of the sine function smoothly increases from 0 to 1. When increasing from 0, the sine function curve is at its steepest point. This means that even if a cell has only a slight correlation with all its neighboring cells, the nonlinear risk amplification factor will be disproportionately increased to issue a strong early warning signal for systemic failure risks that are nascent and may develop into major problems. When the value approaches 1, for example, from 0.8 to 1, the curve of the sine function becomes very flat, and its slope approaches 0. This means that when a certain cell has a stable strong correlation with all its neighboring cells, the influence on the amplification factor becomes negligible, and eventually it smoothly converges to the maximum value of 2. This makes the model very robust when facing extreme systematic failures, avoiding drastic jumps or unbounded growth in scores due to small data disturbances, and ensuring the reliability of the algorithm in industrial environments. Part of it is the sine function The output range is shifted upwards overall, thus strictly limiting the range of the final nonlinear risk amplification factor to within a certain range. This means that the minimum value of a battery cell's dynamic risk score is the parameter volatility index of that battery cell, and the maximum value is the parameter volatility magnified by 100%.

[0057] Thus, when a single point of failure occurs in the battery pack, meaning a single cell is abnormal, Approaching 0, the sine function part is 0, and the amplification factor is... , This approach avoids amplifying the risk of individual cell failures, assessing risk solely based on the volatility of each cell's parameters. When a battery pack experiences a systemic failure, it can lead to anomalies in numerous cells. Approaching 1, the sine function part is 1, and the amplification factor is... , This amplifies the risk of such systemic failures.

[0058] In summary, this step utilizes the nonlinear characteristics of the sine function to weight the spatial correlation. When the spatial correlation weight changes from 0 (isolated fault) to 1 (systematic fault), the amplification factor smoothly and nonlinearly increases from 1 to 2, thereby giving a higher risk score to the systematic fluctuations that occur in the whole area.

[0059] S3: Analyze the repeated cumulative effect of historical damage events of each cell based on historical log data to determine the degree of historical damage of each cell.

[0060] After analyzing the dynamic instability scores of each cell within the battery pack, historical operating data for each cell is further introduced to assess the true risk level of these dynamic instabilities under a specific historical cumulative stress background. The core point is that, for the same dynamic instability score, a cell with accumulated historical damage has a higher risk of failure than one without. Therefore, by analyzing the repeated cumulative effects of historical damage events in each cell, the degree of historical damage to each cell is determined, including: S31: Extract historical damage events and related statistical characteristics for each cell.

[0061] For any given battery cell, first filter out all historical damage events directly related to that battery cell from the historical log data. Then, for each of the filtered historical damage events, extract the basic severity weight of that historical damage event, the total number of times that historical damage event has occurred in all historical damage events of that battery cell, and the time interval between the most recent occurrence of that historical damage event and the current testing time.

[0062] Specifically, typical historical damage events include: Damage caused by instantaneous overcharging of a single cell: Excessive voltage directly damages the crystal structure of the positive electrode material and triggers the oxidative decomposition of the electrolyte, generating gas and increasing internal pressure. This is irreversible damage, permanently reducing battery capacity and safety. Because overcharging damages the cell directly and rapidly, and is one of the direct causes of thermal runaway, it poses the greatest direct threat to the battery pack; therefore, it is assigned the highest basic severity weight: 0.6.

[0063] Damage events caused by high-load discharge: High-current discharge generates enormous heat and mechanical stress. Over time, this can lead to the breakage of active material particles and their peeling off from the current collector, resulting in cumulative potential abnormal states. Because this event is a key factor accelerating battery cycle aging, its damage is complex. Its weight is moderate because it represents a significant loss to battery life and performance. However, its directness in triggering sudden safety accidents is generally lower than that of overcharge events; therefore, it is assigned a moderate base severity weight: 0.3.

[0064] Damage caused by excessive temperature: High temperatures accelerate the excessive thickening of the SEI film on the negative electrode surface. This process continuously consumes the limited active lithium ions within the battery, leading to capacity decay and increased internal resistance. Since this event primarily affects the long-term performance degradation of the battery, compared to the other two events, the process is milder and poses the least direct threat to the battery pack. Therefore, it is assigned a relatively small base severity weight: 0.1.

[0065] S32: Use the exponential decay function to determine the time decay coefficient for each historical damage event.

[0066] For the The first cell Historical damage events The time decay coefficient is: , It is a natural exponential function. It is the first The first cell The time interval between the most recent occurrence of a historical damage event and the current testing time. The preset time decay constant is empirically set to 0.05.

[0067] This is a format for an exponential decay function. The exponential decay function measures the impact of a historical damage event on the current test by the time interval between the most recent occurrence of that event and the current test. Specifically, for any given historical damage event, the larger the time interval between its most recent occurrence and the current test, the smaller the time decay coefficient; conversely, the smaller the time interval, the larger the time decay coefficient. This reflects the principle that more recent damage events have a greater impact on the current test.

[0068] for It also provides another way to determine: The higher the base severity weight of a historical damage event, the more persistent its impact, i.e., the longer its half-life, and the smaller its corresponding time decay constant, and vice versa.

[0069] The base severity weight of each historical damage event is mapped to the half-life of that historical damage event. The time decay constant of the historical event is determined by the half-life. This requires setting two basic parameters: the base half-life, which represents the shortest memory duration, is set to 30 days (empirical value); the impact coefficient is set to 100 (empirical value), which means that for every 0.1 increase in the base severity weight, the half-life increases by an additional 10 days, which is used to amplify the lasting impact of severe events.

[0070] For each historical damage event, first calculate the product of the basic severity weight and the impact coefficient of that historical damage event, and add the basic half-life to obtain the half-life of that historical damage event. Then, calculate the time decay constant based on the half-life of each historical damage event. ,in, This is the half-life of the historical damage event.

[0071] S33: Determine the historical damage level of each cell.

[0072] For each battery cell, the weighted damage level of each historical damage event is obtained by multiplying the basic severity weight, the total number of occurrences, and the time decay coefficient. Finally, the weighted damage levels of all historical damage events of the battery cell are summed to obtain the historical damage level of the battery cell.

[0073] Specifically, it is calculated using the following formula:

[0074] in, For the first The extent of historical damage to each battery cell For the first The first cell A historical damage event, For the first The total number of historical damage events per battery cell For the first The first cell The basic severity weights of each historical damage event reflect the degree of damage caused to the cell's electrochemical system by different types of historical damage events. For the first The first cell The first historical damage event in The total number of historical damage events occurring in a battery cell reflects the repetitive and cumulative effect of historical damage. The long-term damage to a battery cell caused by a single, accidental historical damage event is drastically different from that caused by repeated historical damage events. It is the first The first cell The time interval, in days, between the most recent occurrence of a historical damage event and the current testing time.

[0075] in, It is the first The first cell The time decay coefficient of historical damage events reflects the principle that more recent events are more correlated. The first part multiplies the feature values ​​of these three dimensions to obtain the second part. The first cell Historical damage events The weighted damage level is then calculated, and an accumulation operation is performed using the summation symbol to sum the weighted damage levels of all relevant historical damage events of the cell, thus obtaining the historical damage level that can comprehensively reflect the historical cumulative damage level of the cell.

[0076] This attenuation accumulation model effectively assesses and accumulates the impact of historical damage events recorded in the historical log on the battery cell. It can accurately identify potential cumulative anomalies caused by the long-term, repeated accumulation of minor damage (such as continuous lithium deposition), so as to take preventive maintenance strategies for the battery pack after testing in a timely manner.

[0077] S4: Combine the dynamic instability scores and historical damage levels of each cell to generate a comprehensive risk index for each cell.

[0078] After obtaining the dynamic instability score, which characterizes the current dynamic characteristics of the battery cell, and the historical damage level, which characterizes the historical cumulative damage of the battery cell, no single indicator can fully assess the true risk of the battery cell. A battery cell with a high degree of historical damage, even if its instantaneous dynamic instability score is low, still has a much higher probability of failure under disturbance than a battery cell with a low degree of historical damage. Therefore, this step generates a comprehensive risk index by deeply integrating the dynamic instability score and the degree of historical damage. This comprehensive risk index can more accurately reflect the true failure risk of a battery cell with cumulative damage under dynamic disturbance, thereby providing a more accurate and comprehensive quantitative decision-making basis for subsequent system-level health status assessment.

[0079] Specifically, it is based on the following formula:

[0080] in, For the first The comprehensive risk index of each battery cell For the first Dynamic instability score of individual battery cells This represents the maximum value of the dynamic instability score for all battery cells. It is the hyperbolic tangent function. It is a natural exponential function. For the first The extent of historical damage to each battery cell This represents the average historical damage level of all battery cells. The parameter is used to prevent the denominator from being zero, and its value is [value missing]. .

[0081] In this formula, This is a fundamental risk item, the basic part of the formula, representing the immediate risk derived from the current dynamic instability assessment. It ensures that the final comprehensive risk index of the cell is at least equal to its current dynamic instability score. Even a cell without historical damage events will generate an immediate risk if it currently exhibits severe fluctuations.

[0082] In this formula, This is a risk penalty term, which is obtained by multiplying three parts. It will only produce a large value under specific conditions. The first part... The magnitude of the risk penalty is directly proportional to the basic risk; the higher the dynamic instability of the battery cell, the greater the additional penalty from its historical damage. (Part Two) This is a dynamic risk contribution factor, which assesses the relative severity of the current cell's dynamic instability within its battery pack group. The score of an individual cell was normalized relative to the maximum score in the entire battery pack. The effect of the function is that when a battery cell's... much smaller At that time, the dynamic risk contribution factor is also relatively small; when near At this time, the dynamic risk contribution factor is relatively large, approaching 1, which plays a role in dynamic self-calibration. The greater the dynamic instability of a battery cell, the higher the dynamic risk factor it will contribute. This is a historical damage contribution factor, which assesses the relative severity of the cell's historical damage within its group of battery packs. The historical damage of individual cells was normalized relative to the average level of the entire battery pack. The function has a saturation effect, when a cell has historical damage Far below average When the factor value is close to 0, it is close to 0; when When the value is much higher than the average, the factor value approaches 1, which also achieves dynamic self-calibration: a cell will only contribute a high historical damage factor when its historical damage is much higher than the average.

[0083] The core effect of this formula lies in the precise focus of risk. Through the multiplicative linkage of various parts, it achieves the identification and amplification of the most dangerous cells. The additional risk penalty term will only obtain a significant penalty value when both the dynamic risk contribution factor and the historical damage contribution factor are large at the same time. This means that if a cell with serious historical damage is currently showing violent fluctuations, the additional risk penalty will be the largest, and its comprehensive risk index will be significantly increased. If a cell with serious historical damage is currently showing stable performance, its additional risk penalty will be appropriately reduced. If a cell is currently showing violent fluctuations but has little or no historical damage, it is considered to be just a temporary and accidental fluctuation, which is not enough to constitute a major risk, and the additional risk penalty term will also be appropriately reduced.

[0084] By introducing and These two statistical characteristic parameters can distinguish abnormal cells in a generally healthy battery pack from cells that exhibit similar indicators in a generally aging battery pack. This judgment logic, which combines individual state with group background, improves the accuracy and robustness of the test.

[0085] In summary, this step uses a nonlinear fusion model to accurately assess the risk of each battery cell, resulting in a comprehensive index that accurately reflects the true risk.

[0086] By analyzing dynamic operating conditions and conducting comprehensive risk assessments, it generates a comprehensive risk index for each cell that deeply integrates current dynamic instability and historical accumulated damage. This index reflects the true health status and potential failure risk of the cell better than simple dynamic or historical scores, providing a reliable basis for decision-making in the final system-level health testing.

[0087] like Figure 3 As shown in the figure, taking 50 battery cells as an example, a schematic diagram of the distribution of the comprehensive risk index of the 50 cells is provided. It can be seen from the figure that the comprehensive risk index of the vast majority of cells is at a low level, below 0.0005, indicating that they perform well in both current dynamic testing and historical records. However, the comprehensive risk index of several cells is significantly high, forming several prominent peaks. Among them, the comprehensive risk index of cell number 25 reaches a peak of approximately 0.0032, far exceeding all other cells. This cell meets the high... and high The two conditions present the greatest risk of failure.

[0088] S5: Determine the degree of failure of the battery pack based on the comprehensive risk index of each cell, and determine the test results according to the degree of failure.

[0089] The ultimate risk of a battery pack is determined by two core dimensions: the degree of danger of its weakest point and the breadth of impact of its internal problems. This step aggregates these two dimensions through the following operations to make a final judgment on the battery pack's test results. Specifically, this includes: S51: Determine the extreme value risk index.

[0090] Applying the "barrel theory" to assess the single-point vulnerability of a system, in a battery pack composed of hundreds or thousands of cells connected in series, the failure of any one cell (such as an internal short circuit) can trigger catastrophic consequences for the entire system. Therefore, the overall safety of the system is largely limited by its worst-performing cell. Thus, by iterating through the comprehensive risk index of all cells and finding its maximum value, we obtain the extreme risk index. The extreme risk index represents the most explicit and direct danger currently faced by the battery pack. It plays a single-point veto role in the risk model, ensuring that as long as any cell is in extremely poor condition and carries a very high risk, regardless of how well the other cells perform, the final health score will inevitably decrease, thus avoiding the neglect of single-point failures when assessing the health of the battery pack.

[0091] S52: Assess the extent of risk contagion.

[0092] In a battery pack, the higher the proportion of high-risk cells, the greater the spread of risk and the higher the risk of systemic failure, and vice versa. Therefore, the number of cells with a comprehensive risk index exceeding a preset threshold is counted, and the proportion of these cells in the total number of cells in the battery pack is calculated. , The higher the value, the greater the degree of risk spread, and vice versa. The comprehensive risk index threshold is set based on the following method: the comprehensive risk index of all battery cells is calculated, sorted in ascending order, and the third quartile of the sorted cells is used as the comprehensive risk index threshold.

[0093] S53: Calculate the systemic risk index of the battery pack.

[0094] This will be used to reflect the extent of the risk's spread. A nonlinear amplification is performed to obtain the systemic risk index. The specific nonlinear amplification is based on the following formula:

[0095] in, This is a systemic risk index for the battery pack. This represents the proportion of battery cells whose comprehensive risk index exceeds the comprehensive risk index threshold in the total number of battery cells. The parameter used to prevent the denominator from being zero is set to [value]. .

[0096] This formula reflects the accelerated deterioration effect of systemic risk. In a system, the risk of overall collapse does not increase linearly with the addition of fault points, but rather accelerates. When a large number of battery cells are in a sub-optimal state, it often indicates batch-related manufacturing defects or systemic design problems. The probability of a chain reaction of cell failures and synergistic cell failures in the battery pack increases dramatically. Very small (e.g., only 1% of the cells have an overall risk index exceeding the threshold). The amplification effect is negligible, indicating that the battery pack only has sporadic cell failures. Reaching 0.5, The risk is magnified by a factor of two, indicating that the battery pack is already in a significantly unhealthy state. Up to 0.9, The risk is magnified tenfold by a factor of 10, making it extremely sensitive to cell failures in the battery pack.

[0097] S54: Assess the degree of failure of the battery pack.

[0098] Multiplying the extreme value risk index by the systemic risk index yields the battery pack's failure level, which considers both the weakest link and the prevalence of the problem. Specifically, the systemic failure index... ,in, This is an extreme value risk. This represents the systemic risk factor. When a cell in the battery pack experiences a serious malfunction... Extremely high, even with high risk contagion. Very low, system-level failure index It will also be By directly raising the bar, the risk of a single point of failure, like the weakest link in a barrel, can be accurately identified. When the battery pack's cells experience widespread, slow degradation, the risk of failure for any single cell may not be particularly prominent. The numerical value is not large, but due to the large number of battery cells being in a sub-optimal state, the risk spreads significantly. The index is very high, leading to a high system-level failure rate. It becomes extremely large, and after multiplying the two, This will also be amplified to a very high level, thus accurately capturing this widespread, slow degradation failure.

[0099] S55: Determine the health of the battery pack based on the degree of fault in order to complete the test.

[0100] The degree of battery pack failure Compared with the preset fault severity threshold of 0.8, if If the value does not exceed 0.8, the battery pack is considered to be in good health. If the value exceeds 0.8, the battery pack is deemed to be in an unqualified health condition, and the final test result is obtained.

[0101] like Figure 4 As shown in the diagram, there are two core elements: the left bar represents the battery pack's fault level calculated based on the comprehensive risk index of all cells, with a value of 0.0043; the right bar represents the preset fault level threshold, with a value of 0.8. Comparing the battery pack's fault level with the fault level threshold, we find that 0.0043 is much smaller than 0.8. This indicates that although the battery pack may have individual cells with higher risks (such as cell number 25), the overall fault level obtained by multiplying its extreme risk index by the systemic risk index does not reach the dangerous level that requires a disqualification. Therefore, the system ultimately determines that the battery pack's health status is acceptable.

[0102] The fault severity threshold of 0.8 was determined based on the following method: First, using 1000 pre-acquired battery pack samples, including 800 healthy battery pack samples and 200 faulty battery pack samples, a fault degree is calculated for each battery pack sample according to the method of this technical solution, as shown below. Figure 5 The fault severity distribution shown in the figure reveals that the fault severity of healthy battery pack samples and faulty battery pack samples forms two distinct yet overlapping distribution regions. The fault severity of the 800 healthy battery pack samples is highly concentrated in the low segment, mainly below 0.6, forming a high-density, narrow-band distribution. The fault severity of the 200 faulty samples, on the other hand, is mainly distributed in the high segment, mostly above 0.7, with a wider distribution range. Furthermore, there is a small overlap between these two distributions in the 0.7 to 0.9 interval. A test result set containing 1000 rows of data was obtained, with each row containing two key pieces of information: the fault severity of the battery pack sample and the true label of the battery pack sample. The label of the healthy battery pack samples is 0, and the label of the faulty battery pack samples is 1. Then, the fault severity threshold is defined as a classification rule. Essentially, the threshold is a rule for making a yes or no judgment. In the current scenario, the classification rule is: If the fault level of a battery pack sample is greater than or equal to the fault level threshold, the system determines that it has a fault and its health status is unqualified; otherwise, it determines that it has no fault and its health status is qualified.

[0103] Next, the accuracy of a single threshold for the evaluation results is evaluated using a confusion matrix. Since the fault severity ranges from 0 to 1, a candidate threshold is set at intervals of 0.1, from 0.1 to 1, with candidate thresholds of 0.1, 0.2, 0.3, up to 1. Then, the classification rule corresponding to each candidate threshold is applied to 1000 battery pack samples. The classification result of the candidate threshold is then compared one by one with the true label. This comparison will produce four possible outcomes: The first type is a true positive, where a faulty battery pack sample is also greater than or equal to the candidate threshold and is judged to be faulty and unqualified in health status. This indicates that the judgment result of the candidate threshold is consistent with the true label and the judgment result is correct. The second type is a false positive. A healthy battery pack sample may be judged to have a fault degree that is greater than or equal to the candidate threshold. If it is judged to have a fault and its health status is not qualified, it means that the judgment result of the candidate threshold is inconsistent with the true label and the judgment result is wrong. The third type is a true negative. A healthy battery pack sample has a fault level less than the candidate threshold and is judged to have no fault and be in good health. This means that the judgment result of the candidate threshold is consistent with the true label and the judgment result is correct. The fourth type is a false negative. A faulty battery pack sample is judged to be fault-free and in good health if its fault level is less than the candidate threshold. This indicates that the judgment result of the candidate threshold is inconsistent with the true label and the judgment result is incorrect. By counting the number of 1000 battery pack samples in these four cases, the confusion matrix of the candidate threshold is obtained.

[0104] Specifically, the confusion matrix is ​​constructed as follows: the confusion matrix uses the true labels as rows and the judgment results as columns. It iterates through 1000 battery pack samples, and the judgment result of the candidate threshold for each battery pack sample is assigned to the corresponding cell in the table and counted.

[0105] Table 1 shows the composition of the confusion matrix for this candidate threshold, as follows: Table 1

[0106] The confusion matrix clearly quantifies the correctness and error of the judgment results at the candidate threshold, providing a direct data basis for subsequent calculation of the true positive rate and false positive rate.

[0107] Next, the Yoden index is calculated, and the final failure level threshold is determined based on the Yoden index.

[0108] The Youden index is an important metric for evaluating the performance of binary classification models; a higher value indicates better classification performance. Based on the statistical results of the confusion matrix, two core ratios are calculated: True positive rate The proportion of all faulty battery pack samples that were successfully identified as faulty by this candidate threshold. The higher the threshold, the higher the recall rate of faults when classifying based on the candidate threshold, and the lower the risk of misclassifying a real faulty battery pack as a healthy battery pack. False positive rate The proportion of healthy battery pack samples that were incorrectly identified as faulty by this candidate threshold. The lower the threshold, the lower the false alarm rate for healthy batteries when classifying them based on the candidate threshold, and the lower the risk of misclassifying a truly healthy battery pack as a faulty battery pack.

[0109] Yoden Index The Yoden Index cleverly combines the capabilities of preventing false alarms and preventing missed false alarms. An ideal fault severity threshold will allow... As high as possible, while making As low as possible, so that their difference That will reach its maximum.

[0110] Finally, for each candidate threshold from 0.1 to 1, a classification rule is defined, the confusion matrix is ​​obtained, and the true positive rate is calculated. False positive rate The Yodden Index calculates the true positive rate across all candidate thresholds. False positive rate The Yodden Index yielded the statistical results shown in Table 2: Table 2

[0111] As can be seen from Table 2, when the threshold is set to 0.8, the Youden index reaches its maximum value of 0.985, proving that it is the best balance point to distinguish between healthy battery pack samples and faulty battery pack samples. Therefore, 0.8 is taken as the final fault level threshold.

[0112] In summary, this step successfully transforms scattered cell-level risk information into an indicator that accurately assesses the overall health status of the battery pack, thereby enabling health testing of the battery pack.

[0113] The present invention also provides a battery testing system, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the battery testing method described in any of the above embodiments.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A battery testing method, characterized in that, include: Obtain the voltage data sequence of each cell in the battery pack under test and the historical log data of the battery pack; Based on the voltage data sequence of each cell, the parameter volatility index of each cell is calculated to characterize the degree of fluctuation in the voltage data of each cell. Based on the physical topology of the battery pack, all adjacent cells of each cell are identified. Using the parameter volatility index of each cell and all adjacent cells, the spatial correlation weight between each cell and all adjacent cells is calculated. Combining the parameter volatility index and spatial correlation weight of each cell, the dynamic instability score of each cell is calculated. Historical log data is analyzed to extract all historical damage events corresponding to each cell. By analyzing the repeated cumulative effect of all historical damage events, the historical damage level of each cell is determined. The dynamic instability score and historical damage level of each cell are integrated to generate a comprehensive risk index for each cell. By integrating the comprehensive risk index of all battery cells, the degree of failure of the battery pack is calculated, and the health status of the battery pack is determined based on the comparison result of the degree of failure with the preset failure degree threshold, so as to complete the battery test.

2. The battery testing method according to claim 1, characterized in that, Obtain the voltage data sequence of each cell in the battery pack under test, including: A test time window and a sampling frequency are preset. Within the test time window, the voltage of each cell is continuously collected according to the sampling frequency to form a voltage data sequence of each cell.

3. The battery testing method according to claim 1, characterized in that, The parameter volatility index for each battery cell is determined based on the following method: For any given cell, perform first-order difference processing on the voltage data sequence of that cell to obtain a first-order difference sequence; The absolute values ​​of all data in the first-order difference sequence are first calculated, and then the average value is calculated. This average value is used as the parameter volatility index of the battery cell.

4. The battery testing method according to claim 1, characterized in that, The spatial association weights between each cell and all adjacent cells are obtained using the following formula: ； in, For the first Spatial association weights between each cell and all its adjacent cells For cell numbering, For the first The parameter volatility index of individual battery cells For the first The total number of all adjacent cells of a given cell. For the first The first cell The parameter volatility index of adjacent cells, A preset positive number used to prevent the denominator from being zero. It is the absolute value symbol.

5. The battery testing method according to claim 1, characterized in that, The dynamic instability score of each cell is obtained using the following formula: ； in, For the first Dynamic instability score of individual battery cells For the first The parameter volatility index of individual battery cells For the first Spatial association weights between each cell and all its adjacent cells It is a sine function.

6. The battery testing method according to claim 1, characterized in that, By analyzing the repeated cumulative effects of all historical damage events, the degree of historical damage to each cell is determined, including: For each battery cell, all historical damage events corresponding to that cell are filtered out from historical log data. For each historical damage event, the base severity weight, the total number of times the historical damage event has occurred among all historical damage events, and the time interval between the most recent occurrence of the historical damage event and the current testing time are extracted. Using an exponential decay function, the time decay coefficient of the historical damage event is determined based on the time interval. The base severity weight, the total number of occurrences, and the time decay coefficient are multiplied together to obtain the weighted damage degree of the historical damage event. The weighted damage degrees of all historical damage events are summed to obtain the historical damage degree of the battery cell.

7. The battery testing method according to claim 6, characterized in that, The time decay factor for historical damage events is determined based on the following method: For the The first cell Historical damage events The time decay coefficient is , It is a natural exponential function. The preset time decay constant, It is the first The first cell The time interval between the most recent occurrence of a historical damage event and the time being tested.

8. The battery testing method according to claim 1, characterized in that, The comprehensive risk index for each battery cell is generated by integrating the dynamic instability score and historical damage level of each cell, based on the following formula: ； in, For the first The comprehensive risk index of each battery cell For the first Dynamic instability score of individual battery cells This represents the maximum value of the dynamic instability score for all battery cells. It is the hyperbolic tangent function. It is a natural exponential function. For the first The extent of historical damage to each battery cell This represents the average historical damage level of all battery cells. This is a parameter used to prevent the denominator from being 0.

9. The battery testing method according to claim 1, characterized in that, The degree of battery pack failure is determined based on the following method: The maximum value among all the comprehensive risk indices of battery cells is taken as the extreme value risk index; The number of battery cells whose comprehensive risk index exceeds the preset comprehensive risk index threshold is counted. The proportion of this number of battery cells in the total number of battery cells in the battery pack is then non-linearly amplified to obtain a systemic risk index. Multiplying the extreme risk index by the systematic risk index yields the degree of battery pack failure.

10. A battery testing system, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the battery testing method as described in any one of claims 1-9.

Images