Battery control method of BMS (Battery Management System)

Abstract

The invention relates to the technical field of battery management systems, in particular to a battery control method of a BMS (battery management system), which comprises the following steps of: acquiring a voltage sampling value, a temperature sampling value, an internal resistance sampling value and a charge state sampling value of each single battery according to a BMS sampling period, and constructing a single multi-dimensional feature vector; according to the method, the electrochemical dynamic characteristics in the internal polarization elimination process of the battery can be captured by calculating the voltage relaxation rebound slope. And the calculated value is compared with a slope limit dynamic threshold retrieved according to the current temperature and the charge state, so that non-intrusive online quantitative evaluation of the lithium precipitation risk on the surface of the negative electrode is realized. The charging request current is dynamically adjusted or cut-off control is executed according to the lithium precipitation abnormal state judgment index, it is ensured that the battery always works within the nondestructive electrochemical boundary, the fast charging requirement is met, capacity fading and internal short circuit hidden dangers caused by lithium precipitation are avoided, and the safety of the whole life cycle of the battery is improved.

Description

Technical Field

[0001] This invention relates to the field of battery management system technology, and more particularly to a battery control method for a BMS management system. Background Technology

[0002] The field of battery management system technology involves the monitoring and management of power battery packs and their individual cells throughout their entire life cycle, including the real-time acquisition and monitoring of key physical parameters such as battery voltage, current, temperature and internal resistance.

[0003] Current technologies for monitoring and managing power batteries primarily rely on independent threshold judgments for individual physical parameters such as voltage, temperature, and current. This means that an alarm or protection is triggered only when a parameter exceeds a preset upper or lower limit. This discrete monitoring mode ignores the strong correlations and coordinated changes among the various physical parameters within the battery, making it difficult to identify early-stage hidden faults where individual indicators appear normal but the overall state has deviated from the group distribution. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a battery control method for a BMS management system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a battery control method for a BMS management system, comprising the following steps: Based on the BMS sampling period, the voltage, temperature, internal resistance and state of charge of each individual cell are collected to construct a multidimensional feature vector of the individual cell. Set statistical operations are performed on the multidimensional feature vectors of all individual cells to establish the multidimensional feature statistical covariance matrix of the battery pack. Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, a distance score is calculated and obtained. The distance score is compared with a preset critical threshold. If the distance score is greater than the critical threshold, it is determined to be a multidimensional outlier fault and the fault branch disconnection action is triggered, generating a multidimensional outlier fault isolation control signal. Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption action is triggered during the fast charging process. At the same time, the stationary phase terminal voltage timing data of the specified interval after the current returns to zero is intercepted. Based on the stationary phase terminal voltage timing data, the voltage relaxation rebound slope is calculated. The slope limit dynamic threshold under the current temperature and state of charge is retrieved. The calculated value of the voltage relaxation rebound slope is compared with the slope limit dynamic threshold to obtain the lithium plating abnormal state judgment index. Based on the lithium plating abnormal state judgment index, the degradation adjustment action of the charging request current is triggered to generate a lossless boundary charging cutoff control command.

[0006] Preferably, the steps for obtaining the multidimensional feature statistical covariance matrix of the battery pack are as follows: According to the BMS sampling period, the voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value are read one by one according to the individual cell identifier. The four types of sampling values ​​are indexed and aligned according to a unified timestamp and arranged in a fixed order as voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. The individual cell identifier and sampling timestamp are bound to form a synchronous sequence of individual cell voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. Based on the synchronization sequence of the individual cell voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value, a vector of length four is constructed in the order of voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value, and the corresponding individual cell identifier and sampling timestamp are recorded to generate a multi-dimensional feature vector of the individual cell. Based on the individual multidimensional feature vectors, the individual multidimensional feature vectors of all individual cells within the current BMS sampling period are aggregated, the mean vectors of the four dimensions are calculated, the average of the product of the deviations of each dimension is calculated within the individual cell range, and the matrix rows and columns are filled in a fixed dimension order to generate the multidimensional feature statistical covariance matrix of the battery pack.

[0007] Preferably, the step of obtaining the distance score is as follows: Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, the voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value of the current sampling period are extracted according to the individual cell identifier. The corresponding components of the mean vector are subtracted from each sample value to form the individual cell difference vector. The inverse matrix of the multidimensional feature statistical covariance matrix of the battery pack is calculated to represent the cooperative offset relationship between each dimension, and the individual cell difference vector and covariance inverse matrix pair are generated. The distance score is calculated based on the individual difference vector and the covariance inverse matrix.

[0008] Preferably, the step of acquiring the multidimensional outlier fault isolation control signal is as follows: Based on the distance score, a set critical threshold is invoked, and the distance score is compared with the critical threshold. When the distance score exceeds the critical threshold, it is determined that the corresponding single cell has a multidimensional outlier fault state. An anomaly identification flag is written and the branch disconnection logic is triggered to generate a multidimensional outlier fault isolation control signal.

[0009] Preferably, the step of acquiring the terminal voltage timing data during the resting phase is as follows: Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption is performed during the fast charging process to stop charging and maintain a preset resting time. Starting from the target time when the current returns to zero, the terminal voltage and corresponding timestamp are extracted according to the sampling period to form the terminal voltage timing data during the resting stage.

[0010] Preferably, the step of obtaining the voltage relaxation rebound slope calculation value is as follows: Based on the time series data of the terminal voltage during the resting stage, the time difference and terminal voltage difference of adjacent samples are calculated in ascending order of timestamps. Samples with zero time difference are removed to obtain the linear differential sequence of the terminal voltage during the resting stage, forming a set of voltage change difference and time interval difference pairs. The voltage relaxation rebound slope is calculated based on the pairing set of voltage change difference and time interval difference.

[0011] Preferably, the step of obtaining the lithium plating abnormality determination index is as follows: Based on the current temperature and state of charge values, the effective range of the current temperature and state of charge values ​​in the slope-limited dynamic threshold table is limited. The row index is located by the temperature value and the column index is located by the state of charge value. When the values ​​fall into adjacent intervals, linear interpolation is performed between adjacent entries according to the ratio of the two ends and a single value is output to generate the slope-limited dynamic threshold. Based on the slope limit dynamic threshold, the calculated value of the voltage relaxation rebound slope is read, and the difference between the slope limit dynamic threshold and the calculated value of the voltage relaxation rebound slope is calculated. When the difference is greater than zero, the risk flag is set to true and the comparison timestamp is recorded. When the difference is less than or equal to zero, the risk flag is set to false and the comparison timestamp is recorded, thereby generating a lithium plating abnormal state judgment index.

[0012] Preferably, the step of obtaining the lossless boundary charging cutoff control command is as follows: According to the lithium plating abnormal state judgment index, when the risk flag is true, the charging request current level is downgraded according to the downgrade level mapping table and the cutoff trigger flag is set. When the risk flag is false, the charging request current level is maintained and the cutoff trigger flag is cleared. The execution channel field, target status field and failure protection bit are encapsulated to generate a lossless boundary charging cutoff control command.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by collecting the voltage, temperature, internal resistance, and state of charge of each individual battery cell and constructing a multidimensional feature vector for each cell, and then performing ensemble statistical operations to establish a multidimensional feature statistical covariance matrix for the battery pack, the inherent coupling relationship and dispersion between different physical quantities can be explored, overcoming the limitation of single-parameter monitoring in detecting complex faults. Using this covariance matrix to calculate a distance score and compare it with a critical threshold, accurate identification and location of multidimensional outlier faults are achieved, triggering branch disconnection actions in the early stages of the fault and effectively blocking the risk propagation path. Based on the remaining connected branches after fault isolation, a timed current interruption mechanism is introduced during the fast charging process, and the terminal voltage timing data during the resting phase is intercepted. By calculating the voltage relaxation rebound slope, the electrochemical kinetic characteristics of the internal polarization elimination process of the battery can be captured. Comparing this calculated value with the slope-limited dynamic threshold retrieved based on the current temperature and state of charge, a non-invasive online quantitative assessment of the risk of lithium plating on the negative electrode surface is achieved. Based on the lithium plating abnormality judgment index, the charging request current is dynamically adjusted or the cut-off control is executed to ensure that the battery always works within the non-destructive electrochemical boundary. This not only meets the fast charging requirements, but also avoids the capacity decay and internal short circuit risks caused by lithium plating, thus improving the safety of the battery throughout its entire life cycle. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Please see Figure 1 This invention provides a technical solution, a battery control method for a BMS management system, comprising the following steps: Based on the BMS sampling period, the voltage, temperature, internal resistance and state of charge of each individual cell are collected to construct a multidimensional feature vector of the individual cell. Then, set statistical operations are performed on the multidimensional feature vectors of all individual cells to establish the multidimensional feature statistical covariance matrix of the battery pack. Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, the distance score is calculated and obtained. The distance score is compared with the preset critical threshold. If the distance score is greater than the critical threshold, it is determined to be a multidimensional outlier fault and the fault branch disconnection action is triggered, generating a multidimensional outlier fault isolation control signal. Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption action is triggered during the fast charging process. At the same time, the stationary phase terminal voltage timing data of the specified interval after the current returns to zero is intercepted. Based on the stationary phase terminal voltage timing data, the voltage relaxation rebound slope is calculated. The slope limit dynamic threshold is retrieved under the current temperature and state of charge. The calculated value of voltage relaxation rebound slope is compared with the slope limit dynamic threshold to obtain the lithium plating abnormal state judgment index. Based on the lithium plating abnormal state judgment index, the degradation adjustment action of the charging request current is triggered to generate a lossless boundary charging cutoff control command.

[0017] The steps for obtaining the multidimensional feature statistical covariance matrix of the battery pack are as follows: According to the BMS sampling period, the voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value are read one by one according to the individual cell identifier. The four types of sampling values ​​are indexed and aligned according to a unified timestamp and arranged in a fixed order as voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. The individual cell identifier and sampling timestamp are bound to form a synchronous sequence of individual cell voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. Based on the synchronization sequence of individual cell voltage sampling values, temperature sampling values, internal resistance sampling values, and state of charge sampling values, a vector of length four is constructed in the order of voltage sampling values, temperature sampling values, internal resistance sampling values, and state of charge sampling values, recording the corresponding individual cell identifier and sampling timestamp, and generating a multi-dimensional feature vector for each individual cell. Based on the individual multidimensional feature vectors, the individual multidimensional feature vectors of all individual cells within the current BMS sampling period are aggregated, the mean vectors of the four dimensions are calculated, the average of the product of the deviations of each dimension is calculated within the individual cell range, and the matrix rows and columns are filled in a fixed dimension order to generate the multidimensional feature statistical covariance matrix of the battery pack.

[0018] Specifically, based on the BMS sampling period, the sampling period is set to 10 milliseconds. This duration is determined by the time constant of the charge transfer process in the battery's electrochemical reaction, ensuring the capture of rapidly changing voltage and current transient responses. The analog-to-digital converter is activated to discretize and read the analog voltage signals of all individual cells. Simultaneously, a multiplexer polls the NTC thermistors connected to the surface of each cell to obtain temperature samples. Internal resistance and state-of-charge (SOC) samples, estimated in real-time by an extended Kalman filter algorithm, are retrieved from the memory-mapped area. These four types of data exhibit slight timing deviations at their original acquisition times. A uniform time synchronization tolerance window is set, for example, a tolerance window width of 10% of the sampling period, or 1 millisecond. To balance the errors between data real-time performance and synchronization accuracy, the timestamps of the four types of data within the same sampling period are compared with the reference timestamp. If the absolute value of the timestamp deviation falls within a 1-millisecond tolerance window, it is determined to be a data frame belonging to the same moment. For data exceeding the tolerance window, linear interpolation is performed. The corresponding value at the reference moment is calculated using the values ​​of the two valid sampling points before and after and the time difference. After completing the numerical correction, the voltage, temperature, internal resistance, and state of charge data are sequentially filled in according to a fixed memory address offset. Each group of aligned data packets is labeled with the corresponding physical location number of the individual battery and the uniformly corrected sampling timestamp, forming a synchronization sequence of individual battery voltage sampling values, temperature sampling values, internal resistance sampling values, and state of charge sampling values.

[0019] Based on the synchronous sequence of individual cell voltage, temperature, internal resistance, and state of charge (SCC) sampling values, boundary checks are performed on the validity of the data in the sequence. A pre-defined effective range table for physical quantities is invoked. For example, voltage sampling values ​​are compared to the range of 2.0V to 4.5V, temperature sampling values ​​to the range of -40°C to 85°C, internal resistance sampling values ​​to the range of 0.1mΩ to 10mΩ, and SCC sampling values ​​to the range of 0% to 100%. This effective range table is derived from the operating limit parameters specified in the battery specifications. Data falling within the valid range is retained, while abnormal noise points exceeding the range are clamped using the valid value from the previous moment. Subsequently, following the fixed physical dimension order of voltage, temperature, internal resistance, and state of charge, a contiguous floating-point data storage space is allocated in memory to construct a column vector structure of length four. The unique identifier of the corresponding individual battery is used as the index key of the vector, and the synchronously corrected sampling timestamp is used as the time dimension label of the vector to ensure that each vector has unique traceability in time and space, thus generating a multidimensional feature vector for each individual battery.

[0020] Based on the multidimensional feature vector of each cell, the multidimensional feature vectors of all series-connected cells are read from the buffer pool in the current sampling period. The total number of cells is set to N, for example, N = 96. The four dimensions of voltage, temperature, internal resistance, and state of charge are accumulated over the N cells and divided by the total number N to obtain the mean vector containing the four components. Then, the covariance matrix is ​​constructed. For each cell, its multidimensional feature vector is subtracted from the mean vector to obtain the cell deviation vector. The cell deviation vector is then multiplied by its transpose to obtain a 4x4 cell covariance contribution matrix. The contribution matrices of all cells are summed and divided by the degree of freedom factor. The calculation formula is as follows: ,in, Represents the current sampling time The multidimensional feature statistical covariance matrix of the battery pack is shown below. This represents the total number of individual cells in the battery pack. The index number representing a single cell, ranging from 1 to , Representing the Each individual cell at time The single-unit multidimensional feature vector, Representative moment The mean vector of all individual cell feature vectors. This represents the transpose operation of a vector or matrix. Through this calculation process, the correlation between different physical dimensions is quantified into matrix elements. The calculation results are written into a pre-allocated memory block in row-major order to generate the multidimensional feature statistical covariance matrix of the battery pack.

[0021] The steps to obtain the distance rating are as follows: Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, the voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value of the current sampling period are extracted according to the individual cell identifier. The corresponding components of the mean vector are subtracted to form the individual cell difference vector. The inverse matrix of the multidimensional feature statistical covariance matrix of the battery pack is calculated to represent the cooperative offset relationship between each dimension, and the individual cell difference vector and covariance inverse matrix pair are generated. The distance score is calculated based on the individual difference vector and the inverse covariance matrix. The calculation formula is as follows: ; in, Rate the distance. At the current sampling time, This is the single-unit difference vector at the current sampling time. Let be the inverse covariance matrix at the current sampling time. The single-unit difference velocity vector at the current sampling time is calculated as follows: ,in The single-unit difference vector at the previous sampling time, The BMS sampling period is The state characteristic time constant is dynamically calculated using a built-in function based on the combination of the current temperature T and the state of charge S, and is used to balance the effects of static deviation terms and dynamic change terms.

[0022] Specifically, based on the multidimensional feature covariance matrix of the battery pack and the multidimensional feature vector of each individual cell, the individual cell index table stored in the battery management system is traversed. Using pointer offsets, the data address of each individual cell in the memory stack is located sequentially. The voltage, temperature, internal resistance, and state of charge samples uploaded by the underlying driver and filtered within the current sampling period are read. These four physical quantities are loaded sequentially from high to low bits according to a predefined structure format. Simultaneously, the direct memory access controller is invoked to retrieve the previously calculated mean vector containing the average values ​​of the four dimensions from the shared memory region. The floating-point unit is then activated to perform vector subtraction, subtracting the corresponding dimension value from each component of the four-dimensional feature vector of each individual cell (e.g., subtracting the average voltage from the individual cell voltage, and the average temperature from the individual cell temperature). This eliminates common-mode interference and preserves the individual cell's differences relative to the group, resulting in a value reflecting the individual cell's deviation. The individual cell difference vector is obtained, and then matrix inversion is performed on the covariance matrix of the multidimensional features of the battery pack. Before performing the inversion, the condition number of the covariance matrix is ​​calculated using the singular value decomposition method to evaluate the ill-conditioning of the matrix. If the condition number is less than the preset stability threshold, the Gaussian-Jordan elimination method is directly used to solve the inverse matrix. If the condition number shows that the matrix is ​​close to singular, a small regularization parameter, such as 1e-6, is added to the main diagonal to ensure numerical stability before performing the inversion. Through this inversion process, the covariance information in the original feature space is converted into a precision matrix. The element values ​​in this precision matrix quantify the cooperative change relationship between the four dimensions of voltage, temperature, internal resistance and state of charge, as well as the uncertainty weight of each dimension. Finally, the individual cell difference vector of each individual cell is associated and bound with this shared covariance inverse matrix. A structure pair containing vector pointers and matrix pointers is constructed in the cache to generate a pair of individual cell difference vectors and covariance inverse matrices.

[0023] The distance score calculation formula introduces the inverse covariance matrix. Eliminating scale differences between different physical dimensions (such as voltage in volts and internal resistance in milliohms) enables dimensionless fusion of multidimensional features, while utilizing the state feature time constant. By introducing a velocity term, the algorithm can not only detect static faults where the values ​​exceed limits, but also capture early evolution faults where the values ​​are still within the normal range but the rate of change is abnormal.

[0024] The single-cell difference vector representing the current sampling moment is a column vector containing four components: voltage deviation (V), temperature deviation (°C), internal resistance deviation (mΩ), and state of charge deviation (%). This vector is obtained by subtracting the average value of the battery pack from the real-time collected single-cell data. For example, if the voltage deviation of a certain single cell is collected and calculated by the sensor to be 0.05V, the temperature deviation to be 2.0°C, the internal resistance deviation to be 0.2mΩ, and the SOC deviation to be -1.0%, then... This parameter quantifies the absolute distance of an individual from the group center at the current moment.

[0025] The inverse covariance matrix (precision matrix) representing the current sampling time has dimensions of 4 rows and 4 columns, with units of [missing information]. The matrix, including its cross terms, reflects the statistical certainty of each feature dimension. Dimensions with smaller variances correspond to higher precision values. For example, based on historical operating data, voltage noise is extremely low, so its weight is set higher, while temperature fluctuations are larger, so its weight is lower. Therefore, a diagonal matrix format is used. 400 corresponds to the voltage dimension (meaning that a deviation of 0.05V will have a significant impact), 0.25 corresponds to the temperature dimension, 25 corresponds to the internal resistance dimension, and 1 corresponds to the SOC dimension.

[0026] The unit difference velocity vector represents the velocity vector of the individual cells at the current sampling time, with units of [missing information]. The calculation method is as follows ,in This is the single-unit difference vector from the previous sampling time. Set the BMS sampling period. Seconds (10ms), for example, the previous moment This parameter reflects the instantaneous rate of fault development.

[0027] The time constant representing the state characteristics, in seconds (s), is used to balance the weights of static and dynamic terms, ensuring that their contributions to the final score are matched. It is calculated from the current temperature T and state of charge S through a table lookup or function. The calculation formula is set as follows: ,in The value is 0.1 seconds. Under the current operating conditions, the value is 1.0, that is... The square of this parameter have The dimensions of the quantity, and In The dimensions cancel each other out.

[0028] Calculations based on parameters: Calculate the individual difference velocity vector : ; .

[0029] Calculate the static deviation term : .

[0030] .

[0031] Calculate the dynamic change term : .

[0032] .

[0033] .

[0034] .

[0035] Calculate the final distance score : .

[0036] The results indicate that the current comprehensive outlier distance score for a single cell is 3.05. This value integrates static deviation (contributing 4.0 squared component) and dynamic rate of change (contributing 5.29 squared component). It shows that although the static deviation is acceptable (about 2 standard deviations), its rate of change is fast (the dynamic component contributes more than the static component), and there is a trend of rapid deterioration. Compared with the evaluation method that only uses static distance, this result can provide an earlier warning of potential faults.

[0037] The steps for obtaining the multidimensional outlier isolation control signal are as follows: Based on the distance score, the set critical threshold is called, and the distance score is compared with the critical threshold. When the distance score exceeds the critical threshold, it is determined that the corresponding single cell has a multidimensional outlier fault state. An anomaly identification flag is written and the branch disconnection logic is triggered to generate a multidimensional outlier fault isolation control signal.

[0038] Specifically, based on the distance score, the fault threshold configuration parameters stored in non-volatile memory are read. The setting process for this threshold is based on statistical analysis of massive amounts of historical normal operating data. For example, if 1000 charge-discharge cycles of normal operating data are collected, the distance score at all sampling times is calculated, and the probability density distribution curve of the score is obtained. The upper limit of the distribution covering 99.99% of normal samples is selected as the benchmark. For example, if the statistically obtained mean is 0.5 and the standard deviation is 0.2, then the critical threshold is set to the mean plus 6 times the standard deviation. The calculated distance score is compared with the set critical threshold of 1.7. If the real-time calculated distance score is greater than the critical threshold of 1.7, it is determined that the behavior pattern of the single battery cell has exceeded the normal statistical fluctuation range, confirming that it is in a multidimensional outlier fault state. The corresponding fault flag is immediately set to 1 in the fault status register. At the same time, the drive logic of the hardware protection circuit is started, a high-level pulse signal is sent to the relay control unit to control the main circuit contactor to disconnect to cut off the current circuit, and a control message containing the fault cell number, the time of fault occurrence and the fault level code is generated to generate a multidimensional outlier fault isolation control signal.

[0039] The steps for obtaining the terminal voltage timing data during the resting phase are as follows: Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption is performed during the fast charging process to stop charging and maintain a preset resting time. Starting from the target time when the current returns to zero, the terminal voltage and corresponding timestamp are extracted according to the sampling period to form the terminal voltage timing data during the resting stage.

[0040] Specifically, based on the remaining connected branches after the execution of the multi-dimensional outlier fault isolation control signal, the current network topology status word stored in the internal register of the battery management unit is read. The valid battery branch number that is not marked as faulty and whose relay is in a closed conducting state is identified. It is then confirmed whether the vehicle is in fast charging mode. If in fast charging mode, a request command with a current setting of zero is immediately sent to the charging pile communication protocol interface. This controls the power electronic switch to cut off the main circuit current, and simultaneously starts a timer to maintain the circuit in an open-circuit static state. The duration of this state is determined by a preset static time parameter. The preset static time setting process is based on selecting battery cells from the same batch and conducting step response tests in a laboratory environment, recording the electrical... The average time required for the voltage to reach 95% of its steady-state value from the moment the current is cut off is calculated. For example, test data shows that this time is distributed between 180 and 240 seconds. To ensure data integrity and allow for margin, 300 seconds is selected as a fixed preset resting time. The moment when the current sensor feedback value drops to zero amperes is marked as the target time when the current returns to zero. The analog front-end acquisition module is called to continuously sample the terminal voltage of each individual battery in the remaining connected branches with a fixed sampling period of 10 milliseconds. At the same time, the hardware timestamp corresponding to each sampling action is recorded. The collected voltage values ​​and timestamp data are stored in a circular buffer queue in a one-to-one correspondence until the timer value reaches the preset resting time of 300 seconds, forming the terminal voltage timing data of the resting stage.

[0041] The steps for obtaining the voltage relaxation rebound slope calculation value are as follows: Based on the time series data of the terminal voltage during the resting phase, the time difference and terminal voltage difference of adjacent samples are calculated in ascending order of timestamps. Samples with zero time difference are removed to obtain the linear differential sequence of the terminal voltage during the resting phase, forming a pairing set of voltage change difference and time interval difference. Based on the pairing set of voltage change difference and time interval difference, the voltage relaxation rebound slope is calculated using the following formula: ; in, Here, m represents the calculated voltage relaxation rebound slope, m is the total number of samples in the terminal voltage time series data during the resting phase, and r is the difference index. Let be the difference between adjacent terminal voltages of the r-th sample. Let the timestamp difference of the r-th sample be the value between adjacent timestamps. The time decay weight is calculated using the following formula: , Let r be the absolute timestamp of the r-th sample. This is the initial timestamp for the resting phase. The relaxation characteristic time constant is determined based on the combination of the current temperature T and the state of charge S, and is used to describe the time-domain characteristics of the relaxation rate under different operating conditions.

[0042] Specifically, based on the timing data of the terminal voltage during the resting phase, the recorded voltage and time data pairs are read in batches from the memory buffer. The data columns are then quickly sorted according to the timestamp values ​​in ascending order to ensure the monotonically increasing nature of the timing logic. An empty dynamic array is created to store the differential pairs. A loop traversal program is started, starting from the second data point, sequentially reading the data at the current point index r and the data at the previous point index r-1. The difference between the current time stamp and the previous time stamp is calculated to obtain the time interval difference. The difference between the current terminal voltage and the previous terminal voltage is calculated to obtain the voltage change difference. This process is then applied to the calculation... The calculated time interval difference is validated by setting a minimum time resolution threshold. This threshold is based on the minimum jitter range of the sampling clock, for example, 1 microsecond. The time interval difference is compared with this threshold. If the difference is less than or equal to zero, it is determined to be an invalid sample or a duplicate record. The sample point is directly removed from the sequence to avoid division by zero errors in subsequent calculations. For samples that pass the validation, the voltage change difference and the time interval difference are combined into a data pair and filled into the pre-allocated memory space in sequence to obtain the linear differential sequence of the terminal voltage during the resting stage, forming a pairing set of voltage change difference and time interval difference.

[0043] The formula for calculating the voltage relaxation rebound slope incorporates a time decay weight. By assigning higher computational weight to the voltage change during the initial settling period, the initial rebound slope, which mainly reflects the kinetic characteristics of electrochemical polarization elimination and lithium ion re-intercalation on the surface of solid particles, is extracted. This overcomes the shortcomings of the traditional average slope method in being insensitive to long-tailed diffusion processes and can more sensitively capture abnormal fluctuations in the voltage plateau caused by lithium plating side reactions.

[0044] The total number of samples representing the terminal voltage time series data during the resting phase is a dimensionless integer. This parameter is obtained by reading the array length of the pairing set of voltage change differences and time interval differences generated in the previous steps. For example, after downsampling at intervals of 100 milliseconds (0.1 seconds) within a 300-second resting period, after deducting the starting point, the actual number of effective differential sample pairs obtained is 2999. Here, to demonstrate the calculation process, the first 3 effective sample points are selected for calculation, i.e., let... The value is 4 (indexes from 2 to 4).

[0045] Representing the The voltage difference between adjacent samples, measured in volts (V), is obtained by differentiating the raw data collected by the voltage sensor. It reflects the magnitude of voltage rebound. For example, in three consecutive sampling intervals during the initial resting period, the measured voltage rebound increases of 0.005V, 0.004V, and 0.003V respectively, corresponding to... time value.

[0046] Representing the The difference between adjacent timestamps of each sample, in seconds (s), is obtained by recording and differentially analyzing the high-frequency clock of the BMS. Corresponding to the sampling interval of the voltage change mentioned above, it is set to a fixed sampling period of 0.1s.

[0047] The initial timestamp of the resting phase is in seconds (s). This parameter records the absolute system time at the instant the current is cut off, for example, reading the system clock as 1000.0s.

[0048] Representing the The absolute timestamp of each sample, in seconds (s), and The corresponding values ​​are 1000.1s, 1000.2s, and 1000.3s, respectively.

[0049] The relaxation characteristic time constant, expressed in seconds (s), is used to control the decay rate of the weights. It is obtained by performing electrochemical impedance spectroscopy (EIS) on the battery beforehand, extracting the characteristic frequencies corresponding to the charge transfer impedance at different temperatures T (e.g., every 5 degrees) and different states of charge S (e.g., every 10%), calculating the time constant, and storing it in a lookup table in non-volatile memory. During real-time operation, the current battery temperature is read as 25°C, the estimated state of charge (SOC) is 80%, and the relaxation characteristic time constant under this condition is obtained by bilinear interpolation from the lookup table as 1.0 s.

[0050] Represents the time decay weight, which is dimensionless and is calculated using the following formula: This parameter is dynamically calculated based on real-time timestamps and is used to highlight the changing characteristics in the early stages of relaxation.

[0051] Calculations based on parameters: Calculate the time difference for each sample point and weight : for The time difference is , .

[0052] for The time difference is , .

[0053] for The time difference is , .

[0054] Calculate the numerator : Item 1: .

[0055] Item 2: .

[0056] Item 3: .

[0057] Total numerator = .

[0058] Calculate the denominator term : .

[0059] Item 1: .

[0060] Item 2: .

[0061] Item 3: .

[0062] Sum of denominators = .

[0063] Calculate the final value : V / s.

[0064] The results indicate that the weighted voltage relaxation rebound slope under the current operating conditions is 0.04066 volts per second. This value quantifies the voltage recovery rate of the battery at the moment of power failure. If this value is significantly higher than the normal reference value under the same temperature and SOC (e.g., the normal value is 0.02V / s), it suggests that there may be an abnormal potential difference inside the battery or an additional potential generated by the dissolution of the lithium plating layer. This provides a highly sensitive quantitative indicator for subsequent judgment of lithium plating faults.

[0065] The steps for obtaining the indicators for judging abnormal lithium plating conditions are as follows: Based on the current temperature and state of charge values, the effective range of the current temperature and state of charge values ​​in the slope-limited dynamic threshold table is limited. The row index is located by the temperature value and the column index is located by the state of charge value. When the values ​​fall into adjacent intervals, linear interpolation is performed between adjacent entries according to the ratio of the two ends and a single value is output to generate the slope-limited dynamic threshold. Based on the slope limit dynamic threshold, read the voltage relaxation rebound slope calculation value, calculate the difference between the slope limit dynamic threshold and the voltage relaxation rebound slope calculation value, set the risk flag as true and record the comparison timestamp when the difference is greater than zero, set the risk flag as false and record the comparison timestamp when the difference is less than or equal to zero, and generate the lithium plating abnormal state judgment index.

[0066] Specifically, based on the current temperature and state of charge (SOC) values, the non-volatile storage area is accessed, and a pre-defined slope-limited dynamic threshold table is invoked. This table is constructed based on numerous lithium plating boundary test experiments under various operating conditions, such as lithium plating initiation voltage rebound slope data obtained from three-electrode battery tests within a temperature range of -20°C to 60°C and a SOC range of 0% to 100%. The current temperature value is compared with the temperature node in the threshold table row index. If the current temperature is 27°C, it falls between the 25°C and 30°C nodes. The current SOC value is then compared with the SOC node in the threshold table column index. If the current SOC is 63%, it falls between the 60% and 70% nodes. In between, the interpolation weights for the temperature dimension are calculated, for example (27-25) / (30-25)=0.4, and the interpolation weights for the SOC dimension are calculated, for example (63-60) / (70-60)=0.3. The slope thresholds corresponding to these four adjacent nodes are read, for example, 0.02V / s for 60% SOC at 25 degrees, 0.025V / s for 60% SOC at 30 degrees, 0.018V / s for 70% SOC at 25 degrees, and 0.022V / s for 70% SOC at 30 degrees. First, interpolation is performed in the temperature dimension to obtain two intermediate values, and then interpolation is performed in the SOC dimension using these two intermediate values ​​to finally obtain the accurate limit value under the current operating condition and generate the slope limit dynamic threshold.

[0067] Based on the slope limit dynamic threshold, the previously calculated voltage relaxation rebound slope value is read from the register. The slope limit dynamic threshold is used as the minuend, and the voltage relaxation rebound slope value is used as the subtrahend. A floating-point subtraction operation is performed, and the output result of the comparison operation is stored in a temporary variable. If the calculated difference is greater than zero, it indicates that the current voltage rebound rate is lower than the critical rate of lithium plating risk, and there are no obvious lithium plating characteristics inside the battery. At this time, the lithium plating risk flag in memory is cleared to zero. If the calculated difference is less than or equal to zero, it indicates that the current voltage rebound rate is too fast and has reached or exceeded the boundary condition of lithium plating, suggesting that there may be metallic lithium deposition on the negative electrode surface. At this time, the lithium plating risk flag is immediately set to 1 (i.e., true state). At the same time, the current system clock is read, and the precise timestamp of this comparison is written to the fault log so that the start time of the fault can be traced later to generate lithium plating abnormal state judgment indicators.

[0068] The steps for obtaining the lossless boundary charging cutoff control command are as follows: Based on the lithium plating abnormal state judgment index, when the risk flag is true, the charging request current level is downgraded according to the downgrade level mapping table and the cutoff trigger flag is set. When the risk flag is false, the charging request current level is maintained and the cutoff trigger flag is cleared. The execution channel field, target status field and failure protection bit are encapsulated to generate a lossless boundary charging cutoff control command.

[0069] Specifically, based on the lithium plating anomaly judgment index, the status change of the risk flag bit is monitored in real time. When the risk flag is detected as true, the charging strategy adjustment mechanism is triggered, and the preset degradation level mapping table is accessed. This mapping table defines the current backoff ratio under different risk levels. For example, if the current state is risk-triggered, the corresponding item in the table is looked up to set the target charging current to 50% of the current requested value or directly downgrade to the C / 20 low current maintenance mode. At the same time, the cutoff trigger flag is set to indicate that the protective charging stage has been entered. When the risk flag is false, the original BMS charging request current command is maintained without any degradation intervention, and the cutoff trigger flag is cleared. Finally, according to the format requirements of the CAN bus communication protocol, the calculated final charging current request value is filled into the execution channel field, the cutoff trigger flag is filled into the target status field, and the corresponding failure protection verification bit is set to generate a lossless boundary charging cutoff control command.

[0070] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A battery control method for a BMS management system, characterized in that, Includes the following steps: Based on the BMS sampling period, the voltage, temperature, internal resistance and state of charge of each individual cell are collected to construct a multidimensional feature vector of the individual cell. Set statistical operations are performed on the multidimensional feature vectors of all individual cells to establish the multidimensional feature statistical covariance matrix of the battery pack. Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, a distance score is calculated and obtained. The distance score is compared with a preset critical threshold. If the distance score is greater than the critical threshold, it is determined to be a multidimensional outlier fault and the fault branch disconnection action is triggered, generating a multidimensional outlier fault isolation control signal. Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption action is triggered during the fast charging process. At the same time, the stationary phase terminal voltage timing data of the specified interval after the current returns to zero is intercepted. Based on the stationary phase terminal voltage timing data, the voltage relaxation rebound slope is calculated. The slope limit dynamic threshold under the current temperature and state of charge is retrieved. The calculated value of the voltage relaxation rebound slope is compared with the slope limit dynamic threshold to obtain the lithium plating abnormal state judgment index. Based on the lithium plating abnormal state judgment index, the degradation adjustment action of the charging request current is triggered to generate a lossless boundary charging cutoff control command.

2. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the multidimensional feature statistical covariance matrix of the battery pack are as follows: According to the BMS sampling period, the voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value are read one by one according to the individual cell identifier. The four types of sampling values ​​are indexed and aligned according to a unified timestamp and arranged in a fixed order as voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. The individual cell identifier and sampling timestamp are bound to form a synchronous sequence of individual cell voltage sampling value, temperature sampling value, internal resistance sampling value and state of charge sampling value. Based on the synchronization sequence of the individual cell voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value, a vector of length four is constructed in the order of voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value, and the corresponding individual cell identifier and sampling timestamp are recorded to generate a multi-dimensional feature vector of the individual cell. Based on the individual multidimensional feature vectors, the individual multidimensional feature vectors of all individual cells within the current BMS sampling period are aggregated, the mean vectors of the four dimensions are calculated, the average of the product of the deviations of each dimension is calculated within the individual cell range, and the matrix rows and columns are filled in a fixed dimension order to generate the battery pack multidimensional feature statistical covariance matrix.

3. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the distance score are as follows: Based on the multidimensional feature statistical covariance matrix of the battery pack and the multidimensional feature vector of the individual cell, the voltage sample value, temperature sample value, internal resistance sample value and state of charge sample value of the current sampling period are extracted according to the individual cell identifier. The corresponding components of the mean vector are subtracted from each sample value to form the individual cell difference vector. The inverse matrix of the multidimensional feature statistical covariance matrix of the battery pack is calculated to represent the cooperative offset relationship between each dimension, and the individual cell difference vector and covariance inverse matrix pair are generated. The distance score is calculated based on the individual difference vector and the covariance inverse matrix.

4. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the multidimensional outlier isolation control signal are as follows: Based on the distance score, a set critical threshold is invoked, and the distance score is compared with the critical threshold. When the distance score exceeds the critical threshold, it is determined that the corresponding single cell has a multidimensional outlier fault state. An anomaly identification flag is written and the branch disconnection logic is triggered to generate a multidimensional outlier fault isolation control signal.

5. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the terminal voltage timing data during the static phase are as follows: Based on the remaining connected branches after the execution of the multidimensional outlier fault isolation control signal, a timed current interruption is performed during the fast charging process to stop charging and maintain a preset resting time. Starting from the target time when the current returns to zero, the terminal voltage and corresponding timestamp are extracted according to the sampling period to form the terminal voltage timing data during the resting stage.

6. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the voltage relaxation rebound slope calculation value are as follows: Based on the time series data of the terminal voltage during the resting stage, the time difference and terminal voltage difference of adjacent samples are calculated in ascending order of timestamps. Samples with zero time difference are removed to obtain the linear differential sequence of the terminal voltage during the resting stage, forming a set of voltage change difference and time interval difference pairs. The voltage relaxation rebound slope is calculated based on the pairing set of voltage change difference and time interval difference.

7. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the lithium plating anomaly determination index are as follows: Based on the current temperature and state of charge values, the effective range of the current temperature and state of charge values ​​in the slope-limited dynamic threshold table is limited. The row index is located by the temperature value and the column index is located by the state of charge value. When the values ​​fall into adjacent intervals, linear interpolation is performed between adjacent entries according to the ratio of the two ends and a single value is output to generate the slope-limited dynamic threshold. Based on the slope limit dynamic threshold, the calculated value of the voltage relaxation rebound slope is read, and the difference between the slope limit dynamic threshold and the calculated value of the voltage relaxation rebound slope is calculated. When the difference is greater than zero, the risk flag is set to true and the comparison timestamp is recorded. When the difference is less than or equal to zero, the risk flag is set to false and the comparison timestamp is recorded, thereby generating a lithium plating abnormal state judgment index.

8. The battery control method of the BMS management system according to claim 1, characterized in that, The steps for obtaining the lossless boundary charging cutoff control command are as follows: According to the lithium plating abnormal state judgment index, when the risk flag is true, the charging request current level is downgraded according to the downgrade level mapping table and the cutoff trigger flag is set. When the risk flag is false, the charging request current level is maintained and the cutoff trigger flag is cleared. The execution channel field, target status field and failure protection bit are encapsulated to generate a lossless boundary charging cutoff control command.

Images