Battery internal short circuit detection method and device

Abstract

The invention discloses a battery internal short circuit detection method and device which are applied to a battery pack comprising a plurality of single batteries, and the method comprises the steps: obtaining the current information of the battery pack and the voltage information of each single battery; constructing observed quantity representing monomer difference, and estimating a first difference parameter representing monomer voltage source characteristic difference and a second difference parameter representing monomer impedance characteristic difference on line by adopting nonlinear filtering to obtain a corresponding time sequence; segmenting the time sequence according to a preset operation period, and performing weighted convergence according to weights related to electric quantity changes to obtain corresponding characteristic parameters; and constructing a first statistical discrimination quantity and a second statistical discrimination quantity which reflect extreme value deviation and fluctuation deviation based on the periodic characteristic parameters, performing joint discrimination with a threshold value, and determining an internal short circuit when abnormal monomers are consistent and meet a continuous periodic criterion. According to the method, through difference observation, filtering estimation, periodic weighting and double-discriminant-quantity combined judgment, the discriminant stability and reliability under working condition fluctuation and noise interference are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery internal short circuit detection, in particular to a battery internal short circuit detection method and device. BACKGROUND

[0002] With the rapid development of new energy vehicles, energy storage systems and portable electronic devices, lithium ion batteries are widely used due to their high energy density, long cycle life and other advantages. When the battery is used in groups, a battery management system (BMS) is usually configured to monitor and manage the voltage, current, temperature and other states of the battery to ensure the safe and stable operation of the battery system. Among them, the internal short circuit of the battery is a typical safety hazard, which often has the characteristics of strong latency and hidden development process. If it cannot be identified and measures are not taken in time, it may cause rapid degradation of battery performance and even thermal runaway and other serious consequences. Therefore, it is of great significance to detect and warn the internal short circuit of the battery online.

[0003] In related technologies, the detection of internal short circuit of the battery can usually be based on temperature abnormalities, terminal voltage abnormalities, state of charge (SOC) / open circuit voltage (OCV) deviation, equivalent internal resistance change, energy / capacity inconsistency and other characteristics to make judgments, or use equivalent circuit model (ECM) combined with parameter identification method to estimate the state of the battery and identify abnormalities accordingly. However, the battery group operates in complex conditions, the charge and discharge current fluctuates significantly, and measurement noise and sampling errors are inevitable; at the same time, factors such as individual differences, aging evolution, environmental temperature changes and load changes of the battery will cause natural drift of voltage and internal resistance and other characteristics, making the external characteristics of early internal short circuit not significant, which is easy to be confused with normal inconsistency and aging characteristics. In addition, existing methods often face problems such as insufficient stability of parameter estimation, sensitivity to dynamic conditions, difficulty in uniform calibration of detection thresholds, and difficulty in balancing false positives and false negatives in actual application, resulting in room for improvement in the reliability of early identification of internal short circuit and engineering applicability.

[0004] Therefore, in the online detection and early warning of internal short circuit of the battery group, the working condition fluctuation and noise interference are significant, the intra-group consistency difference and aging drift are difficult to distinguish, and the stability of feature parameter extraction and discrimination is insufficient, which are technical problems that need to be solved. SUMMARY

[0005] The present application provides a battery internal short circuit detection method and device, which aims to solve the problem of insufficient stability of feature parameter extraction and discrimination in the online detection and early warning of internal short circuit of the battery group in the prior art.

[0006] In a first aspect, a battery internal short circuit detection method is provided, the method comprising: Acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell; Based on the electrical signal data, an observation is constructed to characterize the differences between individual cells. Based on this observation, a first difference parameter and a second difference parameter are estimated online to obtain a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter. The first difference parameter characterizes the differences in voltage source characteristics of individual cells, and the second difference parameter characterizes the differences in impedance characteristics of individual cells. The online estimation is achieved using nonlinear filtering. The first time series and the second time series are segmented according to a preset operating cycle. Within each operating cycle, the first time series and the second time series are weighted and converged according to the weights related to the changes in electricity within the operating cycle, so as to obtain the first cycle feature parameter corresponding to the first difference parameter and the second cycle feature parameter corresponding to the second difference parameter. Based on the first cycle characteristic parameter and the second cycle characteristic parameter, a first statistical discriminant and a second statistical discriminant are constructed for internal short circuit detection. The first statistical discriminant is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discriminant is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles. The first statistical discrimination value is combined with the first threshold, and the second statistical discrimination value is combined with the second threshold. When the abnormal cells corresponding to the first statistical discrimination value and the second statistical discrimination value are consistent and the preset continuous period criterion is met, it is determined that the cell has an internal short circuit.

[0007] In the above scheme, optionally, the observation used to characterize the differences between individual cells is a voltage difference observation, which is calculated separately for each individual cell and represents the difference between the terminal voltage of the individual cell and the average terminal voltage of all individual cells in the battery pack.

[0008] In the above scheme, optionally, the first difference parameter is the single-cell open-circuit voltage difference parameter, and the second difference parameter is the single-cell ohmic internal resistance difference parameter; the online estimation is: at each sampling time, based on the voltage difference observation and the corresponding current information of the battery pack, the single-cell open-circuit voltage difference parameter and the single-cell ohmic internal resistance difference parameter are recursively estimated; wherein, the current information of the battery pack includes the main circuit charging and discharging current of the battery pack.

[0009] Optionally, in the above scheme, the model relationship used for recursively estimating the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the single cell in the online estimation is obtained by the following method: establishing a first-order RC equivalent circuit model of a single cell and establishing an average first-order RC equivalent circuit model of the single cells in the battery pack; subtracting the single cell model from the average model to form a difference model, so that the voltage difference observation is associated with the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the single cell; when forming the difference model, the difference terms related to the polarization branch are ignored or incorporated into the noise term.

[0010] Optionally, in the above scheme, the nonlinear filtering is an extended Kalman filter, and the first difference parameter and the second difference parameter are used as filter state variables for recursive estimation.

[0011] In the above scheme, optionally, the operating cycle is a complete charge-discharge cycle, which consists of a discharge stage and a charge stage; wherein, the boundary between the discharge stage and the charge stage is determined by the individual battery voltage reaching a preset discharge cutoff voltage and / or a preset charge cutoff voltage.

[0012] In the above scheme, optionally, the weights related to the changes in power within the operating cycle are determined by the cumulative charge and discharge amount obtained by integrating the current information of the battery pack in ampere-hours, and normalized by the absolute value of the cumulative charge and discharge amount so that the sum of the weights corresponding to each sampling point within the same operating cycle is 1.

[0013] Optionally, in the above scheme, the construction of the first statistical discriminant includes: forming a first set with the first cycle characteristic parameters of each individual battery cell within the same operating cycle; calculating the mean and standard deviation of the first set after removing the maximum and minimum values ​​in the first set; determining the standardized deviation of the minimum value in the first set relative to the mean as the first statistical discriminant, and determining the individual battery cell corresponding to the minimum value as the first candidate abnormal cell; The construction of the second statistical discriminant includes: using multiple consecutive operating cycles of a preset sliding window length as windows, calculating the standard deviation within the window as a fluctuation measure for the second cycle characteristic parameters of each individual battery; forming a second set based on the fluctuation measures of each individual battery within the same sliding window, calculating the mean and standard deviation of the second set after removing the maximum and minimum values ​​in the second set; determining the standardized deviation of the maximum fluctuation measure in the second set relative to the mean as the second statistical discriminant, and determining the individual battery corresponding to the maximum fluctuation measure as the second candidate abnormal individual; wherein, the preset sliding window length is 20 operating cycles.

[0014] In the above scheme, optionally, the first threshold and the second threshold are determined based on the mean and standard deviation of the first statistical discriminant sample and the second statistical discriminant sample in the historical dataset, respectively, wherein the first threshold is the mean of the first statistical discriminant minus 3 times the standard deviation, and the second threshold is the mean of the second statistical discriminant plus 3 times the standard deviation; The joint discrimination includes: when the first statistical discrimination quantity is less than the first threshold, the second statistical discrimination quantity is greater than the second threshold, and the first candidate abnormal cell and the second candidate abnormal cell are the same cell, it is determined that a single cycle criterion is met; when the single cycle criterion is met in three consecutive operating cycles, it is determined that the cell has an internal short circuit.

[0015] Secondly, a battery internal short-circuit detection device, applied to a battery pack comprising multiple individual cells, includes: The data acquisition module is used to acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell. The parameter estimation module is used to construct observations characterizing the differences between individual cells based on the electrical signal data, and to perform online estimation of a first difference parameter and a second difference parameter based on the observations, thereby obtaining a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter; wherein, the first difference parameter is used to characterize the differences in voltage source characteristics of individual cells, and the second difference parameter is used to characterize the differences in impedance characteristics of individual cells; the online estimation is implemented using nonlinear filtering; The period processing module is used to segment the first time series and the second time series according to a preset operating period, and to perform weighted aggregation on the first time series and the second time series respectively according to the weight related to the change in electricity within the operating period in each operating period, so as to obtain the first period feature parameter corresponding to the first difference parameter and the second period feature parameter corresponding to the second difference parameter. The discrimination module constructs a first statistical discrimination quantity and a second statistical discrimination quantity for internal short circuit discrimination based on the first periodic feature parameter and the second periodic feature parameter. It then performs joint discrimination by combining the first statistical discrimination quantity with a first threshold and the second statistical discrimination quantity with a second threshold. When the abnormal cells corresponding to the first statistical discrimination quantity and the second statistical discrimination quantity are consistent and a preset continuous period criterion is met, it determines that the cell has an internal short circuit. The first statistical discrimination quantity is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discrimination quantity is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles.

[0016] Compared with the prior art, this application has at least the following beneficial effects: This application, based on further analysis and research of existing technical problems, recognizes that existing technologies suffer from significant issues in online detection and early warning of short circuits within battery packs, including significant fluctuations in operating conditions and noise interference, difficulty in distinguishing between differences in intra-pack consistency and aging drift, and insufficient stability in the extraction and discrimination of characteristic parameters. By acquiring battery pack current information and individual cell voltage information, and constructing observations that characterize cell differences, this application separates common-mode changes in the battery pack under dynamic operating conditions (e.g., overall pack voltage rises / falls with overall operating conditions) from relative anomalies in individual cells. Furthermore, it performs nonlinear filtering online estimation of the first difference parameter characterizing the voltage source characteristics of individual cells and the second difference parameter characterizing the impedance characteristics of individual cells. The recursive update mechanism of the filter gradually absorbs and suppresses the influence of sampling noise, transient disturbances, and model uncertainties on the parameter sequence, making the obtained time series of the two types of difference parameters more reflective of the continuous evolution of cell differences rather than occasional fluctuations. Subsequently, the two... The time series is segmented according to the operating cycle, and weighted aggregation is performed within the cycle based on the weights related to the change in charge. This is equivalent to giving higher contributions to segments with more information and reducing the impact on segments with insufficient information within each cycle, thereby transforming the unstable estimation driven by instantaneous noise into more stable feature parameters on the cycle scale. Furthermore, this application constructs a first statistical discriminant (reflecting the extreme deviation of voltage source difference) and a second statistical discriminant (reflecting the fluctuation deviation of impedance difference in multiple cycles) based on two types of cycle feature parameters. It requires that the two both point to the same abnormal cell under the statistical benchmark within the battery pack before being jointly discriminated with the threshold, and then superimposed with the continuous cycle criterion. This is equivalent to deducing the judgment from a single feature and a single trigger to a cross-feature consistent and cross-cycle stable evidence chain: only when a cell continuously shows a significant deviation of voltage source difference in the comparison within the group and at the same time shows abnormal fluctuation of impedance difference, and this consistency is maintained in multiple cycles, will the conclusion of internal short circuit be output.

[0017] This application, under the background conditions of significant operating condition fluctuations and noise interference, easy confusion between intra-group consistency differences and aging drift, and insufficient stability of feature extraction and discrimination, improves the reliability of identifying internal short-circuit related anomalies and reduces the risk of misjudgment caused by dynamic coupling and limited sensor accuracy through a derivation chain of differential observation, robust filtering estimation, periodic weighted convergence stabilization, dual discriminants, consistency constraints, and continuous periodic confirmation. Thus, it provides an online-implementable solution to the key problems mentioned in the background art. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a battery internal short circuit detection method provided in one embodiment of this application. Figure 2This is a schematic diagram of the algorithm flow provided in one embodiment of this application; Figure 3 A schematic diagram of a first-order RC equivalent circuit model provided in one embodiment of this application; Figure 4 A schematic diagram of a first-order RC equivalent circuit model for an internal short circuit provided in one embodiment of this application; Figure 5 This is a schematic diagram of the internal short-circuit battery detection results provided in one embodiment of this application; Figure 6 This is a schematic diagram of a normal battery test result provided in one embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] Lithium-ion batteries, due to their high energy density and long cycle life, have become the core energy storage component of electric vehicles. However, lithium-ion batteries have potential safety issues, the most serious of which is thermal runaway. Under actual operating conditions without compression or other abnormal external factors, internal short circuits (ISCs) in lithium-ion batteries are considered one of the main causes of thermal runaway. Early ISCs often lack obvious external characteristics, but once they develop into severe short circuits, they can lead to irreversible safety accidents. Therefore, reliable early detection methods are urgently needed. Existing internal short circuit detection methods can be categorized as follows: one type detects internal short circuits by identifying abnormal voltage drop recovery phenomena. This type of method detects internal short circuits by identifying abnormal voltage drops and subsequent recovery phenomena. For example, some studies have proposed diagnosing ISC faults by analyzing changes in open-circuit voltage (OCV) and deriving the equivalent ISC resistance by extracting OCV characteristics. However, this method may require a high sampling frequency for transient voltage drop-recovery signals, and may fail for certain types of ISCs (such as copper ISCs) due to insufficient voltage sampling frequency of the Battery Management System (BMS). Additionally, the pseudo-open-circuit voltage difference method can detect the ISC of individual cells under dynamic operating conditions, but requires an offline R0-SOC lookup table. Another approach is to detect internal short circuits based on battery pack consistency. This involves monitoring the consistency (such as voltage, capacity, or internal resistance) between individual cells within the battery pack to identify internal short circuits. For example, algorithms based on cycle current detection can promptly detect internal short circuits in any cell within the battery pack. Differential voltage analysis and Mahalanobis distance methods have also been used for early diagnosis and quantitative assessment of battery pack ISCs. Furthermore, some studies have proposed using multistage analysis (MSA) to detect internal short circuits in battery packs with inconsistent State of Health (SOH). However, these methods often rely on other healthy cells in the battery pack as a reference, requiring high overall battery pack consistency, and may be insensitive to small changes in early, soft ISCs. Thirdly, detection is based on battery expansion measurement: internal short circuits are detected by measuring battery expansion, but this method usually requires additional physical sensors. Fourthly, detection is based on electro-thermal coupled models: ISCs are detected and fault severity is classified by constructing an electro-thermal coupled model (ETCM) and analyzing voltage and surface temperature changes. However, the model building and parameter calibration of such methods are relatively complex. This embodiment proposes an early detection method for battery internal short circuits based on a three-level detection architecture of "nonlinear filter parameter estimation + cyclic weighted average parameter extraction + statistical significance judgment," aiming to overcome the shortcomings of the above-mentioned existing methods and fill the following gaps.

[0021] First, it offers sensitive detection of subtle early parameter changes: Traditional internal short-circuit detection methods often struggle to accurately capture the minute but significant parameter changes caused by early internal short circuits. This method, through nonlinear filtering parameter estimation, can more sensitively identify subtle abnormal fluctuations in battery parameters (such as internal resistance or open-circuit voltage). Nonlinear filtering (e.g., extended Kalman filter, EKF) has proven effective in handling model nonlinearity and measurement noise in battery state estimation (such as state of charge, SOC). Applying it to parameter estimation allows for real-time, high-precision estimation of parameters such as resistance and capacitance in the battery equivalent circuit model (ECM), thereby improving the detection accuracy of early ISC.

[0022] Secondly, this method improves the robustness of detection: Existing methods lack sufficient analysis of battery consistency when dealing with individual battery differences and noise, which may lead to false alarms or missed alarms. This method introduces cyclic weighted average parameter extraction and statistical significance judgment, significantly improving the robustness of detection. Cyclic weighted average parameter extraction: By performing cyclic weighted averaging on the estimated parameters over a long time window, instantaneous noise and random fluctuations can be effectively filtered out, extracting a stable trend that reflects the battery's health status. This makes the detection results less susceptible to the influence of occasional measurement errors or operating condition fluctuations, improving the reliability of the judgment. Statistical significance judgment: Combining statistical methods to perform significance testing on the extracted average parameters can distinguish between true parameter changes caused by ISC and minor fluctuations caused by normal operation or aging. This helps avoid misjudging normal battery degradation (such as increased internal resistance) as ISC faults, thereby reducing the false alarm rate.

[0023] Third, it avoids over-reliance on external stress: Many traditional methods may rely on significant changes in battery voltage or capacity, or trigger and detect ISC by applying external stimuli (such as mechanical abuse, electrical abuse, or thermal abuse), which is often not feasible in actual operation. This method comprehensively analyzes multiple parameters of the equivalent circuit model and combines the mean difference model (MDM) to perform real-time comparison and difference analysis of the parameters of each cell or module in the battery pack, thus effectively identifying ISC without the need for specific external stress. MDM can capture subtle differences between battery cells or between the battery cell and the benchmark model, which may be more significant than absolute voltage or capacity changes in the early stages of ISC. This allows the method to extract key characteristic parameters from complex and variable operating signals for comprehensive judgment, achieving more comprehensive early fault identification.

[0024] In summary, this three-level detection architecture, by integrating nonlinear filtering, cyclic weighted average parameter extraction, and statistical significance judgment, and combining equivalent circuit models and mean difference models, fills the gaps in existing methods in terms of improving the accuracy of early detection of internal short circuits, enhancing robustness, and reducing dependence on specific external conditions. This provides a more reliable guarantee for the safe operation of lithium-ion battery systems. Internal short circuits in lithium-ion batteries are one of the main causes of thermal runaway. The method in this embodiment enables early, soft short circuit diagnosis, which is of great significance for improving the safety of lithium batteries.

[0025] In one embodiment, such as Figure 1 As shown, a method for detecting internal short circuits in a battery is provided, including the following steps: Acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell; Based on the electrical signal data, an observation is constructed to characterize the differences between individual cells. Based on this observation, a first difference parameter and a second difference parameter are estimated online to obtain a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter. The first difference parameter characterizes the differences in voltage source characteristics of individual cells, and the second difference parameter characterizes the differences in impedance characteristics of individual cells. The online estimation is achieved using nonlinear filtering. The first time series and the second time series are segmented according to a preset operating cycle. Within each operating cycle, the first time series and the second time series are weighted and converged according to the weights related to the changes in electricity within the operating cycle, so as to obtain the first cycle feature parameter corresponding to the first difference parameter and the second cycle feature parameter corresponding to the second difference parameter. Based on the first cycle characteristic parameter and the second cycle characteristic parameter, a first statistical discriminant and a second statistical discriminant are constructed for internal short circuit detection. The first statistical discriminant is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discriminant is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles. The first statistical discrimination value is combined with the first threshold, and the second statistical discrimination value is combined with the second threshold. When the abnormal cells corresponding to the first statistical discrimination value and the second statistical discrimination value are consistent and the preset continuous period criterion is met, it is determined that the cell has an internal short circuit.

[0026] The battery internal short-circuit detection method of this embodiment is applicable to battery packs containing multiple individual cells, such as power battery packs or energy storage battery clusters. This method is executed by the Battery Management System (BMS) or an external controller during battery pack operation. It collects battery pack current information and individual cell voltage information to form a data basis for subsequent estimation and judgment. The current and voltage information can be collected synchronously according to a fixed sampling period, which can be set according to the system's computing power and data refresh rate.

[0027] This embodiment constructs an observational system based on collected electrical signal data to characterize the differences between individual cells. Subsequently, for each individual cell, the observational system is used to estimate a first difference parameter and a second difference parameter online. The first difference parameter characterizes the differences in voltage source characteristics of the individual cell (e.g., differences related to equivalent open-circuit voltage), and the second difference parameter characterizes the differences in impedance characteristics of the individual cell (e.g., differences related to equivalent ohmic internal resistance). The online estimation is implemented using nonlinear filtering, which updates the estimated value of the difference parameter at each sampling time based on the observational system and the current input, thereby obtaining a first time series of the first difference parameter and a second time series of the second difference parameter.

[0028] In this embodiment, the two time series are segmented according to a preset operating cycle. The operating cycle can be selected from a period closely related to the changes in the battery's electrochemical state (e.g., a complete charge-discharge cycle). Within each operating cycle, the first and second time series are weighted and converged according to the weights associated with the changes in charge capacity during that cycle, resulting in a first cycle characteristic parameter corresponding to a first difference parameter and a second cycle characteristic parameter corresponding to a second difference parameter. The weights can depend on the cumulative charge-discharge capacity or its normalized value within the cycle to ensure that the contribution of different sampling points to the cycle characteristics is adjustable.

[0029] This embodiment constructs a first statistical discriminant and a second statistical discriminant based on the first cycle characteristic parameters and the second cycle characteristic parameters. The first statistical discriminant is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and can be used to locate individual cells that exhibit abnormal extreme values ​​within the pack. The second statistical discriminant is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles, and can be used to locate individual cells that exhibit abnormal fluctuations in multiple cycle dimensions.

[0030] This embodiment uses a first statistical discriminant quantity and a first threshold, and a second statistical discriminant quantity and a second threshold for joint discrimination. When the abnormal cells pointed to by the two statistical discriminant quantities are consistent and a preset continuous cycle criterion is met, it is determined that the cell has an internal short circuit. The continuous cycle criterion can be set to meet the cycle criterion once for several consecutive operating cycles, thereby avoiding misjudgments caused by occasional disturbances in a single cycle.

[0031] This embodiment employs a combined mechanism of difference observations, online estimation using nonlinear filtering, periodic weighted convergence, joint discrimination using dual statistical discriminants, and continuous periodic confirmation. This method can still obtain stable difference characteristics that can be used for intra-group comparison even in the presence of operating condition fluctuations and noise. Furthermore, it improves the reliability and consistency of internal short-circuit discrimination through the synergistic constraint of extreme value deviation and fluctuation deviation.

[0032] In this embodiment, the observation used to characterize the differences between individual cells is a voltage difference observation. The voltage difference observation is calculated for each individual cell and represents the difference between the terminal voltage of the individual cell and the average terminal voltage of all individual cells in the battery pack.

[0033] In this embodiment, the observed quantity is specifically defined as the voltage difference observed quantity. The voltage difference observed quantity is calculated for each individual cell in the battery pack. The calculation uses the individual cell terminal voltage data obtained at the same sampling time to ensure that the voltage difference reflects the relative difference within the pack at the same moment rather than the deviation caused by time misalignment.

[0034] The voltage difference observation represents the difference between the terminal voltage of a single cell and the average terminal voltage of all cells in the battery pack. The average value can be obtained by summing the voltages of all cells by the BMS and dividing by the number of cells. For single cell voltage data with missing measurements, communication abnormalities, or obvious outliers, a validity check can be performed in the engineering implementation to remove invalid data before calculating the average value, so as to ensure that the average value is representative (this removal is optional and does not change the definition of the observation "difference between the single cell and the average value of the pack").

[0035] Using the difference between the individual terminal voltage and the average terminal voltage within the group as an observation, the consistency difference within the group can be directly characterized, and the influence of common mode changes such as the overall rise / fall of the package voltage on the estimation can be weakened, so that the subsequent difference parameter estimation can focus more on the relative anomalies of the individual units.

[0036] In this embodiment, the first difference parameter is the single-cell open-circuit voltage difference parameter, and the second difference parameter is the single-cell ohmic internal resistance difference parameter; the online estimation is: at each sampling time, based on the voltage difference observation and the corresponding current information of the battery pack, the single-cell open-circuit voltage difference parameter and the single-cell ohmic internal resistance difference parameter are recursively estimated; wherein, the current information of the battery pack includes the main circuit charging and discharging current of the battery pack.

[0037] In this embodiment, the first difference parameter is defined as the individual cell open-circuit voltage difference parameter, used to characterize the difference between individual cells at the equivalent voltage source level; the second difference parameter is defined as the individual cell ohmic internal resistance difference parameter, used to characterize the difference between individual cells at the equivalent ohmic impedance level. The aforementioned difference parameters can be understood as the deviation of the individual cell parameter from the group average parameter or the difference quantity related to the group statistical benchmark, so as to establish a correspondence with the voltage difference observation in the above embodiment.

[0038] Online estimation is performed recursively at each sampling time: using the voltage difference observation at that sampling time and the battery pack current information at the corresponding sampling time as input, the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of each cell are jointly recursively estimated. The battery pack current information includes the charging and discharging current of the battery pack main circuit, which can be obtained from the BMS current sampling circuit and aligned with the voltage sampling and holding time or interpolated and aligned at the software layer.

[0039] By limiting the estimation objects to open-circuit voltage difference and ohmic internal resistance difference, and recursively updating them at each sampling time, the intra-group difference in voltage difference observation can be decomposed into two dimensions: voltage source difference and impedance difference, providing structured input for subsequent joint discrimination based on extreme value deviation and fluctuation deviation.

[0040] In this embodiment, the model relationship used in the online estimation to recursively estimate the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the individual cells is obtained in the following way: a first-order RC equivalent circuit model of the individual cell is established and an average first-order RC equivalent circuit model of the individual cells in the battery pack is established. The difference between the individual cell model and the average model is calculated to form a difference model, so that the voltage difference observation is associated with the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the individual cells. When forming the difference model, the difference terms related to the polarization branch are ignored or incorporated into the noise term.

[0041] In this embodiment, the model relationships used for recursive estimation are obtained through differential modeling: First, a first-order RC equivalent circuit model of a single cell is established to describe the relationship between the cell's terminal voltage, open-circuit voltage, ohmic internal resistance, and polarization branch; simultaneously, an average first-order RC equivalent circuit model of the cells within the battery pack is established to describe the relationship between the average terminal voltage and average parameters within the pack. Then, the difference between the single-cell model and the average model is calculated to form a differential model, establishing a correspondence between the voltage difference observation and the single-cell open-circuit voltage difference parameters and the single-cell ohmic internal resistance difference parameters that can be used for estimation.

[0042] When formulating the difference model, the difference term related to the polarization branch can be implemented in two ways: First, the polarization branch difference term can be directly ignored, which is suitable for cases where the polarization difference is relatively small or the estimation window is long; second, the polarization branch difference term can be incorporated into the noise term and treated as part of the model uncertainty in the noise modeling of the filter, thus eliminating the need for explicit estimation of the polarization difference parameters. Both methods can maintain the difference model's focus on estimating the open-circuit voltage difference and the ohmic internal resistance difference.

[0043] By employing a difference modeling approach that compares the individual model with the average model, the voltage difference observation can be directly correlated with the difference parameters. At the same time, by ignoring or incorporating noise into the polarization difference term, the model complexity can be reduced and the number of parameters that need to be identified online can be decreased, thus facilitating online recursion.

[0044] In this embodiment, the nonlinear filtering is an extended Kalman filter, and the first difference parameter and the second difference parameter are used as filter state variables for recursive estimation.

[0045] In this embodiment, the nonlinear filtering is specifically implemented using an Extended Kalman Filter (EKF). The filtered state variables are selected as a first difference parameter and a second difference parameter (corresponding to the single-cell open-circuit voltage difference parameter and the single-cell ohmic internal resistance difference parameter under the limitations of the above embodiment). The EKF includes two stages: prediction and update. In the prediction stage, the state variables are updated a priori based on the assumption of slow parameter variation. In the update stage, an observation update is constructed based on the voltage difference observation and the battery pack current input, and the state variables are corrected.

[0046] Process noise and observation noise can be obtained from empirical values, offline calibration, or online adaptive methods. Process noise is used to characterize the uncertainty of the difference parameter changing slowly over time, while observation noise is used to characterize voltage measurement errors, difference model simplification errors, etc. In engineering implementation, upper and lower limits can be set for the noise covariance matrix to avoid filter divergence; alternatively, strategies to maintain or reduce the update gain can be adopted under conditions where the current is close to zero and the amount of information is insufficient (this is an optional implementation).

[0047] This embodiment uses EKF to recursively estimate the difference parameters, which can smoothly track the parameters under noise and operating condition fluctuations, making the obtained difference parameter time series more suitable for subsequent periodic feature extraction and statistical discrimination.

[0048] In this embodiment, the operating cycle is a complete charge-discharge cycle, which consists of a discharge phase and a charge phase; wherein, the boundary between the discharge phase and the charge phase is determined by the individual battery voltage reaching a preset discharge cutoff voltage and / or a preset charge cutoff voltage.

[0049] In this embodiment, the operating cycle is defined as a complete charge-discharge cycle, which consists of one discharge phase and one charge phase. The boundary between the discharge phase and the charge phase can be determined by the individual cell voltage reaching a preset discharge cutoff voltage and / or a preset charge cutoff voltage. For example, when the voltage of any individual cell under any specified strategy reaches a cutoff threshold, the corresponding phase is considered to have ended and the next phase begins, thus determining the cycle boundary.

[0050] For battery packs with a large number of individual cells, strategies such as determining the boundary based on the cell that reaches the cutoff threshold first, the average voltage reaching the threshold, or a certain percentage of cells reaching the threshold can be used to avoid frequent fluctuations in cycle division caused by individual anomalies. For repeated threshold crossings caused by voltage sampling fluctuations, hysteresis judgment or continuous satisfaction of time criteria can be used to stabilize the cycle boundary.

[0051] Using a complete charge-discharge cycle as the operating period is beneficial for extracting cycle characteristic parameters after the battery state has undergone a relatively complete change process. This makes the convergence results of the difference parameters more consistent with the actual operation of the battery and facilitates cross-cycle statistical comparisons.

[0052] In this embodiment, the weights related to the changes in charge within the operating cycle are determined by the cumulative charge and discharge amount obtained by integrating the current information of the battery pack in ampere-hours, and normalized using the absolute value of the cumulative charge and discharge amount so that the sum of the weights corresponding to each sampling point within the same operating cycle is 1.

[0053] In this embodiment, the weights are determined by the cumulative charge and discharge quantities obtained by integrating the battery pack current information in ampere-hours. The ampere-hour integration accumulates the main circuit charge and discharge current over time during the operating cycle, obtaining the cumulative charge change corresponding to each sampling point within the cycle. The cumulative quantity can be achieved by discrete summation, taking into account the sampling period.

[0054] To ensure that the weights reflect the degree of change in charge level without being affected by the sign of the charging / discharging current, the absolute value of the cumulative charge / discharge amount is normalized so that the sum of the weights of all sampling points within the same operating cycle is 1. Subsequently, these weights are used to weight and converge the first and second time series respectively, forming the first-cycle characteristic parameters and the second-cycle characteristic parameters. In engineering implementation, a minimum lower limit can be set for the weights to prevent instability in weight values ​​due to extremely small currents.

[0055] By using weights that are correlated with and normalized to the cumulative charge and discharge amount for periodic aggregation, the periodic characteristic parameters can better reflect the differences in battery capacity during significant ranges, thereby improving the representativeness and consistency of the periodic characteristic parameters.

[0056] In this embodiment, the construction of the first statistical discriminant includes: forming a first set with the first cycle characteristic parameters of each individual battery cell within the same operating cycle; calculating the mean and standard deviation of the first set after removing the maximum and minimum values ​​in the first set; determining the standardized deviation of the minimum value in the first set relative to the mean as the first statistical discriminant; and determining the individual battery cell corresponding to the minimum value as the first candidate abnormal cell.

[0057] The construction of the second statistical discriminant includes: using multiple consecutive operating cycles of a preset sliding window length as windows, calculating the standard deviation within the window as a fluctuation measure for the second cycle characteristic parameters of each individual battery; forming a second set based on the fluctuation measures of each individual battery within the same sliding window, calculating the mean and standard deviation of the second set after removing the maximum and minimum values ​​in the second set; determining the standardized deviation of the maximum fluctuation measure in the second set relative to the mean as the second statistical discriminant, and determining the individual battery corresponding to the maximum fluctuation measure as the second candidate abnormal individual; wherein, the preset sliding window length is 20 operating cycles.

[0058] Within the same operating cycle, the first-cycle characteristic parameters of all individual cells are collected to form a first set. To improve the robustness of the statistical benchmark within the group, the maximum and minimum values ​​in the first set are first removed, and then the mean and standard deviation are calculated for the remaining data. Subsequently, the standardized deviation of the minimum value of the first set from the aforementioned mean is taken as the first statistical discriminant, and the individual cell that produces this minimum value is marked as the first candidate anomalous cell. This implements the method for constructing the corresponding extreme value deviation discriminant.

[0059] Using multiple consecutive operating cycles of a preset sliding window length as windows, the standard deviation of the second-cycle characteristic parameters of each individual cell is calculated within the window as a fluctuation metric. Subsequently, within the same sliding window, the fluctuation metrics of each individual cell are aggregated to form a second set. Similarly, after removing the maximum and minimum values, the mean and standard deviation are calculated. The standardized deviation of the maximum fluctuation metric in the second set relative to the aforementioned mean is taken as the second statistical discriminant, and the individual cell generating this maximum fluctuation metric is marked as the second candidate anomalous cell. The sliding window length is set to 20 operating cycles; the window can be updated periodically.

[0060] In the initial stage of system startup or when the number of effective running cycles is less than 20, the second statistical discriminant can be not output if the window is not full, or a short window can be formed using the current number of cycles and marked as a warning state, in order to ensure the consistency between the calculation definition and the engineering output.

[0061] This embodiment uses a dual statistical method of extreme value deviation discriminant and multi-period fluctuation deviation discriminant to characterize the abnormal single-entity features from the perspectives of intra-group horizontal differences and cross-period vertical fluctuations, providing more sufficient discriminant basis for subsequent joint discrimination.

[0062] In this embodiment, the first threshold and the second threshold are determined based on the mean and standard deviation of the first statistical discriminant sample and the second statistical discriminant sample in the historical dataset, respectively. The first threshold is the mean of the first statistical discriminant minus 3 times the standard deviation, and the second threshold is the mean of the second statistical discriminant plus 3 times the standard deviation.

[0063] The joint discrimination includes: when the first statistical discrimination quantity is less than the first threshold, the second statistical discrimination quantity is greater than the second threshold, and the first candidate abnormal cell and the second candidate abnormal cell are the same cell, it is determined that a single cycle criterion is met; when the single cycle criterion is met in three consecutive operating cycles, it is determined that the cell has an internal short circuit.

[0064] In this embodiment, the first threshold and the second threshold are obtained based on historical datasets. The historical datasets can be derived from operational data of the same battery pack under different operating conditions, or from production testing and road condition data. The mean and standard deviation of the first statistical discriminant sample in the historical dataset are calculated, and the mean and standard deviation of the second statistical discriminant sample are calculated to form the statistics required for threshold calibration.

[0065] The first threshold is determined by subtracting three standard deviations from the mean of the first statistical discriminant, and the second threshold is determined by adding three standard deviations to the mean of the second statistical discriminant. Joint discrimination includes: when the first statistical discriminant is less than the first threshold, the second statistical discriminant is greater than the second threshold, and the first candidate abnormal cell and the second candidate abnormal cell are the same cell, a single cycle criterion is satisfied.

[0066] To achieve the final determination, a continuous cycle criterion is further introduced: when the above single-cycle criterion is true for three consecutive operating cycles, it is determined that the single cell has an internal short circuit. In the engineering implementation, a counter can be set to record the number of consecutive true cycles. When a cycle fails to meet the criterion, the counter is reset or decremented to meet the determination logic of three consecutive cycles.

[0067] This embodiment uses a combination of a 3σ threshold formed by historical sample statistics and a criterion confirmed over three consecutive periods to distinguish between occasional disturbances and persistent anomalies in a single period, thereby improving the reusability of the discrimination threshold and the stability of the judgment logic.

[0068] In one embodiment, a battery internal short circuit early detection method employing a three-level detection architecture—nonlinear filtering parameter estimation, cyclic weighted average parameter extraction, and statistical significance judgment—is described below: A three-level detection architecture is constructed, consisting of "nonlinear filtering parameter estimation + cyclic weighted average parameter extraction + statistical significance judgment." The correlation principle between characteristic parameters and internal short circuits is as follows: When an internal short circuit occurs in a battery, a continuous micro-current discharge occurs locally, leading to irreversible changes such as electrode active material loss and electrolyte decomposition. These changes are directly reflected in the equivalent circuit parameters: ΔE (voltage difference) corresponds to the difference in the battery's open-circuit voltage; an internal short circuit will lower the open-circuit voltage, causing ΔE to deviate negatively; ΔR (internal resistance difference) corresponds to the sum of the battery's ohmic internal resistance and polarization internal resistance; an internal short circuit will cause local heating and increased interface impedance, causing ΔR to increase positively. By monitoring the changing trends of these two core parameters, early identification of internal short circuit faults can be achieved.

[0069] The instantaneous characteristic parameters (ΔE, ΔR) at each sampling time are estimated using a nonlinear filtering algorithm. Using a complete charge-discharge cycle as a unit, the average value of the instantaneous parameters within the cycle is calculated to obtain the cycle-weighted average characteristic parameters; Based on battery pack consistency, statistical methods are used to analyze the deviation of the cyclic weighted average parameters to determine whether an ISC fault exists.

[0070] In one embodiment, an internal short-circuit detection method for a lithium battery pack is provided, the algorithm flow of which is as follows: Figure 2 As shown, this embodiment uses a first-order RC equivalent circuit model (ECM) to describe the dynamic characteristics of a single electrical switch. This model achieves a balance between accuracy and computational complexity, making it suitable for real-time estimation scenarios in BMS. Figure 3 and Figure 4 As shown, the model structure includes: Open circuit voltage source It reflects the electrochemical equilibrium potential of the battery and has a non-linear relationship with SOC (State of Charge). Ohmic resistance It mainly consists of electrolyte resistance, electrode material resistance, and current collector contact resistance; polarization resistor With polarization capacitor Parallel networks: describe the electrochemical polarization and concentration polarization effects of batteries; and This indicates the voltage and current measured by the BMS.

[0071] The derivation of the terminal voltage equation for a single battery cell is as follows: .

[0072] in The polarization voltage satisfies the dynamic equation of the RC circuit: .

[0073] Based on the mean difference model (MDM), the average terminal voltage equations of all cells in the battery pack are averaged to obtain the pack-average terminal voltage equation: .

[0074] The difference parameter is defined by subtracting the equation for a single cell from the equation for the group average. Voltage difference:

[0075] Open circuit voltage difference:

[0076] Internal resistance difference:

[0077] Polarization voltage difference:

[0078] Ignoring the dynamic changes in polarization voltage difference (because the polarization characteristics of the cells within the battery pack are relatively consistent), With smaller amplitudes and slower changes, a simplified battery differential model (CDM) voltage equation is ultimately obtained, which serves as the core equation for parameter estimation: .

[0079] When an internal short circuit occurs in a battery cell, its equivalent circuit needs to incorporate an internal short-circuit resistor. (Connected in parallel across the original equivalent circuit), at this time the actual terminal voltage of the battery cell and equivalent internal resistance Things have changed.

[0080] Terminal voltage change: Internal short circuit causes continuous discharge of the cell, and the active material of the electrode participates in side reactions, open circuit voltage. Decrease, let the decrease be . , .

[0081] Equivalent internal resistance change: internal short-circuit resistance With the original ohmic internal resistance The parallel connection is formed, and the localized heating caused by the short circuit leads to an increase in interface impedance. The final equivalent internal resistance... :

[0082] in, This is the increase in internal resistance caused by heat generation.

[0083] Difference parameter changes: Substituting the parameters of the short-circuited cell into the parameter estimation model, at this time: , The change depends on and The size. When no internal short circuit occurs. , , .

[0084] The Extended Kalman Filter (EKF) is used as the core parameter estimation algorithm, and its specific implementation is as follows: Based on the battery differential model (CDM) of MDM, the state equation and observation equation are established: Equations of state: ,in It is a state vector (voltage difference, internal resistance difference). It is an identity matrix (assuming the parameters change slowly). This is process noise.

[0085] Observation equation: ,in The observed value is the difference between the battery voltage and the group average voltage. This is the charging and discharging current. To observe noise.

[0086] Through the prediction-update iteration process of EKF, real-time estimation of ΔE and ΔR in nonlinear systems is achieved, adapting to the nonlinear characteristics of battery electrochemical processes.

[0087] Charge / discharge cycle segmentation: Each charge / discharge cycle is divided into segments for parameter estimation. The start and end points of the cycle are based on the battery (ternary lithium battery) voltage reaching 2.9V (discharge cutoff) and 4.1V (charge cutoff).

[0088] Weighted average principle: The larger the charge / discharge capacity, the more significant the change in the battery's electrochemical state, and the more accurately the estimated parameters reflect the true state. Therefore, the weighting coefficient is proportional to the cumulative charge / discharge capacity at that moment.

[0089] Weighted average calculation method: Suppose there are a total of [number] charges / discharge cycles in a certain period of time. At the sampling time, the first The EKF estimate at time t is , The cumulative charge and discharge capacity is (Unit: Ah, positive during discharge, negative during charging, initial) ),but: Cumulative charge / discharge capacity calculation: ,in For the first Current at any given moment (A) The sampling interval is (h).

[0090] Weighting coefficient calculation: ,satisfy .

[0091] Weighted cyclic average voltage difference: .

[0092] Weighted average internal resistance difference: .

[0093] Internal Short Circuit (ISC) Detection Logic: When an internal short circuit occurs in the battery cell: , The change depends on and The magnitude and direction of the change are uncertain, but its fluctuation will increase. Therefore, it exhibits a synergistic change of "negative deviation of ΔE + increased fluctuation of ΔR". The detection logic needs to be implemented through four steps: multi-index quantification, threshold determination, and joint confirmation. This eliminates interference from normal aging and measurement noise while ensuring accurate early ISC identification. The detailed detection logic is as follows.

[0094] Negative significance detection: Based on "parameter changes during internal short circuit", ISC causes partial micro-discharge in the cell, leading to loss of electrode active materials and electrolyte decomposition, ultimately reducing the open-circuit voltage; while It is a "weighted average of the differences between the cell open-circuit voltage and the group average open-circuit voltage", therefore, the ISC cell's... It will show a significant negative deviation. The specific steps are as follows: Parameter set construction: Collect the target cycle-weighted average parameters of all cells in the battery pack. This constitutes the parameter set: ,in: This represents the total number of battery cells in the battery pack. For the first Cycle-weighted average parameters of individual cells .

[0095] Statistical calculation: Remove the maximum and minimum values ​​of the cycle-averaged parameters in the battery pack to obtain a new set of parameters. Calculate the mean of the remaining parameters. and standard deviation .

[0096] Detection index formula: Take set The smallest parameter value (i.e., the cell most likely to have an ISC) Value), define the negative significance index: .

[0097] Positive significance test of fluctuation: When ISC occurs, two factors lead to... Increased volatility: Internal short-circuit resistor With the original ohmic internal resistance Parallel connection makes the equivalent internal resistance It varies with the degree of short circuit; Short circuit localized heating leads to an increase in interface impedance Unstable.

[0098] Normally aging battery cells The increase is only slow (with small fluctuations), therefore, ISC can be distinguished from normal aging by the "parameter fluctuation amplitude". The specific calculation steps are as follows: Fluctuation statistics window definition: A fixed sliding window is defined as "20 consecutive charge-discharge cycles" (window length selection criteria: too short and it is easily affected by single-cycle noise interference; too long and it delays detection). For each battery cell, calculate 20 cycles within the calculation window. Standard deviation .

[0099] Parameter set construction: Collect parameters of all cells in the battery pack. This constitutes the parameter set: ,in: This represents the total number of battery cells in the battery pack. For the first The first 20 windows of each cell Standard deviation .

[0100] Statistical calculation: Remove the maximum and minimum values ​​of the cycle-averaged parameters in the battery pack to obtain a new set of parameters. Calculate the mean of the remaining parameters. and standard deviation .

[0101] Detection index formula: Take set The largest parameter value (i.e., the cell most likely to have an ISC) Value), define the negative significance index: .

[0102] The method for determining the threshold is based on the "law of large numbers," which is the sum of a sufficiently large number of batteries. and The overall distributions of these data points approximate a normal distribution. The threshold is determined using the 3σ criterion (99.73% of the data fall within this range). Within this range (if it exceeds this range, it is considered abnormal), ensuring the reliability and engineering applicability of the threshold. The specific calculation steps are as follows: Big data collection: Data from 10,000 battery packs were selected, and calculations were performed according to methods 6.1-6.3. and To form an indicator dataset , .

[0103] Mean and variance calculation: Then calculate the dataset. mean and standard deviation , mean and standard deviation .

[0104] Threshold determination: Threshold ,when At that time, it was considered that there was a possibility of an internal short circuit; Threshold ,when At that time, it was considered that there was a possibility of an internal short circuit.

[0105] Multiple indicators combined for confirmation: Preliminary judgment condition: In a single charge-discharge cycle, a certain cell simultaneously meets the following conditions. and If so, it is preliminarily determined to be a suspected ISC battery cell.

[0106] Final judgment criteria: If the initial judgment criteria are met for three consecutive charge-discharge cycles, the final judgment is an ISC fault.

[0107] Exclusion criteria: If only a single indicator threshold is met or there is no continuous satisfaction, it is judged as a normal state (which may be due to normal aging or transient operating condition interference), and the changes in subsequent cycle parameters need to be continuously monitored.

[0108] In this embodiment, as Figure 5 and Figure 6 As shown, the six subgraphs are as follows: Curve graph Curve graph, standardized after a single charge-discharge cycle Curve graph, standardized after a single charge-discharge cycle Curve graph Curve graph and The graph (the first 19 values ​​are filled with 0 because a sliding window of 20 is used). Figure 5 and Figure 6 These are normal battery test results; no abnormalities were triggered.

[0109] In one embodiment, a battery internal short circuit detection device is provided, comprising: The data acquisition module is used to acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell. The parameter estimation module is used to construct observations characterizing the differences between individual cells based on the electrical signal data, and to perform online estimation of a first difference parameter and a second difference parameter based on the observations, thereby obtaining a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter; wherein, the first difference parameter is used to characterize the differences in voltage source characteristics of individual cells, and the second difference parameter is used to characterize the differences in impedance characteristics of individual cells; the online estimation is implemented using nonlinear filtering; The period processing module is used to segment the first time series and the second time series according to a preset operating period, and to perform weighted aggregation on the first time series and the second time series respectively according to the weight related to the change in electricity within the operating period in each operating period, so as to obtain the first period feature parameter corresponding to the first difference parameter and the second period feature parameter corresponding to the second difference parameter. The discrimination module constructs a first statistical discrimination quantity and a second statistical discrimination quantity for internal short circuit discrimination based on the first periodic feature parameter and the second periodic feature parameter. It then performs joint discrimination by combining the first statistical discrimination quantity with a first threshold and the second statistical discrimination quantity with a second threshold. When the abnormal cells corresponding to the first statistical discrimination quantity and the second statistical discrimination quantity are consistent and a preset continuous period criterion is met, it determines that the cell has an internal short circuit. The first statistical discrimination quantity is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discrimination quantity is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles.

[0110] The specific implementation details of each module can be found in the above description of the limitations of the battery internal short circuit detection method, and will not be repeated here.

[0111] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for detecting internal short circuits in a battery, characterized in that, The method includes: Acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell; Based on the electrical signal data, an observation is constructed to characterize the differences between individual cells. Based on this observation, a first difference parameter and a second difference parameter are estimated online to obtain a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter. The first difference parameter characterizes the differences in voltage source characteristics of individual cells, and the second difference parameter characterizes the differences in impedance characteristics of individual cells. The online estimation is achieved using nonlinear filtering. The first time series and the second time series are segmented according to a preset operating cycle. Within each operating cycle, the first time series and the second time series are weighted and converged according to the weights related to the changes in electricity within the operating cycle, so as to obtain the first cycle feature parameter corresponding to the first difference parameter and the second cycle feature parameter corresponding to the second difference parameter. Based on the first cycle characteristic parameter and the second cycle characteristic parameter, a first statistical discriminant and a second statistical discriminant are constructed for internal short circuit detection. The first statistical discriminant is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discriminant is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles. The first statistical discrimination value is combined with the first threshold, and the second statistical discrimination value is combined with the second threshold. When the abnormal cells corresponding to the first statistical discrimination value and the second statistical discrimination value are consistent and the preset continuous period criterion is met, it is determined that the cell has an internal short circuit.

2. The battery internal short circuit detection method according to claim 1, characterized in that, The observation used to characterize the differences between individual cells is a voltage difference observation, which is calculated for each individual cell and represents the difference between the terminal voltage of the individual cell and the average terminal voltage of all individual cells in the battery pack.

3. The battery internal short circuit detection method according to claim 2, characterized in that, The first difference parameter is the single-cell open-circuit voltage difference parameter, and the second difference parameter is the single-cell ohmic internal resistance difference parameter; the online estimation is: at each sampling time, based on the voltage difference observation and the corresponding current information of the battery pack, the single-cell open-circuit voltage difference parameter and the single-cell ohmic internal resistance difference parameter are recursively estimated; wherein, the current information of the battery pack includes the main circuit charging and discharging current of the battery pack.

4. The battery internal short circuit detection method according to claim 3, characterized in that, The model relationship used in the online estimation to recursively estimate the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the individual cells is obtained in the following way: a first-order RC equivalent circuit model of the individual cell is established and an average first-order RC equivalent circuit model of the individual cells in the battery pack is established. The difference between the individual cell model and the average model is calculated to form a difference model, so that the voltage difference observation is associated with the open-circuit voltage difference parameter and the ohmic internal resistance difference parameter of the individual cells. When forming the difference model, the difference terms related to the polarization branch are ignored or incorporated into the noise term.

5. The battery internal short circuit detection method according to claim 1, characterized in that, The nonlinear filtering is an extended Kalman filter, which uses the first difference parameter and the second difference parameter as filter state variables for recursive estimation.

6. The battery internal short circuit detection method according to claim 1, characterized in that, The operating cycle is a complete charge-discharge cycle, which consists of a discharge phase and a charge phase. The boundary between the discharge phase and the charge phase is determined by the individual battery voltage reaching a preset discharge cutoff voltage and / or a preset charge cutoff voltage.

7. The battery internal short circuit detection method according to claim 1, characterized in that, The weights related to the changes in charge within the operating cycle are determined by the cumulative charge and discharge amount obtained by integrating the current information of the battery pack in ampere-hours, and normalized by the absolute value of the cumulative charge and discharge amount so that the sum of the weights corresponding to each sampling point within the same operating cycle is 1.

8. The battery internal short circuit detection method according to claim 1, characterized in that, The construction of the first statistical discriminant includes: forming a first set based on the first cycle characteristic parameters of each individual battery cell within the same operating cycle; calculating the mean and standard deviation of the first set after removing the maximum and minimum values ​​in the first set; determining the standardized deviation of the minimum value in the first set relative to the mean as the first statistical discriminant; and determining the individual battery cell corresponding to the minimum value as the first candidate abnormal cell. The construction of the second statistical discriminant includes: using multiple consecutive operating cycles of a preset sliding window length as windows, calculating the standard deviation within the window as a fluctuation measure for the second cycle characteristic parameters of each individual battery; forming a second set based on the fluctuation measures of each individual battery within the same sliding window, calculating the mean and standard deviation of the second set after removing the maximum and minimum values ​​in the second set; determining the standardized deviation of the maximum fluctuation measure in the second set relative to the mean as the second statistical discriminant, and determining the individual battery corresponding to the maximum fluctuation measure as the second candidate abnormal individual; wherein, the preset sliding window length is 20 operating cycles.

9. The battery internal short circuit detection method according to claim 8, characterized in that: The first threshold and the second threshold are determined based on the mean and standard deviation of the first statistical discriminant sample and the second statistical discriminant sample in the historical dataset, respectively. The first threshold is the mean of the first statistical discriminant minus 3 times the standard deviation, and the second threshold is the mean of the second statistical discriminant plus 3 times the standard deviation. The joint discrimination includes: when the first statistical discrimination quantity is less than the first threshold, the second statistical discrimination quantity is greater than the second threshold, and the first candidate abnormal cell and the second candidate abnormal cell are the same cell, it is determined that a single cycle criterion is met; when the single cycle criterion is met in three consecutive operating cycles, it is determined that the cell has an internal short circuit.

10. A battery internal short circuit detection device, applied to a battery pack comprising multiple individual cells, characterized in that, include: The data acquisition module is used to acquire electrical signal data reflecting the operating status of the battery pack, including the current information of the battery pack and the voltage information of each individual cell. The parameter estimation module is used to construct observations characterizing the differences between individual cells based on the electrical signal data, and to perform online estimation of a first difference parameter and a second difference parameter based on the observations, thereby obtaining a first time series corresponding to the first difference parameter and a second time series corresponding to the second difference parameter; wherein, the first difference parameter is used to characterize the differences in voltage source characteristics of individual cells, and the second difference parameter is used to characterize the differences in impedance characteristics of individual cells; the online estimation is implemented using nonlinear filtering; The period processing module is used to segment the first time series and the second time series according to a preset operating period, and to perform weighted aggregation on the first time series and the second time series respectively according to the weight related to the change in electricity within the operating period in each operating period, so as to obtain the first period feature parameter corresponding to the first difference parameter and the second period feature parameter corresponding to the second difference parameter. The discrimination module constructs a first statistical discrimination quantity and a second statistical discrimination quantity for internal short circuit discrimination based on the first periodic feature parameter and the second periodic feature parameter. It then performs joint discrimination by combining the first statistical discrimination quantity with a first threshold and the second statistical discrimination quantity with a second threshold. When the abnormal cells corresponding to the first statistical discrimination quantity and the second statistical discrimination quantity are consistent and a preset continuous period criterion is met, it determines that the cell has an internal short circuit. The first statistical discrimination quantity is used to reflect the degree of deviation of the first difference parameter from the extreme value of the statistical benchmark within the battery pack, and the second statistical discrimination quantity is used to reflect the degree of fluctuation deviation of the second difference parameter from the statistical benchmark within the battery pack over multiple operating cycles.

Images