Battery fault diagnosis method and system, storage medium and electric vehicle

Abstract

The invention discloses a battery fault diagnosis method and system, a storage medium and an electric vehicle, and belongs to the technical field of lithium ion battery safety management. Synchronously acquiring voltage transient response data of each single battery in the battery pack under disturbance; dynamically identifying and determining a normal battery cluster representing consistency at the current moment, and establishing a dynamic reference voltage response; for a suspected battery, the voltage transient response of the suspected battery is compared with the dynamic reference voltage response, differential characteristics of multiple dimensions are extracted to carry out collaborative analysis and decision fusion so as to distinguish and judge the fault type, corresponding early warning or battery management actions are executed according to the fault type, and through injection current step disturbance, the battery management accuracy is improved. Linear and nonlinear dynamic characteristic distortion which is difficult to perceive and cannot be represented under steady-state operation of the battery can be excited, extremely high detection sensitivity is achieved for weak faults such as high-resistance internal short circuit and early aging, and preposition of a safety early warning window is achieved.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery safety management technology, and in particular to a battery fault diagnosis method, system, storage medium, and electric vehicle. Background Technology

[0002] With the rapid development of electric vehicles, large-scale energy storage, and other fields, the operational safety of lithium-ion batteries, as core energy storage units, has become a focus of industry attention. Battery failures, especially internal short-circuit faults that can lead to thermal runaway, aging failures caused by long-term operation, and electrical connection loosening faults caused by vibration and corrosion, pose significant threats to battery system safety. Therefore, there is an urgent need to develop efficient, accurate, and online battery fault diagnosis technologies.

[0003] Currently, common battery fault diagnosis methods have the following limitations: 1) Insufficient sensitivity to early and weak faults (such as high-resistance internal short circuits), making it difficult to achieve early safety warnings; 2) Difficulty in effectively distinguishing fault types, especially easily confusing aging batteries with early internal short circuit batteries, or misjudging connection faults as battery-related problems; 3) Diagnostic thresholds or model parameters are often fixed values, unable to adaptively compensate for parameter drift caused by initial inconsistencies, temperature fluctuations, and overall aging of the battery pack, resulting in high false alarm and false alarm rates.

[0004] Existing technical solutions, such as those based on frequency domain impedance analysis, charging curve characteristic peak analysis, or multivariate statistical decision-making, either rely on specific operating conditions such as the complete charging process, require high-precision models, or can only target a single fault type. These solutions are insufficient to meet the needs of real-time, robust, and accurate diagnosis of multiple types of faults in the complex operating environment of real vehicles. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides a battery fault diagnosis method, system, storage medium, and electric vehicle.

[0006] In a first aspect, a battery fault diagnosis method is provided, the method comprising:

[0007] S1: Inject current step disturbance signal during battery pack operation;

[0008] S2: Synchronously collect voltage transient response data of each individual cell in the battery pack under the current step disturbance signal.

[0009] S3: Based on the voltage transient response data, dynamically identify and establish a normal battery cluster that represents consistency at the current moment, and establish a dynamic reference voltage response based on the average response of the cluster.

[0010] S4: For suspected batteries in abnormal battery clusters, compare their voltage transient response with the dynamic reference voltage response, and extract multi-dimensional differential features from different stages of the response process.

[0011] S5: Based on the differentiated features of multiple dimensions, perform collaborative analysis and decision fusion to distinguish and determine the fault type, which includes at least one of internal short circuit, aging and electrical connection fault;

[0012] S6: Execute the corresponding warning or battery management action according to the fault type.

[0013] Furthermore, the current step disturbance signal includes at least a first step and a second step;

[0014] The amplitude and duration of the second step are both greater than those of the first step;

[0015] Furthermore, the amplitude of the current step disturbance signal is dynamically calculated based on the real-time voltage of the battery and an internal resistance parameter to ensure that the safety voltage constraint is met.

[0016] Furthermore, the safety voltage constraint includes:

[0017] Signal-to-noise ratio constraint: The instantaneous ohmic voltage drop generated by the first step is greater than a preset multiple of the effective value of the system voltage measurement noise;

[0018] Safety voltage constraint: During the second step, the peak voltage of the battery shall not exceed its maximum allowable charging voltage minus a safety margin.

[0019] Dynamic excitation constraint: The rise time and fall time of the current step signal are much smaller than the time constant of the fastest polarization process of the battery under test.

[0020] Furthermore, S1 also includes adaptive adjustment of the current step disturbance according to the battery's state of charge, specifically:

[0021] Within the normal diagnostic range of SOC=20%-80%, the current step disturbance is implemented normally according to the original amplitude;

[0022] When the SOC is within the safety limit zone of < 20% or > 80%, the current step amplitude is automatically reduced or the diagnostics are paused.

[0023] Furthermore, S3 specifically includes:

[0024] For each individual cell, extract its average voltage within a very short time window after the first step begins, and subtract it from the steady-state average voltage before the step to obtain the ohmic internal resistance voltage drop ΔV_i_ohm corresponding to the cell.

[0025] The numerical distribution of the ohmic internal resistance voltage drop ΔV_i_ohm of all individual cells is statistically analyzed, and its mean μ and standard deviation σ are calculated. Cells that meet the condition |ΔV_i_ohm -μ| < 2σ are classified as candidate normal cell clusters.

[0026] The voltage response curves V_i(t) of the normal battery cluster during the complete observation period are time-aligned, and then the average voltage is calculated at each synchronized sampling time point to generate a dynamic reference voltage response V_ref(t).

[0027] Calculate the average ohmic internal resistance of a normal battery cluster: R_ohm_ref = average(ΔV_i_ohm) / ΔI1, ΔI1=I_pulse - I_steady, where I_pulse is the disturbance current and I_steady is the steady-state charging current.

[0028] The average (ΔV_i_ohm) = V_step - V_steady, where V_steady is the average voltage of the dynamic reference voltage response V_ref(t) within a very short time window before the step, and V_step is the average voltage of the dynamic reference voltage response V_ref(t) within a very short time window after the step.

[0029] S4 specifically includes:

[0030] For each tagged suspect battery, its transient voltage response V_s(t) is compared with the dynamic reference voltage response V_ref(t), and multiple feature parameters are extracted from different stages of the response process:

[0031] First characteristic F1: F1 = |R_ohm_s - R_ohm_ref| / R_ohm_ref, where R_ohm_s = ΔV_s_ohm / ΔI1, and ΔV_s_ohm is the ohmic internal resistance voltage drop of the suspect battery ΔV_i_ohm.

[0032] Second feature F2: F2 = |k_s - k_ref| / k_ref, where k_s and k_ref are the average slopes of voltage rise of the transient voltage response V_s(t) and the dynamic reference voltage response V_ref(t) within a window after the start of the second step, respectively.

[0033] The third feature F3: F3 = |τ_s-τ_ref| / τ_ref, where τ_s and τ_ref are the recovery time constants of the suspect battery and the normal battery, respectively, obtained by fitting the voltage recovery curve segment that lasts for several seconds after the second step ends with a first-order exponential function.

[0034] Fourth feature F4: F4 = |r_s - r_ref| / r_ref, where r_s and r_ref are the ratios of the steady-state polarization voltage increments generated by the suspect battery and the normal battery under the first and second jumps, respectively.

[0035] Fifth feature F5: F5 = |E_s - E_ref| / E_ref, where E_s is the root mean square of the high-frequency signal corresponding to the suspect battery, and E_ref is the average value of the high-frequency energy value of each normal battery.

[0036] Furthermore, S5 specifically includes:

[0037] The extracted original feature values ​​F1 to F5 are divided by the preset significant anomaly threshold of each feature to convert them into dimensionless anomaly indexes f1 to f5.

[0038] Based on the different characteristics of the diagnostic sensitivity and specificity of various faults, differentiated weights are assigned to them: W = [W1,W2, W3, W4, W5] = [0.05, 0.3, 0.25, 0.3, 0.1].

[0039] Calculate the internal short-circuit confidence C_isc = (W2*f2 + W4*f4 +α*W1*f1) / (W2 + W4 +α*W1);

[0040] Calculate the aging confidence level C_aging = (W3*f3 +*W1*f1) / (W3 +β*W1);

[0041] Calculate the confidence level of electrical connection faults: C_conn = (W5*f5 +γ*W1*f1) / (W5 +γ*W1);

[0042] Where α, β, and γ are adjustment coefficients;

[0043] The calculated confidence levels for internal short circuits, aging, and electrical connection faults are compared with preset thresholds and cross-validated to make a final determination.

[0044] Furthermore, the calculated confidence levels for internal short circuits, aging, and electrical connection faults are compared with preset thresholds and cross-validated to make a final judgment. Specifically, this includes:

[0045] If the confidence level of an internal short circuit exceeds its preset threshold, an internal short circuit is initially suspected. Then, the internal short circuit resistance R_isc is estimated, or the final internal short circuit resistance value R_isc,f is calculated. If R_isc or R_isc,f is less than a critical value, an internal short circuit fault is finally confirmed. Here, R_isc ≈ R_ohm_ref / F1 - R_ohm_ref, or R_isc,f = (C_F1 * R_isc1 + C_F2 * R_isc2) / (C_F1 + C_F2), where C_F1 = (W1 * f1) / (W1 * f1 + W2 * f2), C_F2 = (W2 * f2) / (W1 * f1 + W2 * f2), R_isc1 = R_isc, R_isc2 = (F2 * R_pol,ref) / (1 - F2), R_pol,ref = V_pol,ref2 / ΔI2,V_pol,ref2 is the polarization steady-state voltage taken from V_ref(t) of the second step;

[0046] If the aging confidence exceeds its preset threshold, and the internal short circuit confidence is less than the lower limit of the internal short circuit confidence, while the internal short circuit resistance is greater than the upper limit of the internal short circuit resistance, then it is determined to be an aging fault.

[0047] If the confidence level of an electrical connection fault exceeds its preset threshold, and f1 also exceeds the threshold, then it is determined to be an electrical connection fault.

[0048] Secondly, a battery fault diagnosis system is provided for implementing the aforementioned battery internal short circuit diagnosis method, the system comprising:

[0049] An active excitation module is used to inject current step disturbance signals during battery pack operation;

[0050] The high-speed synchronous acquisition module is used to synchronously acquire the voltage transient response data of each individual cell in the battery pack under the disturbance of the current step disturbance signal;

[0051] The data processing and diagnostic module is used to dynamically identify and establish a normal battery cluster with consistent characteristics at the current moment based on the voltage transient response data, and to establish a dynamic reference voltage response based on the average response of the cluster.

[0052] It is also used to compare the voltage transient response of suspected batteries in abnormal battery clusters with the dynamic reference voltage response, and extract multi-dimensional differential features from different stages of the response process.

[0053] It is also used for collaborative analysis and decision fusion based on the differentiated features of multiple dimensions to distinguish and determine the fault type, which includes at least one of internal short circuit, aging and electrical connection fault;

[0054] The control and communication module is used to coordinate the operation of the system and output diagnostic results and control signals.

[0055] Thirdly, a computer-readable storage medium is provided, storing a battery fault diagnosis program, which, when executed by a processor, implements the battery internal short circuit diagnosis method.

[0056] Fourthly, an electric vehicle is provided, including the aforementioned battery fault diagnosis system.

[0057] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows:

[0058] Strong early warning capability: By actively injecting specific waveform (double step) current disturbance, it can excite linear and nonlinear dynamic characteristic distortions that are difficult to detect and characterize under steady-state operation of the battery. It has extremely high detection sensitivity for weak faults such as high resistance internal short circuit and early aging, realizing the early warning window of safety.

[0059] Accurate fault differentiation: The system systematically extracts multi-dimensional features from transient responses and constructs differentiated fusion judgment logic. A single diagnosis can simultaneously and accurately distinguish internal short circuits, aging, and connection faults, realizing a comprehensive assessment of the mechanical-electrical integrated health status of the battery system.

[0060] High adaptive robustness: The introduction of a dynamic benchmark establishment mechanism utilizes real-time data from each diagnosis to self-calibrate the reference standard, effectively overcoming the problem of fixed thresholds or model failures caused by battery pack parameter drift, and significantly improving the adaptability and reliability of the diagnostic method under actual complex working conditions.

[0061] High value in quantitative assessment: It can not only qualitatively determine the type of fault, but also quantitatively estimate key parameters such as internal short-circuit resistance, providing direct data support for assessing the severity of faults, tracking deterioration trends, and developing predictive maintenance strategies. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of a battery fault diagnosis method provided by the present invention;

[0064] Figure 2 This is a schematic diagram of a current step disturbance signal provided by the present invention;

[0065] Figure 3 This invention provides a structural frame for a battery fault diagnosis system.

[0066] Figure reference numerals: 100-Battery fault diagnosis system; 101-Active excitation module; 102-High-speed synchronous acquisition module; 103-Data processing and diagnosis module; 104-Control and communication module. Detailed Implementation

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0069] Example 1

[0070] See Figure 1 A battery fault diagnosis method includes the following steps: Step (1): Injecting a current step disturbance signal during the operation of the battery pack.

[0071] It should be noted that this step aims to safely and effectively stimulate the battery's dynamic response. In practice, an active excitation module consisting of a microcontroller and power MOSFETs / IGBTs is used to inject a set of [unclear - possibly a specific type of MOSFET] during the constant current charging phase of the battery pack. Figure 2 The double-step current disturbance signal is shown. The structure and preferred parameters of the current disturbance signal are as follows:

[0072] The first step is a small-amplitude, rapid step, with typical preferred parameters: amplitude ΔI1 is 0.05C to 0.1C of the battery's rated capacity, and duration t1 is between 1 and 10 ms. This step is used to excite the battery's ohmic response, mainly reflecting its ohmic internal resistance.

[0073] The second step is a large-amplitude, relatively long step, with typical preferred parameters: amplitude ΔI2 is 0.2C to 0.3C of the battery's rated capacity, and duration t2 is between 50-200ms. This step is used to excite the battery's polarization response, mainly reflecting polarization internal resistance and dynamic characteristics.

[0074] For example, taking a commercial vehicle power battery system with a total energy of 282kWh and a voltage platform of 640V as an example, its rated capacity is approximately 440Ah. According to the preferred proportion of the present invention, during the diagnostic process, a first step current is injected at time t0 during the stable constant current charging phase (400A), and its excitation process is defined as stage T1. The current amplitude is approximately 22A (0.05C), and after a duration of 5ms, the charging current returns to 400A. After waiting for a preset time (100ms in this embodiment), a second step current is injected at time t1, and its excitation process is defined as stage T2. The current amplitude is approximately 88A (0.2C), and after a duration of 50ms, the charging current returns to 400A. This excitation signal can generate a significant voltage jump of approximately 88mV on a typical cell internal resistance (e.g., 1mΩ), which is much higher than the measurement noise. At the same time, its amplitude is much smaller than the system's own fast charging current (e.g., 400A) and safety limits, thereby ensuring the safety of the entire process while maintaining high diagnostic sensitivity. The lower limit of the step amplitude is selected in this example to prioritize absolute safety during the initial demonstration. In practical applications, it can be adjusted to a better value within the scope defined in the claims, depending on the specific battery characteristics and diagnostic requirements.

[0075] It should also be noted that the generation of the current step disturbance signal is not a mechanical random selection of the above fixed parameter range, but rather a dynamic calculation or verification based on real-time operating conditions to meet the following core constraints and ensure the safety and effectiveness of the diagnosis.

[0076] Signal-to-noise ratio constraint: The system ensures that the instantaneous ohmic voltage drop ΔV_ohm ≈ ΔI1 * R_ohm_est generated by the first step is significantly higher than the system measurement noise V_noise (typically 1-2mV). Here, R_ohm_est refers to the estimated ohmic internal resistance used to predict the voltage jump and design an effective excitation signal before implementing this current step. Typically, the factory-set typical value for this battery model can be used directly. Alternatively, the average ohmic internal resistance of a healthy battery calculated in the most recent diagnostic can be used. In this embodiment, ΔV_ohm ≥ 10 * V_noise is required. If this requirement is not met, ΔI1 is automatically increased slightly.

[0077] Safety Voltage Constraint: To prevent the battery voltage from exceeding the safety limit during excitation, the system performs dynamic safety calculations on the amplitude of the second step. During the calculation, it must be ensured that the estimated peak voltage V_peak satisfies: V_peak ≤ V_charge_max - ΔV_margin. Where V_charge_max is the maximum allowable charging voltage of a single battery cell, and ΔV_margin is the preset safety margin voltage (e.g., 50 mV). The peak voltage can be estimated as the sum of the current battery voltage V_cell, the ohmic voltage drop ΔI² × R_ohm_est, and the estimated polarization voltage drop. Therefore, the upper limit of the dynamic safety amplitude for the second step is:

[0078] ΔI2_max_safe = (V_charge_max - ΔV_margin - V_cell) / (R_ohm_est +K_pol)

[0079] In the formula, R_ohm_est is a parameter related to the battery's ohmic internal resistance. During the first system run or the first excitation of this diagnostic cycle, this parameter can be the battery's nominal internal resistance at the factory. After completing the first step response analysis, a more accurate ohmic internal resistance value can be obtained and used to update this parameter for subsequent diagnostics. K_pol is an empirical coefficient used to compensate for polarization voltage drop. The final injected second step amplitude ΔI2 is the smaller of ΔI2_max_safe and the maximum current allowed by the system hardware ΔI_max_hardware.

[0080] Dynamic excitation constraint: The rise time and fall time of the current step signal (typically less than 100 microseconds) are much smaller than the time constant of the fastest polarization process of the battery under test (usually greater than 1 second), thus ensuring that the excitation signal can be regarded as an ideal step input to effectively excite the dynamic characteristics of the battery.

[0081] In addition, to ensure the safety of the diagnostic process and avoid overcharging or over-discharging of the battery due to the diagnostic activity itself, this solution will adaptively adjust the current step disturbance according to the battery's state of charge (SOC). The specific implementation strategy is as follows:

[0082] Normal diagnostic zone (SOC 20%-80%): In this voltage plateau region, the battery has a wide safety margin. The system implements current step disturbances at the original amplitude (e.g., 0.05C-0.3C) to obtain voltage response data with high signal-to-noise ratio and obvious characteristics.

[0083] Safety Limit Zone (SOC < 20% or SOC > 80%): When the battery's SOC enters this range, its terminal voltage is close to the upper or lower limit of the charge / discharge cutoff voltage, and the electrochemical system is on the edge of stability. Applying a large current jump at this point can easily lead to failure due to the instantaneous changes in ohmic resistance voltage drop and polarization voltage. For example, at high SOC, the voltage may surge instantaneously above the charging cutoff voltage, triggering the BMS overvoltage protection and potentially accelerating side reactions such as oxygen evolution in the positive electrode material, posing a risk of thermal runaway. At low SOC, the voltage may drop instantaneously below the discharge cutoff voltage, triggering the BMS undervoltage protection and potentially causing the copper current collector on the negative electrode to dissolve, resulting in irreversible damage.

[0084] Therefore, the system initiates safety strategies within this range, such as Strategy 1: Automatically reduce the current step amplitude: for example, reduce the amplitude to 0.02C to 0.05C of the rated capacity. The aim is to perform limited diagnosis using a sufficiently small disturbance signal that will not cause voltage overshoot, while ensuring absolute safety. At this time, the signal-to-noise ratio decreases, and the diagnostic sensitivity will decrease accordingly. Strategy 2: Suspend diagnosis: As the most conservative strategy, when the system assesses extremely high risk (such as extreme SOC, unsuitable temperature, etc.), it can choose to temporarily skip the current diagnostic cycle to prioritize battery safety. In this embodiment, the pause diagnosis strategy is preferred. This adaptive strategy achieves the optimal balance between fault diagnosis effectiveness and battery operational safety by monitoring SOC in real time and dynamically managing the intensity of diagnostic behavior.

[0085] Step (2): Synchronously collect voltage transient response data of each individual cell in the battery pack under the disturbance of current step disturbance signal.

[0086] Simultaneously and subsequently, the system uses a multi-channel synchronous high-speed analog-to-digital converter (ADC) at a sampling rate of no less than 10 kHz to synchronously acquire the transient response data V_i(t) of the terminal voltage of each individual cell in the battery pack (where i is the cell number), and simultaneously acquires the bus current I(t). The acquisition must cover the complete process of steady state before the step, transient during the step, and recovery after the step, and all data must be accompanied by precise timestamps.

[0087] Step (3): Based on the voltage transient response data, dynamically identify and establish a normal battery cluster that represents consistency at the current moment, and establish a dynamic reference voltage response based on the average response of the cluster.

[0088] This step aims to eliminate the effects of initial inconsistencies in the battery pack and overall aging, enabling adaptive diagnostic benchmarks.

[0089] First, calculate the ohmic voltage drop: For each individual cell, extract its average voltage within a very short time window (e.g., 1-10 milliseconds) after the first step, and subtract it from the steady-state average voltage before the step to obtain the ohmic internal resistance voltage drop ΔV_i_ohm for that cell. This ΔV_i_ohm value constitutes a primary characteristic quantity for judging the initial consistency of the cell.

[0090] Then, normal battery clusters are identified: Based on the consistency assumption, the numerical distribution of the ohmic internal resistance voltage drop ΔV_i_ohm of all batteries is statistically analyzed, and its mean μ and standard deviation σ are calculated. Batteries that meet the condition |ΔV_i_ohm -μ| < 2σ are classified as candidate normal batteries. If the number of batteries in this candidate cluster exceeds a preset proportion of the total number of batteries (e.g., 75%), it is officially confirmed as a normal battery cluster for this diagnosis. Batteries outside this range are marked as suspect batteries.

[0091] A dynamic baseline is then established: the voltage response curves V_i(t) of all members in the normal battery cluster over the complete observation period are time-aligned, and the average voltage is calculated at each synchronized sampling time point to generate a dynamic baseline voltage response curve V_ref(t). This curve represents the standard response of a healthy battery group to the same excitation under the current operating conditions. Simultaneously, the average ohmic internal resistance R_ohm_ref of the normal battery cluster is calculated as average(ΔV_i_ohm) / ΔI1.

[0092] In this embodiment, the steps for calculating the average ohmic internal resistance of the normal battery cluster are as follows:

[0093] The precise time point t0 at which the current step begins can be determined from the dynamic reference voltage response curve V_ref(t).

[0094] The average voltage of the curve within a very short time window before the step jump (e.g., 1 ms before t0) is taken as the reference steady-state voltage V_steady.

[0095] The average voltage of the curve within a very short time window after the step jump (e.g., 5ms after t0) is taken as the instantaneous voltage V_step after the step jump.

[0096] Calculate the average (ΔV_i_ohm) = V_step - V_steady.

[0097] R_ohm_ref = average(ΔV_i_ohm) / ΔI1, where ΔI1 is the change in current from the steady-state charging current I_steady to the disturbance current I_pulse, i.e., ΔI1 = I_pulse - I_steady.

[0098] Step (4): For suspected batteries in abnormal battery clusters, compare their transient voltage response with the dynamic reference voltage response, and extract multi-dimensional differential features from different stages of the response process, specifically including:

[0099] For each tagged suspect battery, its transient voltage response V_s(t) is compared in detail with the dynamic reference voltage response curve V_ref(t) established in the previous step, and multiple feature parameters are extracted from different stages of the response process:

[0100] The first characteristic (F1) reflects the relative change in ohmic internal resistance. The calculation formula is: F1 = |R_ohm_s - R_ohm_ref| / R_ohm_ref, where R_ohm_s is the ohmic internal resistance of the suspect battery s calculated based on the first step response (R_ohm_s = ΔV_s_ohm / ΔI1). ΔV_s_ohm is the ΔV_i_ohm value of the suspect battery s calculated in step (3). This characteristic is sensitive to any fault that causes a change in ohmic internal resistance (such as internal short circuits, aging, or connection corrosion).

[0101] The second characteristic (F2) reflects the relative change in polarization kinetics. By analyzing the average slopes k_s and k_ref of the voltage rise of V_s(t) and V_ref(t) within a time window (e.g., 10 to 20 milliseconds) after the start of the second step, the relative deviation is calculated as: F2 = |k_s - k_ref| / k_ref. An internal short circuit shunts the polarization current, slowing down the polarization voltage buildup process, and significantly increasing F2.

[0102] The third feature (F3) reflects the relative difference in the voltage recovery time constant. For the voltage recovery curve segment lasting several seconds after the second step, a first-order exponential function is used for fitting. In this embodiment, V(t) = A*exp(-t / τ) + B is used. The parameter V(t) represents the measured battery terminal voltage at time t, which is the dependent variable. The parameter t is the independent time variable, representing the time elapsed since the start of the recovery process. The parameter A represents the initial amplitude of the voltage recovery, which is the maximum deviation of the voltage from the steady-state value at the start of the recovery process. The parameter τ is the recovery time constant, which characterizes how quickly the voltage recovers to the steady-state value and is a key indicator for assessing the battery's aging state. The parameter B is the steady-state voltage value after the recovery process ends, representing the battery's equilibrium state under a given charging current. V(t) and t are the raw measured values ​​directly obtained from the high-speed data acquisition module; they are the inputs to the fitting algorithm. A, τ, and B are the parameters obtained through the fitting algorithm.

[0103] Using the above formulas, the recovery time constants τ_s and τ_ref for the suspect battery and the normal battery are obtained respectively. The result is calculated as: F3 = |τ_s - τ_ref| / τ_ref. Battery aging typically leads to a decrease in the diffusion ability of lithium ions in the active material, resulting in a slower voltage recovery process, an increase in τ, and consequently, an increase in F3.

[0104] The fourth characteristic (F4) reflects the relative deviation of the linearity of the step response. The steady-state polarization voltage increments (minus the ohmic voltage drop) generated by the suspect cell and the normal cell for the first and second step responses are calculated separately, and their ratios are obtained to obtain r_s and r_ref. F4 is defined as |r_s - r_ref| / r_ref.

[0105] Specifically, the steady-state polarization voltage values ​​of the suspect battery and the normal battery with respect to the first step (ΔI1) and the second step (ΔI2) are obtained, denoted as V_pol,s1, V_pol,s2 and V_pol,ref1, V_pol,ref2 respectively. Then, the response proportions of the suspect battery (r_s = V_pol,s2 / V_pol,s1) and the normal battery (r_ref = V_pol,ref2 / V_pol,ref1) are calculated. Finally, the linearity characteristic is quantified using the following formula: F4 = |r_s - r_ref| / r_ref. In the small to medium current range, the polarization response of a healthy battery is approximately linear, meaning the r value is relatively stable. Severe internal short circuits disrupt this linear relationship, leading to a significant change in F4.

[0106] The fifth characteristic (F5) reflects anomalies in electrical connection conditions. High-frequency components of the voltage signal within a very short time window (e.g., 1 millisecond) following a current step (especially the first step) are extracted (this can be obtained through high-pass digital filtering, with a cutoff frequency of, for example, 1 kHz). The root mean square energy E_s of the high-frequency signal corresponding to the suspect battery is calculated. The same processing is applied to the voltage data of each battery in the normal battery cluster to obtain the high-frequency energy value of each battery. The average value is taken as the reference energy E_ref. Then, F5 = |E_s - E_ref| / E_ref. Loosening or corrosion of electrical connection points (such as bolts and welds) can cause nonlinear contact resistance, resulting in significant high-frequency voltage oscillations during sudden current changes, thus increasing F5.

[0107] Step (5): Perform collaborative analysis and decision fusion based on differentiated features across multiple dimensions to distinguish and determine fault types, specifically including:

[0108] Feature normalization: The extracted original feature values ​​F1 to F5 are divided by the preset "significant anomaly threshold" of each feature to transform them into dimensionless anomaly indexes f1 to f5, so as to eliminate the influence of dimensions and unify the scale.

[0109] Differentiated weight allocation and confidence calculation: Based on the diagnostic sensitivity and specificity of different features for various faults, differentiated weights are assigned, and the sum of the weight coefficients must be 1. In this embodiment, the specific weight configuration is as follows: The weight vector W = [W1, W2, W3, W4, W5] = [0.05, 0.3, 0.25, 0.3, 0.1] is set, where W2 and W4 emphasize internal short-circuit indication, W3 emphasizes aging indication, W5 emphasizes connection fault indication, and W1 represents the common manifestations of various faults. Subsequently, the comprehensive confidence scores for the three types of faults are calculated respectively:

[0110] Internal short-circuit confidence level: C_isc = (W2*f2 + W4*f4 + α*W1*f1) / (W2 + W4 + α*W1)

[0111] Aging confidence level: C_aging = (W3*f3 + β*W1*f1) / (W3 + β*W1)

[0112] Connection failure confidence: C_conn = (W5*f5 + γ*W1*f1) / (W5 + γ*W1)

[0113] Where α, β, and γ are adjustment coefficients, which can be taken as 0.5 in this embodiment.

[0114] Fault determination and parameter estimation: The calculated overall confidence level is compared with a preset threshold, and cross-validation is performed to make a final determination.

[0115] If C_isc exceeds its threshold (e.g., 0.7), an internal short circuit is initially suspected. Further, the internal short circuit resistance can be estimated based on the first characteristic F1, the second characteristic F2, and the reference ohmic internal resistance R_ohm_ref. If the short circuit resistance is less than a critical value (e.g., 100 Ω), it is ultimately confirmed as an internal short circuit fault.

[0116] If C_aging exceeds its threshold (e.g., 0.6), and the internal short-circuit confidence C_isc is below a lower threshold (e.g., 0.4), while the estimated R_isc is large (e.g., > 1000 Ω), then it is determined to be an aging fault.

[0117] If C_conn exceeds its threshold (e.g., 0.5), and f1 also exceeds a lower threshold (e.g., 0.3), then an electrical connection fault is identified.

[0118] The above criteria can be met independently, supporting the reporting of multiple concurrent faults for the same battery, such as "internal short circuit with loose connection", thereby enabling the integrated diagnostic system to support the reporting of multiple concurrent faults for the same battery.

[0119] To clearly illustrate the above logic, an example is given below: Suppose that after the system completes calculations for a suspected battery, it obtains an estimated internal short-circuit resistance of 65Ω, an overall confidence level of 0.85 for the internal short circuit, an overall confidence level of 0.40 for aging, a connection state characteristic value of 0.60, and a relative rate of change characteristic value of ohmic internal resistance of 0.35. Based on the criteria:

[0120] Since both 65Ω < 100Ω and 0.85 > 0.7 are true, the fault is determined to be an internal short circuit.

[0121] 65Ω < 1000Ω, does not meet the aging judgment criteria.

[0122] Since both 0.60 > 0.5 and 0.35 > 0.3 are true, the connection is determined to be faulty.

[0123] Ultimately, the system determined that the battery simultaneously suffered from an internal short circuit and a connection fault. This example demonstrates the ability of the method of the present invention to effectively distinguish and comprehensively diagnose multiple coexisting fault types.

[0124] Short-circuit resistance can be estimated in various ways, the core of which lies in using the changes in battery parameters reflected by fault characteristics.

[0125] Method 1: Estimation is performed directly based on the first characteristic F1 and the reference ohmic internal resistance R_ohm_ref. According to the equivalent circuit principle, the internal short-circuit resistance R_isc is connected in parallel with the battery's ohmic internal resistance R_ohm, and its estimated value can be obtained through the following formula:

[0126] R_isc ≈ R_ohm_ref / F1- R_ohm_ref

[0127] Method 2: Two estimated values, R_isc1 and R_isc2, are derived from the first feature F1 and the second feature F2 using the model. Then, a weighted average is calculated based on their feature confidence levels to obtain the final estimated value, R_isc,f. The specific implementation is as follows:

[0128] Obtaining Dynamic Reference Parameters: First, based on the average response curve V_ref(t) of the normal battery cluster established in this diagnosis, the following key reference parameters are obtained: Reference polarization resistance R_pol,ref: From the second step V_ref(t), the steady-state polarization voltage V_pol,ref2 is taken and calculated according to the formula R_pol,ref = V_pol,ref2 / ΔI2. Reference ohmic internal resistance R_ohm_ref: This is a known quantity from the above steps.

[0129] Feature normalization: The aforementioned steps can be used to convert F1 and F2 into dimensionless anomaly indexes f1 and f2.

[0130] Calculate two independent resistance estimates separately: R_isc1 is directly estimated based on the first characteristic F1 and the reference ohmic internal resistance R_ohm_ref, and the result of this step is obtained from Method 1. The estimate R_isc2 based on the second characteristic F2 is calculated using the formula R_isc2 = (F2 * R_pol,ref) / (1 - F2), reflecting the dynamic current shunting.

[0131] Determine the dynamic fusion weights for this diagnosis: Calculate the confidence weights based on the significance of each feature in this measurement.

[0132] C_F1 = (W1 *f1) / (W1 * f1 + W2 * f2)

[0133] C_F2 = (W2 * f2) / (W1 * f1 + W2 *f2)

[0134] W1 and W2 are the preset base weights for the corresponding features. This design enables the system to adaptively adjust the confidence level of each predicted value based on the quality of the current data.

[0135] The final result of the weighted fusion calculation is: R_isc,f = (C_F1 * R_isc1 + C_F2 * R_isc2) / (C_F1 + C_F2).

[0136] In this invention, Method 1 is computationally efficient and has clear physical meaning, while Method 2 employs dynamic weighting, which improves the reliability and robustness of the results. Those skilled in the art can choose either of the above methods based on their actual needs for diagnostic accuracy and computational complexity. The estimated internal short-circuit resistance and various comprehensive confidence levels are obtained through the above methods.

[0137] Step (6): Execute the corresponding warning or battery management action according to the fault type.

[0138] The data processing and diagnostic module, through the control and communication module (such as a CAN bus controller), reports the diagnostic results (including the faulty battery number, fault type, confidence level, estimated internal short-circuit resistance value, etc.) to the upper-level battery management system (BMS) main controller or cloud monitoring platform in real time. The BMS can then execute tiered pre-control strategies based on the fault type and severity, such as triggering audible and visual alarms on the instrument panel, recording fault logs, limiting the power output of the module containing the faulty battery, and providing specific fault location and type prompts on the maintenance interface.

[0139] It is worth noting that the battery fault diagnosis method in this embodiment has the following significant advantages:

[0140] Strong early warning capability: By actively injecting specific waveform (double step) current disturbance, it can excite linear and nonlinear dynamic characteristic distortions that are difficult to detect and characterize under steady-state operation of the battery. It has extremely high detection sensitivity for weak faults such as high resistance internal short circuit and early aging, realizing the early warning window of safety.

[0141] Accurate fault differentiation: The system systematically extracts multi-dimensional features from transient responses and constructs differentiated fusion judgment logic. A single diagnosis can simultaneously and accurately distinguish internal short circuits, aging, and connection faults, realizing a comprehensive assessment of the mechanical-electrical integrated health status of the battery system.

[0142] High adaptive robustness: The introduction of a dynamic benchmark establishment mechanism utilizes real-time data from each diagnosis to self-calibrate the reference standard, effectively overcoming the problem of fixed thresholds or model failures caused by battery pack parameter drift, and significantly improving the adaptability and reliability of the diagnostic method under actual complex working conditions.

[0143] High value in quantitative assessment: It can not only qualitatively determine the type of fault, but also quantitatively estimate key parameters such as internal short-circuit resistance, providing direct data support for assessing the severity of faults, tracking deterioration trends, and developing predictive maintenance strategies.

[0144] Example 2

[0145] A battery fault diagnosis system is provided to implement the battery fault diagnosis method in Embodiment 1. Its structural block diagram is shown below. Figure 3 As shown. This battery fault diagnosis system 100 can be integrated into the vehicle's BMS or used as a standalone testing device. The system includes:

[0146] Active excitation module 101: Composed of a high-speed digital controller (such as a DSP or high-performance MCU) and a precision power amplifier circuit (such as an H-bridge or linear amplifier). It is responsible for receiving control commands and generating and injecting safe and controllable current step disturbance signals during battery pack operation.

[0147] The high-speed synchronous acquisition module 102 includes multiple high-precision, high common-mode rejection ratio differential operational amplifiers and a multi-channel synchronous sampling ADC chip. It is responsible for synchronously measuring the terminal voltage and bus current of all individual cells under current step disturbance signal disturbance at a high sampling rate, and sending the digitized voltage transient response data to the processing unit.

[0148] Data processing and diagnosis module 103: This is the core computing unit of the system, typically implemented by a high-performance microprocessor (such as the ARM Cortex-M7 / M33 series). Internally, it contains the software algorithm for implementing the battery fault diagnosis method in Example 1. It is responsible for dynamically identifying and establishing a normal battery cluster with consistent characteristics at the current moment based on voltage transient response data, and establishing a dynamic reference voltage response using the average response of this cluster. For suspected batteries in abnormal battery clusters, it compares their voltage transient response with the dynamic reference voltage response, extracting multi-dimensional differential features from different stages of the response process. Based on these multi-dimensional differential features, it performs collaborative analysis and decision fusion to distinguish and determine the fault type, which includes at least one of internal short circuit, aging, and electrical connection faults.

[0149] Control and Communication Module 104: Responsible for overall system scheduling and timing synchronization between modules. It is equipped with an on-board network interface (such as CAN FD, Ethernet) to receive diagnostic commands from the BMS master controller, report detailed diagnostic results, and trigger corresponding alarm signals, and perform corresponding warning or battery management actions according to the fault type.

[0150] Example 3

[0151] A computer-readable storage medium stores a battery fault diagnosis program, which, when executed by a processor, implements the battery internal short circuit diagnosis method in Embodiment 1.

[0152] Example 4

[0153] An electric vehicle includes the battery fault diagnosis system of Embodiment 2. The battery fault diagnosis system of Embodiment 2 is installed in the electric vehicle and identifies and detects whether there is a fault in the battery of the electric vehicle through the active excitation module, high-speed synchronous acquisition module, data processing and diagnosis module, and control and communication module in the battery fault diagnosis system.

[0154] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0155] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0156] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A battery failure diagnosis method characterized by comprising: The method comprises: S1: injecting a current step disturbance signal during the operation of the battery pack; S2: synchronously collecting voltage transient response data of each single battery in the battery pack under the disturbance of the current step disturbance signal; S3: dynamically identifying and establishing a normal battery cluster representing consistency at the current time based on the voltage transient response data, and establishing a dynamic reference voltage response based on the average response of the cluster; S4: for the suspected battery of the abnormal battery cluster, comparing its voltage transient response with the dynamic reference voltage response, and extracting multiple dimensional differential features from different stages of the response process; S5: based on the differential features in multiple dimensions, performing collaborative analysis and decision fusion to distinguish and determine the fault type, which includes at least one of internal short circuit, aging and electrical connection fault; S6: performing corresponding early warning or battery management action according to the fault type.

2. The battery failure diagnosis method according to claim 1, characterized by, The current step disturbance signal at least includes a first step and a second step; The amplitude and duration of the second step are greater than those of the first step; And the amplitude of the current step disturbance signal is dynamically calculated according to the real-time voltage of the battery and an internal resistance parameter to ensure that the safety voltage constraint is met.

3. The battery failure diagnosis method according to claim 2, characterized by, The safety voltage constraint includes: Signal-to-noise ratio constraint: the instantaneous ohmic voltage drop generated by the first step is greater than a preset multiple of the effective value of the system voltage measurement noise; Safety voltage constraint: the peak voltage of the battery during the second step does not exceed the maximum allowed charging voltage minus a safety margin; Dynamic excitation constraint: the rise time and fall time of the current step signal are much smaller than the time constant of the fastest polarization process of the battery to be tested.

4. The battery failure diagnosis method according to claim 3, characterized by, In S1, the current step disturbance is adaptively adjusted according to the state of charge of the battery, specifically: When in the normal diagnosis zone of SOC = 20%-80%, the current step disturbance is normally implemented with the original amplitude; When in the safety limit zone of SOC < 20% or SOC > 80%, the current step amplitude is automatically reduced or the diagnosis is suspended.

5. The battery failure diagnosis method according to claim 2 or 3, characterized by, S3 specifically includes: For each single battery, the average voltage of the battery in a very short time window after the start of the first step is extracted, and the difference between the average voltage before the step and the ohmic internal resistance voltage drop ΔV_i_ohm of the battery is obtained; The numerical distribution of the ohmic internal resistance voltage drop ΔV_i_ohm of all single batteries is calculated, and the mean μ and standard deviation σ are calculated, and the batteries satisfying the condition |ΔV_i_ohm - μ| < 2σ are classified as candidate normal battery clusters; The voltage response curves V_i(t) of the normal battery cluster in the complete observation period are time-aligned, and then the voltage average value at each synchronized sampling time point is calculated to generate the dynamic reference voltage response V_ref(t); The average ohmic internal resistance of the normal battery cluster is calculated: R_ohm_ref = average(ΔV_i_ohm) / ΔI1, ΔI1 = I_pulse - I_steady, I_pulse is the disturbance current, and I_steady is the charging steady-state current; The average (ΔV_i_ohm) = V_step - V_steady, V_steady is the average voltage of the dynamic reference voltage response V_ref(t) in a very short time window before the step, and V_step is the average voltage of the dynamic reference voltage response V_ref(t) in a very short time window after the step.

6. The battery failure diagnosis method according to claim 5, characterized by, The S4 specifically includes: For each labeled suspect battery, compare its voltage transient response V_s(t) with the dynamic reference voltage response V_ref(t), and extract a plurality of characteristic parameters from different stages of the response process: The first characteristic F1: F1 = |R_ohm_s - R_ohm_ref| / R_ohm_ref, wherein R_ohm_s = ΔV_s_ohm / ΔI1, and ΔV_s_ohm is the ohmic internal resistance voltage drop ΔV_i_ohm of the suspect battery; The second characteristic F2: F2 = |k_s - k_ref| / k_ref, wherein k_s and k_ref are the average voltage rise slopes of the voltage transient response V_s(t) and the dynamic reference voltage response V_ref(t) in a time window after the second step starts, respectively; The third characteristic F3: F3 = |τ_s - τ_ref| / τ_ref, wherein τ_s and τ_ref are the recovery time constants of the suspect battery and the normal battery obtained by fitting the voltage recovery curve segment of the suspect battery and the normal battery with a first-order exponential function for several seconds after the second step ends; The fourth characteristic F4: F4 = |r_s - r_ref| / r_ref, wherein r_s and r_ref are the ratios of the steady-state polarization voltage increments of the suspect battery and the normal battery under the first step and the second step, respectively; The fifth characteristic F5: F5 = |E_s - E_ref| / E_ref, wherein E_s is the root mean square of the high-frequency signal of the suspect battery, and E_ref is the average value of the high-frequency energy values of the normal batteries.

7. The battery failure diagnosis method according to claim 6, characterized by, The S5 specifically includes: Divide the extracted original characteristic values F1 to F5 by their respective preset significant abnormality thresholds to convert them into dimensionless abnormality degree indicators f1 to f5; According to the diagnostic sensitivity and specificity of different characteristics to various types of faults, assign them differentiated weights W = [W1, W2, W3, W4, W5] = [0.05, 0.3, 0.25, 0.3, 0.1]; Calculate the aging confidence C_aging = (W3*f3 +*W1*f1) / (W3 +β*W1); Calculate the electrical connection fault confidence C_conn = (W5*f5 +γ*W1*f1) / (W5 +γ*W1); Wherein α, β, γ are adjustment coefficients; The calculated internal short circuit confidence, aging confidence, and electrical connection fault confidence are compared with preset thresholds and cross-verified to make a final determination.

8. The battery failure diagnosis method according to claim 7, characterized by, The comparison of the calculated internal short circuit confidence, aging confidence, and electrical connection fault confidence with preset thresholds and cross-verification to make a final determination specifically includes: if the internal short circuit confidence exceeds its preset threshold, a preliminary determination is made that there is a suspected internal short circuit, and then the internal short circuit resistance R_isc or the final internal short circuit resistance value R_isc,f is estimated, if R_isc or R_isc,f is less than a critical value, the internal short circuit fault is finally confirmed, wherein R_isc≈R_ohm_ref / F1- R_ohm_ref, or R_isc,f = (C_F1* R_isc1 + C_F2 * R_isc2) / (C_F1 + C_F2), wherein C_F1 = (W1*f1) / (W1*f1 + W2*f2), C_F2 = (W2*f2) / (W1*f1 + W2*f2), R_isc1=R_isc, R_isc2 = (F2 *R_pol,ref) / (1-F2), R_pol,ref = V_pol,ref2 / ΔI2, V_pol,ref2 is the polarization steady-state voltage taken from V_ref(t) of the second step; If the aging confidence exceeds its preset threshold, and the internal short circuit confidence is less than the lower limit of the internal short circuit confidence, and the internal short circuit resistance is greater than the upper limit of the internal short circuit resistance, it is determined to be an aging fault. If the electrical connection fault confidence exceeds its preset threshold, and f1 also exceeds the threshold, it is determined to be an electrical connection fault.

9. A battery failure diagnosis system characterized by comprising: The system for implementing the battery internal short circuit diagnosis method as claimed in any one of claims 1-8 comprises: An active excitation module for injecting a current step disturbance signal during the operation of the battery pack; A high-speed synchronous acquisition module for synchronously acquiring voltage transient response data of each single battery in the battery pack under the disturbance of the current step disturbance signal; A data processing and diagnosis module for dynamically identifying and establishing a normal battery cluster representing consistency at the current time based on the voltage transient response data, and establishing a dynamic reference voltage response based on the average response of the cluster; Also for the suspected battery of the non-normal battery cluster, the voltage transient response of the battery is compared with the dynamic reference voltage response, and multiple dimensional differential features are extracted from different stages of the response process; Also for collaborative analysis and decision fusion based on multiple dimensional differential features to distinguish and determine the fault type, the fault type includes at least one of internal short circuit, aging, and electrical connection fault; A control and communication module for coordinating system work and outputting diagnosis results and control signals.

10. A computer-readable storage medium, characterized in that, A battery fault diagnosis program is stored, and when the battery fault diagnosis program is executed by a processor, the battery internal short circuit diagnosis method as claimed in any one of claims 1-8 is implemented.

11. An electric vehicle, characterized by The battery failure diagnosis system according to claim 9 is included.

Images