Energy storage battery cluster safety state dynamic evaluation method and system based on feature decoupling

Abstract

The application relates to the technical field of energy storage batteries, and discloses a method and system for dynamically evaluating the safety state of an energy storage battery cluster based on feature decoupling, which comprises the following steps: S1, obtaining time sequence operation data of a plurality of single batteries in a battery system, wherein the time sequence operation data at least comprises the voltage, temperature of each single battery and the loop current of the battery system; S2, identifying the operation condition of the battery system according to the loop current, wherein the operation condition comprises a static condition and a non-static condition; S3, when the static condition is identified, performing an internal short-circuit micro-analysis step based on common-mode rejection; S4, when the non-static condition is identified, performing a thermal runaway early warning step based on the fusion of electro-thermal mechanisms; and S5, generating and outputting corresponding early warning signals according to the judgment results of the internal short-circuit micro-analysis step and / or the thermal runaway early warning step. The application can adapt to battery aging and accurately extract fault features from complex operation data, and has high sensitivity and extremely low false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of energy storage battery technology, and specifically to a method and system for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling. Background Technology

[0002] In existing safety management of energy storage power stations and new energy vehicles, battery thermal runaway remains the primary threat to system safety. The accumulation of inconsistencies in individual battery cells, the gradual development of micro-internal short circuits, and the eventual outbreak of abnormal heat generation constitute a typical evolutionary chain from hidden danger to disaster. However, most existing early warning methods rely too heavily on static threshold monitoring of single physical quantities (such as voltage and temperature), for example, setting an alarm to be triggered when the voltage difference exceeds 100mV or the temperature exceeds 55℃. Such methods ignore the inherent aging characteristics of batteries throughout their entire life cycle: as the number of cycles increases, the battery's internal resistance and capacity dispersion naturally increase. Fixed thresholds are prone to causing a large number of invalid false alarms in the middle and later stages of battery life, leading to desensitization to alarm signals among maintenance personnel and masking true early faults.

[0003] Currently, some progress has been made in existing research on battery fault detection. For example, Chinese Patent Document 1 (CN116008820B) proposes a method for detecting inconsistencies in vehicle battery cells. It distinguishes between charging and discharging conditions, calculates the average voltage difference and voltage difference respectively to determine whether there are inconsistencies in the battery, and introduces trend analysis based on historical operating conditions. Chinese Patent Document 2 (CN117907844A) discloses a method for detecting short-circuit anomalies in a battery system. It calculates the characteristic vector of each individual cell, the SOC deviation characteristic factor, and the equivalent short-circuit resistance value, and provides a comprehensive early warning based on the ranking of the outlier degree of these three factors. Chinese Patent Document 3 (CN120629941A) proposes a fault detection method for lithium-ion batteries. It determines the operating state based on the total current of the battery pack. In the static state, it performs self-discharge detection through the voltage drop rate and temperature rise rate. In the charging and discharging state, it uses deep learning to extract temperature-voltage coupling features for consistency analysis. However, the aforementioned existing technologies still have significant limitations: While Patent Document 1 distinguishes between charge and discharge conditions, its voltage change under static conditions is still based on a direct slope, failing to isolate common-mode interference from polarization rebound; Patent Document 2 relies on the equivalent short-circuit resistance value, which requires complex parameter identification, resulting in high engineering difficulty and computational burden; While the deep learning model used in Patent Document 3 under charge and discharge conditions can extract features, its purely data-driven black-box nature makes it lack physical interpretability, and it has extremely high requirements for the quality and quantity of training data, limiting its generalization ability. More importantly, most of the aforementioned existing technologies still focus on the direct processing of apparent data, failing to delve into the coupling nature of the electrochemical and thermophysical processes inside the battery. In particular, in distinguishing between normal physical effects and early fault characteristics, there is a lack of effective mechanism-driven signal processing methods, resulting in detection accuracy and timeliness that cannot meet practical engineering needs.

[0004] In summary, the core technical challenges in achieving early warning of battery thermal runaway lie in three aspects: First, the nonlinear interference of polarization effect is difficult to eliminate: In the initial stage of rest after the battery is discharged, the voltage change is mainly determined by polarization rebound (nonlinear rising process) and micro-internal short circuit leakage (linear falling process). Existing methods directly calculate the falling slope of the original voltage, which cannot effectively separate these two factors, causing the early micro-internal short circuit characteristics to be completely submerged in the strong polarization background.

[0005] Second, the thermo-electric coupling mechanism is complex: the temperature rise of the battery during operation comes from both the Joule heating of the current (normal physical process) and the exothermic reaction of internal chemical side reactions (fault signal). Traditional technology only monitors the absolute temperature or the rate of temperature rise. Under high current charging and discharging conditions, normal Joule heating temperature rise is easily misjudged as a precursor to thermal runaway, resulting in a high false alarm rate.

[0006] Third, it lacks the ability to adapt throughout the entire life cycle: from the time a battery leaves the factory to its retirement, the distribution characteristics of its voltage, temperature and other parameters will drift significantly. Fixed thresholds or models based on the initial state cannot follow this change, thus losing their ability to make judgments after the battery ages.

[0007] The aforementioned challenges collectively limit the accurate identification of early-stage safety hazards in batteries using existing technologies. Summary of the Invention

[0008] The present invention aims to provide a dynamic assessment method and system for the safety status of energy storage battery clusters based on feature decoupling. It can adapt to battery aging and accurately extract fault features from complex operating data, and has high sensitivity and extremely low false alarm rate.

[0009] To achieve the above objectives, the present invention provides the following basic solution.

[0010] Option 1 The dynamic assessment method for the safety status of energy storage battery clusters based on feature decoupling includes the following steps: S1, acquire timing data of multiple individual cells in the battery system, wherein the timing data includes at least the voltage and temperature of each individual cell and the loop current of the battery system. S2, Identify the operating conditions of the battery system based on the circuit current, including the static operating conditions and the non-static operating conditions; S3, when identified as a static operating condition, perform the internal short-circuit microanalysis step based on common-mode suppression: For each individual cell, construct its relative voltage sequence with respect to the average voltage of all individual cells in the battery system. Preset time period under static working conditions Within this process, calculate the differential leakage rate for each individual cell: ; Based on the comparison results of the differential leakage rate of each individual cell and the group statistical distribution of the differential leakage rate in the battery system, it is determined whether there are abnormal individual cells caused by micro-internal short circuits. S4, when a non-static operating condition is identified, execute the thermal runaway early warning step based on the fusion of electrothermal mechanisms: Within a preset time window under non-static conditions, the thermal-electric coupling characteristic value of each individual cell is calculated. The thermal-electric coupling characteristic value represents the abnormal temperature rise residual after deducting Joule heating, reversible entropy heating, and heat dissipation terms, and its calculation formula is as follows: ; in, ; ; Let the loop current be the current at the k-th sampling point. Let be the equivalent internal resistance of the i-th single cell at the k-th sampling point. The open-circuit voltage temperature coefficient, The equivalent heat dissipation coefficient, Let be the heat capacity parameter of the i-th individual cell. This refers to the ambient temperature or the reference ambient temperature within the cluster. The predicted temperature for the k-th sampling point. Let M be the measured temperature at the k-th sampling point, and M be the number of sampling points participating in the residual calculation within the preset time window. The sampling time interval; Based on the comparison of the thermal-electric coupling characteristic values ​​of each individual cell with the group statistical distribution of the thermal-electric coupling characteristic values ​​within the battery system, it is determined whether there is a potential risk of thermal runaway due to abnormal chemical heat sources. S5. Based on the judgment results of the internal short-circuit microanalysis step and / or thermal runaway early warning step, generate and output the corresponding early warning signal.

[0011] Furthermore, when the absolute value of the circuit current is less than the preset static current threshold and the duration exceeds the first preset time threshold, it is determined to be a static working condition; otherwise, it is determined to be a non-static working condition. The non-static working condition includes charging working condition and discharging working condition. S4 also includes: performing a conformance assessment step during charging. For each individual battery cell, calculate its performance over the entire charging cycle. Intra-consistency anomaly index ; ; in, Let be the voltage of the i-th individual cell at time t. Let be the average voltage of all individual cells in the battery system at time t; The moment charging begins. This is the moment when charging ends; Real-time calculation of the average CAI value of all individual cells in the battery system and standard deviation ; When the i-th single cell satisfy At that time, it was determined that the single cell had a potential risk of consistent degradation.

[0012] Furthermore, in S3, the step of determining whether there is an abnormal individual cell due to a micro-internal short circuit, based on the comparison result of the differential leakage rate of each individual cell with the group statistical distribution of the differential leakage rate within the battery system, specifically includes: Real-time calculation of the average differential leakage rate of all individual cells in the battery system. and standard deviation ; When the differential leakage rate of the i-th single cell satisfy When the cell is in a state of abnormality, it is determined that there is a micro-internal short circuit. Here, k is a preset sensitivity coefficient with a value range of 3 to 6. The mean and standard deviation A sliding time window is used for dynamic updates to adapt to population characteristic drift caused by battery aging.

[0013] Furthermore, in S4, the determination of whether there is a potential risk of thermal runaway due to abnormal chemical heat sources, based on the comparison results of the thermal-electric coupling characteristic values ​​of each individual cell with the group statistical distribution of the thermal-electric coupling characteristic values ​​within the battery system, specifically includes: Real-time calculation of the mean thermal-electric coupling characteristics of all individual cells in the battery system and standard deviation ; When the thermal-electric coupling characteristic value of the i-th single cell satisfy At that time, it was determined that the single cell showed signs of impending thermal runaway. The mean and standard deviation It uses a sliding time window for dynamic updates.

[0014] Furthermore, the warning signal includes three levels of warning: When only the consistency anomaly index is present When an anomaly is detected, a first-level warning signal is output, indicating that equalization maintenance is required; When the differential leakage rate When an anomaly is detected, a second-level warning signal is output, indicating that the machine needs to be stopped for inspection. When the thermal-electric coupling characteristic value An anomaly is detected, or the differential leakage rate is determined. With the aforementioned thermo-electric coupling characteristic value Simultaneously, when an anomaly is detected, a third-level warning signal is output, indicating that the circuit needs to be cut off urgently and the fire emergency plan needs to be activated.

[0015] Furthermore, in S3, the preset time period The start time After being set to the second preset time threshold after the start of the static operation, the polarization voltage drastic change zone in the initial static phase is avoided; the second preset time threshold is 20 to 40 minutes.

[0016] Furthermore, prior to S2, a data cleaning step is included: removing time frame data from the time-series running data where the voltage of a single battery cell exceeds a preset normal voltage range, or the temperature exceeds a preset normal temperature range, or the data is missing.

[0017] Furthermore, in S4, the preset time window under non-static operating conditions is a continuous charging period or a continuous discharging period, and the thermo-electric coupling characteristic value The calculation is performed only when the absolute value of the loop current is continuously greater than the preset current threshold within the preset time window, so as to avoid calculation noise interference under low current conditions.

[0018] Furthermore, it also includes: recording the differential leakage rate of each individual cell under multiple consecutive static conditions. The leakage rate trend curve of the single cell is constructed; when the slope of the leakage rate trend curve is positive and exceeds the preset trend threshold, a micro-internal short circuit degradation trend warning signal is generated.

[0019] Option 2 A dynamic safety status assessment system for energy storage battery clusters based on feature decoupling is used to execute the dynamic safety status assessment method for energy storage battery clusters based on feature decoupling as described in Scheme 1; including: The data acquisition module is used to acquire the time-series operating data of multiple individual cells in the battery system in real time at a preset sampling frequency; The operating condition identification module is used to identify the operating condition of the battery system based on the circuit current. The internal short-circuit microanalysis module is used to perform internal short-circuit microanalysis based on common-mode suppression when the operating condition identification module determines that the operating condition is static. Thermal runaway early warning module is used to execute thermal runaway early warning based on electrothermal mechanism fusion when the operating condition identification module determines that the operating condition is not static. The graded early warning output module is used to generate and output early warning signals of corresponding levels based on the judgment results of the internal short circuit analysis module and / or the thermal runaway early warning module.

[0020] The working principle and advantages of this invention are as follows: This invention presents a dynamic safety status assessment method and system for energy storage battery clusters based on feature decoupling. It can adapt to battery aging and accurately extract fault features from complex operational data, exhibiting high sensitivity and an extremely low false alarm rate. The key points are: This solution, based on a profound understanding of the electrochemical polarization mechanism and thermophysical laws of batteries, constructs an original "3+1" dimension analysis framework. Using mathematical methods, it precisely decouples three physically distinct fault characteristics—"consistency," "internal short circuit," and "thermal anomaly"—from the complex raw BMS data. Unlike conventional techniques that conflate these three or simply superimpose them, this solution designs a dedicated mechanism-driven feature extraction channel for each fault mode: during the charging phase, consistency indicators are extracted by accumulating voltage trajectory deviations; during the resting phase, micro-internal short circuit leakage rate is extracted by suppressing common-mode interference through relative voltage differential; and during the operating phase, abnormal heat source characteristics are extracted by eliminating Joule heating effects through a thermo-electric decoupling model. These three fault feature dimensions work in parallel, each targeting a different physical failure mode of the battery, without interfering with and complementing each other. More importantly, this solution constructs a dynamic statistical decision-making dimension that spans the entire battery lifecycle. This is a 3-Sigma adaptive decision-making mechanism based on a sliding time window. The system calculates the characteristic distribution benchmark of the current battery cluster in real time at the current life stage, thereby forming a dynamic safety boundary that adaptively adjusts with the battery aging state. This completely eliminates the rigid thresholds of traditional "pressure difference > 100mV, temperature > 55℃" and solves the fundamental defect of frequent false alarms after battery aging caused by fixed thresholds.

[0021] In particular, unlike conventional methods, this approach does not passively accept the coupling states of various physical quantities in the raw data for the three fault characteristic dimensions. Instead, it actively utilizes the strong correlation of polarization responses of series-connected battery packs under the same operating conditions to subtract the average voltage within the cluster from the individual cell voltage. This defines the nonlinear polarization rebound as group common-mode interference and suppresses it, allowing the linear micro-internal short-circuit leakage characteristics to be orthogonally separated from the strong background noise. This process completely overturns the long-standing technical bias of those skilled in the art of "waiting for polarization to be eliminated before performing self-discharge detection," physically removing polarization interference. This enables highly sensitive capture of weak leakage signals at the millivolt-per-hour level even in the initial static stage, achieving a qualitative improvement in the signal-to-noise ratio of fault characteristics.

[0022] Under non-static conditions, traditional technologies only monitor absolute temperature or the rate of temperature rise. Normal Joule heating during high-current charging and discharging can easily be misinterpreted as a precursor to thermal runaway. This is an inevitable consequence of simply treating the battery as a "black box" and ignoring its internal energy conversion processes. This solution, based on the principle of energy conservation, first uses Joule heating, reversible entropy heating, heat dissipation terms, and heat capacity parameters to recursively predict temperature. Then, it constructs a thermo-electric coupling characteristic value using the residual between the measured temperature and the predicted temperature, thereby decoupling the normal electrothermal process from the fault temperature rise component caused by abnormal chemical reactions. This characteristic value has a clear physical meaning, its calculation does not require a large amount of training data, and it maintains interpretability under different rates and operating conditions, completely solving the persistent problem of frequent false alarms due to aging drift in the later stages of the battery's lifespan caused by fixed thresholds. In contrast, conventional multi-parameter fusion or machine learning methods can only learn statistical correlations and cannot establish causal relationships, making them prone to failure when operating conditions change.

[0023] From a technical perspective, this solution achieves three breakthroughs that are difficult to match with existing technologies. First, extremely high sensitivity: Through a differential algorithm, common-mode interference from polarization rebound and sensor zero-point drift is successfully eliminated, resulting in internal short-circuit detection accuracy at the mV / h level. It can capture minute leakage signals even in the early stages of quiescent operation, whereas traditional methods are completely ineffective against such early faults due to their inability to isolate polarization interference. Second, full lifecycle adaptability: By abandoning fixed thresholds such as "100mV" and adopting dynamic relative thresholds based on population statistics, the algorithm is applicable from the early to the late stages of battery life, eliminating the need for frequent manual parameter calibration and significantly reducing maintenance costs. Third, extremely low false alarm rate: By introducing the residual between the predicted and measured temperatures under thermal equilibrium conditions, it effectively distinguishes between "normal temperature rise under high-rate operating conditions" and "abnormal temperature rise due to faults," solving the pain point of false temperature alarms caused by drastic load fluctuations in energy storage frequency regulation and peak shaving scenarios, making the early warning system truly practical.

[0024] From a non-obviousness perspective, when faced with the challenge of battery thermal runaway early warning, a common approach for those skilled in the art is to add more parameters or employ a more complex classifier to the existing single-threshold method, such as feeding features like voltage, temperature, and internal resistance into a neural network for training. While this "data-driven" approach superficially integrates multi-source information, it does not fundamentally solve the problem of coupling interference between various physical quantities and relies on a large amount of labeled data, resulting in poor generalization ability.

[0025] This approach takes a contrarian approach, returning to the fundamental physical mechanisms. It creatively transforms classical electrochemical polarization theory and thermodynamic principles into engineering-feature-based signal processing methods, orthogonally decomposing coupled physical quantities at the feature level. For example, those skilled in the art generally consider the data from the initial voltage drop curve during the resting phase unusable, believing that self-discharge assessment can only be performed after polarization is completely eliminated. However, this invention utilizes the high consistency of polarization rebound within the population, transforming it from "interference" into a "reference benchmark," thus achieving reverse thinking. Similarly, this approach transforms the Joule heating, reversible entropy heating, heat dissipation, and heat capacity terms from the first law of thermodynamics into a recursive normal temperature prediction model. By statistically comparing temperature residuals across the population, it fully explores, utilizes, and transforms the relative consistency of heat dissipation conditions in practical engineering. This "mechanism-driven" construction approach ensures that each feature has a clear physical meaning (giving the solution excellent interpretability and engineering maintainability), and its dynamic threshold adapts to the population distribution (giving the solution the ability to generalize and discover unknown fault types), without relying on any historical label data (reducing the data threshold and time cost for algorithm implementation). Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the method flow of the dynamic evaluation method and system for the safety status of energy storage battery clusters based on feature decoupling, as described in Embodiment 1 of the present invention. Figure 2 This is a comparative analysis of the voltage response of the present invention and the prior art under static conditions. Detailed Implementation

[0027] The following detailed explanation illustrates the specific implementation methods: Example 1 The basic implementation examples are as follows: Figure 1 As shown: A dynamic assessment method for the safety status of energy storage battery clusters based on feature decoupling includes the following steps: S1 acquires real-time timing data of multiple individual cells in the battery system under test at a preset sampling frequency (e.g., 1Hz). This timing data includes at least the voltage and temperature of each individual cell, as well as the loop current of the battery system. This data is typically collected directly by the BMS and uploaded to the monitoring unit via a communication interface.

[0028] To improve the accuracy of subsequent analysis and avoid misjudgments caused by abnormal data points, a data cleaning step is preferably included before data analysis: removing data frames in the time-series running data that do not conform to physical laws or are obviously erroneous.

[0029] Specifically, time frame data in the time-series operation data that has a single battery voltage exceeding a preset normal voltage range (e.g., 2V to 5V), a temperature exceeding a preset normal temperature range (e.g., -20℃ to 80℃), or missing data, lost communication packets, or empty fields are excluded.

[0030] S2, Identify the operating conditions of the battery system based on the circuit current. The operating conditions include static conditions and non-static conditions. When the absolute value of the circuit current is less than a preset static current threshold and the duration exceeds a first preset time threshold, it is determined to be a static condition. Otherwise, it is determined to be a non-static condition. The non-static conditions include charging conditions and discharging conditions.

[0031] Specifically, the loop current data I(t) after data cleaning is acquired. A preset resting current threshold is set. ,Should It is a positive number close to zero, such as 0.5A or 1A, which can be set according to the accuracy of the current sensor and the specifications of the battery system.

[0032] In this embodiment, when the absolute value of the loop current |I(t)| is less than the preset resting current threshold... If this state persists for more than a first preset time threshold (e.g., 30 minutes), the battery system is determined to be in a "resting state". The resting state means that the battery is neither charged nor discharged. At this time, the electrochemical reactions inside the battery are mainly affected by self-discharge and polarization elimination effects.

[0033] Conversely, when the absolute value of the loop current is greater than or equal to... When the duration exceeds a short time threshold (e.g., 5 seconds, to avoid interference from transient current fluctuations), it is determined to be a "non-static condition". Non-static conditions further include charging conditions (…). ) and discharge conditions ( ).

[0034] After identifying the operating conditions, the continuous data stream is sliced ​​according to the time label. For stationary operating conditions, continuous stationary data segments are extracted; for non-stationary operating conditions, continuous charging or discharging data segments are extracted, with each data segment serving as an analysis unit.

[0035] S3, when identified as a static operating condition, executes an internal short-circuit micro-analysis step based on common-mode suppression; the core of this step lies in eliminating the common-mode polarization rebound signal shared by all batteries through "relative voltage" transformation, thereby highlighting the differential-mode leakage characteristics caused by the micro-internal short circuit. Specifically: For each individual cell i, its voltage is obtained at each time point t under static conditions. Simultaneously, the average voltage of all N individual cells in the battery system at the same moment is calculated and denoted as . .

[0036] Then construct its relative voltage sequence with respect to the average voltage of all individual cells in the battery system. : ; in, Let be the voltage of the i-th individual cell at time t. Let t be the average voltage of all individual cells in the battery system at time t.

[0037] The physical meaning of this expression is that after batteries are connected in series, during the initial resting period, all individual cells experience the same charge and discharge history, resulting in highly similar polarization rebound curves. By subtracting the group mean, the common-mode interference of polarization rebound can be significantly suppressed, while the linear voltage decay (i.e., differential-mode signal) unique to each individual cell caused by a micro-internal short circuit is preserved and amplified.

[0038] To further stabilize the detection, the present invention preferably selects a preset time period after the start of the static condition. Analysis will be performed. Specifically, the preset time period... The start time After being set to the second preset time threshold after the start of the static operation, the polarization voltage drastic change zone in the initial static phase is avoided; the second preset time threshold is 20 to 40 minutes.

[0039] In a preferred embodiment Take the sample after 30 minutes of settling. Take the time before the end of the settling period, or take a fixed time window length, for example, from... Start with 2 hours. Choosing a time period with good linearity can ensure the accuracy of subsequent linear fitting.

[0040] Furthermore, the preset time period under static working conditions Within this, the differential leakage rate of each individual cell is calculated. Differential leakage rate characterizes the rate of change of relative voltage over time, i.e., the rate of voltage decay caused by micro-internal short circuits.

[0041] In a direct calculation method, .

[0042] After obtaining the differential leakage rate of all individual cells, an adaptive decision-making mechanism based on population statistical distribution is used to identify anomalies. This involves comparing the differential leakage rate of each individual cell with the population statistical distribution of the differential leakage rate within the battery system to determine if any individual cells exhibit abnormalities due to micro-internal short circuits. Specifically, this includes: Real-time calculation of the average differential leakage rate of all individual cells in the battery system. and standard deviation ; When the differential leakage rate of the i-th single cell satisfy When the cell is in a state of abnormality, it is determined that there is a micro-internal short circuit. Here, k is a preset sensitivity coefficient with a value range of 3 to 6.

[0043] Due to the micro-internal short circuit, the relative voltage shows a decreasing trend. The values ​​are typically negative, and the larger the absolute value (i.e., the more negative), the more severe the anomaly. The "less than" relationship in the above criterion is precisely to capture this significant negative deviation.

[0044] The mean and standard deviation A sliding time window is used for dynamic updates to adapt to population characteristic drift caused by battery aging. For example, the results calculated under all idle operating conditions within the last 7 days are used. The values ​​form a statistical sample, which is calculated in real time. and This allows the decision boundary to adaptively follow the overall aging state of the battery system.

[0045] For ease of understanding, the following is combined with Figure 2 The signal processing effect of the present invention will be explained intuitively.

[0046] in, Figure 2 A illustrates the typical performance of existing technologies when directly monitoring the original cell voltage curves. In the initial resting period, the voltage of all cells shows an upward trend, even the red-marked faulty cells (with micro-internal short circuits) in the figure. This is because the voltage rebound caused by the elimination of polarization effects in the initial resting period (e.g., reaching +50mV) is far greater than the voltage drop caused by the micro-internal short circuit (e.g., only -2mV). Under these conditions, if the slope of the original voltage drop is directly calculated, the slope of all cells is positive, and the difference between faulty and normal cells is completely submerged in the common trend of polarization rebound. Therefore, existing technologies struggle to identify micro-internal short circuits in the early resting period using the original voltage curves, requiring several hours or even longer resting times until the polarization completely subsides, creating a significant detection blind zone.

[0047] In stark contrast, Figure 2B shows the results after processing with the relative voltage differential transformation (i.e., individual cell voltage minus the cluster average voltage) using the method proposed in this invention. This invention utilizes the strong correlation of the polarization responses of individual cells in a battery cluster series topology to define the nonlinear electrochemical polarization rebound component as intra-cluster common-mode interference, which is then suppressed by a simple calculation of individual cell voltage minus the cluster average voltage.

[0048] After this decoupling process, the present invention achieves two key signal transformations: First, background noise whitening: the voltage time-domain response of a normal cell, which was originally a non-stationary exponentially increasing curve (polarization-dominated), is mapped to a stationary random process fluctuating around the zero axis, completely eliminating the influence of the nonlinear trend term and making the statistical characteristics of the data more stable. Second, fault feature reconstruction: the ohmic leakage characteristics caused by micro-internal short circuits are orthogonally separated from the high-intensity polarization background noise, exhibiting a significant monotonic linear decay characteristic in the relative voltage sequence, while the relative voltages of other normal cells fluctuate randomly around the zero axis. This transformation greatly improves the signal-to-noise ratio of weak fault signals, allowing the system to accurately locate abnormal cells with micro-internal short circuits by calculating the linear regression slope of the relative voltage difference sequence during the early stage of rest (e.g., within a time window of 30 minutes to 2 hours after the start of rest) without waiting for the battery to be completely depolarized. This invention significantly overcomes the technical bottleneck of existing technologies having detection blind zones under polarization interference.

[0049] S4, when a non-static operating condition is identified, a thermal runaway early warning step based on electrothermal mechanism fusion is executed. The core of this step lies in using a physical model to eliminate normal temperature rises caused by Joule heating of the current, thereby identifying additional heat sources generated by abnormal internal chemical reactions (such as SEI film decomposition, positive electrode oxygen release, etc.). Specifically: Within a preset time window under non-static conditions, the thermal-electric coupling characteristic value of each individual cell is calculated. .

[0050] The preset time window under non-static operating conditions is a continuous charging period or a continuous discharging period, such as a continuous 10-minute charging segment.

[0051] The thermal-electric coupling characteristic value represents the abnormal temperature rise residual after deducting Joule heating, reversible entropy heating, and heat dissipation terms, and its calculation formula is as follows: ; in, ; ; Let the loop current be the current at the k-th sampling point. Let be the equivalent internal resistance of the i-th single cell at the k-th sampling point. The open-circuit voltage temperature coefficient, The equivalent heat dissipation coefficient, Let be the heat capacity parameter of the i-th individual cell. This refers to the ambient temperature or the reference ambient temperature within the cluster. The predicted temperature for the k-th sampling point. Let M be the measured temperature at the k-th sampling point, and M be the number of sampling points participating in the residual calculation within the preset time window. This represents the sampling time interval.

[0052] In the above formula, This represents the heat increment generated by the i-th individual cell during the k-th sampling interval due to the normal electrothermal process, where... Corresponding to the Joule heating term, Corresponding to the reversible entropy heat term; It is derived from the previous predicted temperature, normal heat increment, heat dissipation term, and heat capacity parameter; This represents the average residual between the measured temperature and the predicted temperature.

[0053] If the temperature rise of a single battery cell mainly comes from Joule heating, reversible entropy heating, and normal heat dissipation balance, then its The measured temperature should be relatively small and similar to that of other healthy cells within the battery system. Conversely, if an exothermic side reaction occurs within a cell, the measured temperature will consistently exceed the model-predicted temperature, leading to its... Abnormally high.

[0054] To avoid computational noise in transitional sections with low current or drastic current fluctuations, in another preferred embodiment, the thermo-electric coupling characteristic value... The calculation is performed only when the absolute value of the loop current is continuously greater than a preset current threshold (e.g., the current value corresponding to a 0.2C multiplier) within the preset time window, in order to avoid computational noise interference under low current conditions. If there is a long period of low current within the window, the window is removed or the overall window length is extended to ensure statistical stability.

[0055] Similar to S3, this step uses dynamic population statistical distribution to identify anomalies—based on the comparison of the population statistical distribution of the thermal-electric coupling characteristic values ​​of each individual cell with those of the thermal-electric coupling characteristic values ​​within the battery system, it determines whether there is a potential risk of thermal runaway due to an abnormal chemical heat source; specifically including: Real-time calculation of the mean thermal-electric coupling characteristics of all individual cells in the battery system and standard deviation ; When the thermal-electric coupling characteristic value of the i-th single cell satisfy When this occurs, it is determined that the single cell exhibits an abnormal pre-thermal runaway phenomenon; that is, there is an additional heat source exceeding the normal Joule heating effect.

[0056] The mean and standard deviation The results are dynamically updated using a sliding time window (e.g., the calculation results of all non-static operating condition windows within the most recent day).

[0057] Furthermore, in order to comprehensively assess the health status of the battery and promptly identify inconsistencies in capacity and internal resistance caused by manufacturing differences or long-term operation, a consistency assessment procedure is performed under charging conditions during non-static operation: For each individual battery cell, calculate its performance over the entire charging cycle. Intra-consistency anomaly index This index quantifies the cumulative deviation of the individual voltage curve from the population average voltage curve, and its calculation formula is as follows: ; in, Let be the voltage of the i-th individual cell at time t. Let be the average voltage of all individual cells in the battery system at time t; The moment charging begins. This is the point at which charging ends (e.g., when the full charge voltage is reached or the charging current drops to the cutoff value).

[0058] Real-time calculation of the average CAI value of all individual cells in the battery system and standard deviation It also uses a sliding time window for dynamic updates.

[0059] When the i-th single cell satisfy If this occurs, the individual battery cell is deemed to have a potential for consistent degradation. This indicator reflects a long-term, slow increase in dispersion, typically caused by factors such as increased internal resistance and uneven capacity decay.

[0060] S5, based on the judgment results of the internal short-circuit microanalysis step and / or the thermal runaway early warning step, generate and output corresponding early warning signals. Specifically, the early warning signals include three levels of early warning: When only the consistency anomaly index is present When an anomaly is detected, a first-level warning signal is output, indicating that the consistency within the battery pack has deteriorated and equalization maintenance or recharging of specific cells is required.

[0061] When the differential leakage rate When an anomaly is detected, a second-level warning signal is output, indicating the presence of a micro-internal short circuit. Since a micro-internal short circuit is an important precursor to thermal runaway, this warning indicates a potential safety risk to the battery system. It is recommended to arrange for a shutdown and professional inspection, and replace the abnormal cell if necessary.

[0062] When the thermal-electric coupling characteristic value An anomaly is detected, or the differential leakage rate is determined. With the aforementioned thermo-electric coupling characteristic value When an anomaly is detected (i.e., the same cell simultaneously exhibits micro-short circuit and abnormal temperature rise characteristics), a third-level warning signal is output. This warning indicates that the battery has entered the thermal runaway acceleration stage, and emergency measures should be taken immediately, prompting the need to urgently disconnect the circuit and activate the fire emergency plan.

[0063] To further improve the lead time for early warning and prevent missed detections due to occasional fluctuations in a single test, this invention also includes trend analysis of differential leakage rate under multiple consecutive static operating conditions.

[0064] Specifically, the differential leakage rate of each individual cell is recorded during multiple consecutive periods of static operation (e.g., static periods each day). And with time as the horizontal axis, Plot the leakage rate trend curve of this single cell on the vertical axis. Linear regression can be used to fit this curve to obtain the trend slope of the leakage rate change curve. When the trend slope is positive (indicating that the leakage rate is gradually increasing) and exceeds a preset trend threshold (e.g., 0.1 mV / h / day), even if the current... It has not yet exceeded the aforementioned static statistical threshold ( This also generates early warning signals for micro-internal short-circuit degradation trends. This trend analysis can enable the early detection of slowly deteriorating faults.

[0065] This embodiment also provides a dynamic safety status assessment system for energy storage battery clusters based on feature decoupling, used to execute the dynamic safety status assessment method for energy storage battery clusters based on feature decoupling as described above; it includes the following functional modules: The data acquisition module is used to acquire the time-series operating data of multiple individual cells in the battery system in real time at a preset sampling frequency; The operating condition identification module is used to identify the operating condition of the battery system based on the circuit current. The internal short-circuit microanalysis module is used to perform internal short-circuit microanalysis based on common-mode suppression when the operating condition identification module determines that the operating condition is static. Thermal runaway early warning module is used to execute thermal runaway early warning based on electrothermal mechanism fusion when the operating condition identification module determines that the operating condition is not static. The graded early warning output module is used to generate and output early warning signals of corresponding levels based on the judgment results of the internal short circuit analysis module and / or the thermal runaway early warning module.

[0066] The methods and systems described in this invention can be deployed on various hardware platforms. For example, they can be directly integrated into the microcontroller of a battery management system (BMS) to run in real time as embedded software, enabling vehicle-mounted warnings. Alternatively, they can be deployed on a cloud-based big data platform to receive data uploaded from multiple BMSs, perform centralized analysis and model training, and then push the warning results to a local monitoring terminal or the user's mobile phone. Regardless of the deployment method, it does not depart from the protection scope of this invention.

[0067] This embodiment provides a dynamic evaluation method and system for the safety status of energy storage battery clusters based on feature decoupling. It can adapt to battery aging and accurately extract fault features from complex operating data, and has high sensitivity and extremely low false alarm rate.

[0068] Example 2 The dynamic assessment method for the safety status of energy storage battery clusters based on feature decoupling has been adjusted as follows, based on Example 1: The differential leakage rate The calculation uses linear regression fitting: during the preset time period Within, the relative voltage sequence for each individual cell. Perform a least-squares linear fit, and use the slope of the fitted line as the differential leakage rate. .

[0069] This embodiment provides a dynamic evaluation method for the safety status of energy storage battery clusters based on feature decoupling. Compared with Embodiment 1, the differential leakage rate calculation utilizes multiple data points, resulting in more robust calculation results.

[0070] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A dynamic assessment method for the safety status of energy storage battery clusters based on feature decoupling, characterized in that, Includes the following steps: S1, acquire timing data of multiple individual cells in the battery system, wherein the timing data includes at least the voltage and temperature of each individual cell and the loop current of the battery system. S2, Identify the operating conditions of the battery system based on the circuit current, including the static operating conditions and the non-static operating conditions; S3, when identified as a static operating condition, perform the internal short-circuit microanalysis step based on common-mode suppression: For each individual cell, construct its relative voltage sequence with respect to the average voltage of all individual cells in the battery system. Preset time period under static working conditions Within this process, calculate the differential leakage rate for each individual cell: ; Based on the comparison results of the differential leakage rate of each individual cell and the group statistical distribution of the differential leakage rate in the battery system, it is determined whether there are abnormal individual cells caused by micro-internal short circuits. S4, when a non-static operating condition is identified, execute the thermal runaway early warning step based on the fusion of electrothermal mechanisms: Within a preset time window under non-static conditions, the thermal-electric coupling characteristic value of each individual cell is calculated. The thermal-electric coupling characteristic value represents the abnormal temperature rise residual after deducting Joule heating, reversible entropy heating, and heat dissipation terms, and its calculation formula is as follows: ; in, ; ； Let the loop current be the current at the k-th sampling point. Let be the equivalent internal resistance of the i-th single cell at the k-th sampling point. The open-circuit voltage temperature coefficient, The equivalent heat dissipation coefficient, Let be the heat capacity parameter of the i-th individual cell. This refers to the ambient temperature or the reference ambient temperature within the cluster. The predicted temperature for the k-th sampling point. Let M be the measured temperature at the k-th sampling point, and M be the number of sampling points participating in the residual calculation within the preset time window. The sampling time interval; Based on the comparison of the thermal-electric coupling characteristic values ​​of each individual cell with the group statistical distribution of the thermal-electric coupling characteristic values ​​within the battery system, it is determined whether there is a potential risk of thermal runaway due to abnormal chemical heat sources. S5. Based on the judgment results of the internal short-circuit microanalysis step and / or thermal runaway early warning step, generate and output the corresponding early warning signal.

2. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, When the absolute value of the circuit current is less than the preset static current threshold and the duration exceeds the first preset time threshold, it is determined to be a static working condition; otherwise, it is determined to be a non-static working condition. The non-static working condition includes charging working condition and discharging working condition. S4 also includes: performing a conformance assessment step during charging. For each individual battery cell, calculate its performance over the entire charging cycle. Intra-consistency anomaly index ; ； in, Let be the voltage of the i-th individual cell at time t. Let be the average voltage of all individual cells in the battery system at time t; The moment charging begins. This is the moment when charging ends; Real-time calculation of the average CAI value of all individual cells in the battery system and standard deviation ; When the i-th single cell satisfy At that time, it was determined that the single cell had a potential risk of consistent degradation.

3. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, In S3, determining whether there are abnormal individual cells due to micro-internal short circuits based on the comparison results of the differential leakage rate of each individual cell with the group statistical distribution of the differential leakage rate within the battery system specifically includes: Real-time calculation of the average differential leakage rate of all individual cells in the battery system. and standard deviation ; When the differential leakage rate of the i-th single cell satisfy When the cell is in a state of abnormality, it is determined that there is a micro-internal short circuit. Here, k is a preset sensitivity coefficient with a value range of 3 to 6. The mean and standard deviation A sliding time window is used for dynamic updates to adapt to population characteristic drift caused by battery aging.

4. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, In S4, the determination of whether there is a potential risk of thermal runaway due to abnormal chemical heat sources, based on the comparison results of the thermal-electric coupling characteristic values ​​of each individual cell with the group statistical distribution of the thermal-electric coupling characteristic values ​​within the battery system, specifically includes: Real-time calculation of the mean thermal-electric coupling characteristics of all individual cells in the battery system and standard deviation ; When the thermal-electric coupling characteristic value of the i-th single cell satisfy At that time, it was determined that the single cell showed signs of impending thermal runaway. The mean and standard deviation It uses a sliding time window for dynamic updates.

5. The dynamic assessment method for the safety status of energy storage battery clusters based on feature decoupling according to claim 2, characterized in that, The warning signals include three levels of warning: When only the consistency anomaly index is present When an anomaly is detected, a first-level warning signal is output, indicating that equalization maintenance is required; When the differential leakage rate When an anomaly is detected, a second-level warning signal is output, indicating that the machine needs to be stopped for inspection. When the thermal-electric coupling characteristic value An anomaly is detected, or the differential leakage rate is determined. With the aforementioned thermo-electric coupling characteristic value Simultaneously, when an anomaly is detected, a third-level warning signal is output, indicating that the circuit needs to be cut off urgently and the fire emergency plan needs to be activated.

6. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, In S3, the preset time period The start time After being set to the second preset time threshold after the start of the static operation, the polarization voltage drastic change zone in the initial static phase is avoided; the second preset time threshold is 20 to 40 minutes.

7. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, Before S2, there is also a data cleaning step: removing time frame data from the time-series running data where the voltage of a single battery cell exceeds the preset normal voltage range, or the temperature exceeds the preset normal temperature range, or the data is missing.

8. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, In S4, the preset time window under non-static operating conditions is a continuous charging period or a continuous discharging period, and the thermo-electric coupling characteristic value The calculation is performed only when the absolute value of the loop current is continuously greater than the preset current threshold within the preset time window, so as to avoid calculation noise interference under low current conditions.

9. The method for dynamic evaluation of the safety status of energy storage battery clusters based on feature decoupling according to claim 1, characterized in that, Also includes: Under multiple consecutive static conditions, the differential leakage rate of each individual cell was recorded in each static condition. The leakage rate trend curve of the single cell is constructed; when the slope of the leakage rate trend curve is positive and exceeds the preset trend threshold, a micro-internal short circuit degradation trend warning signal is generated.

10. A dynamic safety status assessment system for energy storage battery clusters based on feature decoupling, characterized in that, The method for performing dynamic safety status assessment of energy storage battery clusters based on feature decoupling as described in any one of claims 1-9 includes: The data acquisition module is used to acquire the time-series operating data of multiple individual cells in the battery system in real time at a preset sampling frequency; The operating condition identification module is used to identify the operating condition of the battery system based on the circuit current. The internal short-circuit microanalysis module is used to perform internal short-circuit microanalysis based on common-mode suppression when the operating condition identification module determines that the operating condition is static. Thermal runaway early warning module is used to execute thermal runaway early warning based on electrothermal mechanism fusion when the operating condition identification module determines that the operating condition is not static. The graded early warning output module is used to generate and output early warning signals of corresponding levels based on the judgment results of the internal short circuit analysis module and / or the thermal runaway early warning module.

Images