Method and system for detecting abnormal battery cells of an electrochemical energy storage system

Abstract

The embodiment of the application discloses an abnormal battery cell detection method and system of an electrochemical energy storage system. Firstly, the voltage, current and temperature data of part of the battery cells of the battery module are collected; the temperature of all the battery cells is preliminarily estimated through local inverse distance weighted interpolation, and the voltage residual is used for dynamic correction and smoothing to obtain a calibrated temperature field; further, the temperature gradient index of each battery cell based on the spatial neighborhood relationship and the voltage gradient index based on the electrical connection are calculated; finally, the joint criterion is constructed by fusing the temperature gradient index, the voltage gradient index and the temperature change rate, and compared with the adaptive dynamic threshold to realize the accurate diagnosis and early warning of the abnormal battery cells. The application effectively overcomes the problem of high early abnormality missing report and false report rate caused by the temperature monitoring blind area and the lack of spatial correlation analysis in the prior art, and significantly improves the sensitivity and reliability of the safety monitoring of the energy storage system without increasing the hardware cost.

Description

Technical Field

[0001] This invention relates to the field of battery safety management technology for electrochemical energy storage systems, specifically to a method and system for detecting abnormal battery cells in an electrochemical energy storage system. Background Technology

[0002] With the rapid development of new energy power generation and the increasing demand for peak shaving and frequency regulation in power systems, the application scale of electrochemical energy storage systems in grid-side, user-side, and new energy supporting scenarios continues to expand. Energy storage systems are typically composed of a large number of lithium-ion battery cells connected in series and parallel, and their operational safety is closely related to the consistency between the cells. The battery management system (BMS), as the core monitoring unit, needs to monitor key parameters such as voltage and temperature of each cell in real time to achieve early anomaly detection and safety warnings.

[0003] However, in practical engineering applications, existing battery anomaly detection technologies still have many limitations:

[0004] 1. Detection scheme based on individual cell voltage consistency: Anomalies are judged by comparing the deviation or range of the voltage of each individual cell with the average voltage of the module. Although this method is simple and easy to implement, it is not sensitive to early thermal-related anomalies caused by slight increase in internal resistance, poor contact, etc., and the static threshold is difficult to adapt to the dynamic baseline drift caused by the slow degradation of battery performance.

[0005] 2. Direct Temperature Monitoring and Threshold Alarm-Based Solution: This solution relies on temperature sensors (such as NTC thermistors) deployed on some individual battery cells for measurement. When the temperature exceeds a preset threshold, an alarm is triggered. This solution provides a direct response to severe faults such as localized overheating. However, due to the limited sensor density (usually lower than that of voltage sensors), it cannot achieve full-cell temperature coverage, resulting in monitoring blind spots.

[0006] 3. Data-driven model-based battery state estimation and fault diagnosis solutions: These methods utilize historical operational data to train machine learning models to estimate battery state of charge (SOC), state of health (SOH), or identify faults. While these methods possess a degree of intelligence, they have high requirements for data quality, feature engineering, and computing resources. Furthermore, the models have poor interpretability and are difficult to deploy in real-time on resource-constrained edge devices.

[0007] 4. Precise diagnostic solutions based on electrochemical impedance spectroscopy (EIS) or high-frequency parameter analysis: This method measures changes in internal battery parameters using excitation signals to achieve high-precision aging and fault diagnosis. However, this method requires dedicated hardware, is time-consuming and costly, and is primarily suitable for offline laboratory analysis; it is difficult to apply to online monitoring of large-scale energy storage systems.

[0008] In summary, existing technologies generally suffer from the following shortcomings:

[0009] 1) Temperature monitoring blind spots and hardware dependence

[0010] Option 2 (direct temperature monitoring) is limited by sensor density and cost, and cannot achieve direct, full-coverage monitoring of the temperature of all battery cells, resulting in significant monitoring blind spots. Option 4 (EIS) requires additional dedicated hardware, significantly increasing system complexity and cost, and is not suitable for the universal deployment of large-scale energy storage systems.

[0011] 2) Lack of spatial correlation analysis in battery anomaly diagnosis

[0012] Anomalies in individual battery cells often exhibit significant spatial correlations. For example, an increase in local internal resistance can cause differences in voltage and temperature response between that cell and its neighboring cells. However, the aforementioned approaches primarily analyze individual battery parameters from a time-series perspective or compare them with overall averages, failing to fully utilize the fixed spatial arrangement of batteries within the module. Anomalies (such as localized overheating or internal short circuits) often exhibit spatial propagation and neighborhood correlation characteristics; ignoring this dimension will result in the loss of crucial early diagnostic information.

[0013] 3) Insensitive to early and weak monomeric abnormalities

[0014] Option 1 (voltage consistency) is insufficient in identifying early thermally related anomalies (such as localized micro-overheating) caused by slight increases in internal resistance or contact impedance, which have not yet led to significant voltage deviations. Its static threshold setting also cannot effectively respond to dynamic baseline drift caused by slow degradation of battery performance.

[0015] 4) The system has high implementation costs and limited engineering applicability.

[0016] Option 3 (Complex Data-Driven Model) and Option 4 (EIS) typically have high requirements for data sampling frequency, computing resources, or testing conditions, making them difficult to deploy stably and efficiently in existing BMS or other edge hardware platforms operating in online environments with second-level sampling frequencies. Their complexity conflicts with the demands for reliability, real-time performance, and low cost in practical engineering. Summary of the Invention

[0017] To address this, embodiments of the present invention provide a method and system for detecting abnormal battery cells in an electrochemical energy storage system, thereby solving the technical problem that existing technologies cannot accurately detect early abnormalities in energy storage batteries under engineering-feasible conditions due to incomplete temperature sensor coverage and the lack of integration with voltage spatial neighborhood analysis.

[0018] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0019] According to a first aspect of the present invention, a method for detecting abnormal battery cells in an electrochemical energy storage system is provided, the method comprising:

[0020] The operating data of the target battery module is collected, including at least the voltage of all battery cells, the total current of the module, and the measured values ​​of temperature sensors arranged in some of the battery cells.

[0021] Based on the measured values ​​of the temperature sensor, the first estimated temperature of all battery cells in the target battery module is estimated by spatial interpolation method, and the voltage residual is calculated based on the voltage of all battery cells. The first estimated temperature is then corrected and smoothed according to the voltage residual to obtain the calibration temperature.

[0022] For each battery cell, its temperature neighborhood gradient exponent is calculated based on the calibration temperature, and its voltage neighborhood gradient exponent is calculated based on its voltage.

[0023] Based on the temperature neighborhood gradient exponent, the voltage neighborhood gradient exponent, and the time-series change rate of the calibration temperature, a joint criterion for determining the abnormality of a battery cell is constructed and compared with a preset dynamic threshold. When the joint criterion of any battery cell exceeds the dynamic threshold, the battery cell is determined to be an abnormal cell.

[0024] Further, the first estimated temperature of all battery cells in the target battery module is estimated using a spatial interpolation method, including:

[0025] Based on the Euclidean distance between the geometric center of the battery cell and the temperature sensor, a weighted mapping matrix from the temperature sensor to the battery cell is constructed.

[0026] For each battery cell, multiple nearby temperature sensors are selected, and interpolation calculations are performed using a local inverse distance weighting algorithm to obtain the first estimated temperature of the battery cell.

[0027] Further, the first estimated temperature is corrected based on the voltage residual, specifically including:

[0028] Calculate the sustained deviation of the voltage residual relative to its baseline obtained by recursive least squares filtering;

[0029] When the continuous deviation exceeds the preset residual threshold, it is determined that the corresponding battery cell has a thermal risk.

[0030] The coefficient for temperature correction is dynamically adjusted based on the total current of the module, and the correction coefficient is positively correlated with the square of the absolute value of the current.

[0031] The second estimated temperature is obtained by superimposing the product of the dynamically adjusted correction factor and the continuous deviation on the first estimated temperature of the battery cell with thermal risk.

[0032] Furthermore, temperature smoothing specifically involves:

[0033] The second estimated temperature is subjected to an exponentially weighted moving average to obtain the calibration temperature, wherein the smoothing coefficient is adaptively adjusted according to the rate of change of current.

[0034] Furthermore, the temperature neighborhood gradient exponent is calculated as follows:

[0035] Calculate the deviation between the calibration temperature of the target battery cell and the calibration temperature of its physically adjacent battery cells;

[0036] Different weighting coefficients are assigned to the nearest or diagonal neighborhoods based on the adjacency relationship, and the deviations are weighted and combined to obtain the temperature neighborhood gradient index.

[0037] Furthermore, the voltage neighborhood gradient exponent is calculated as follows:

[0038] Calculate the first-order component, which is the relative deviation of the voltage of the target battery cell from the average voltage of other battery cells in its electrical connection neighborhood.

[0039] Calculate the second-order component, which is the spatial second-order gradient of the target cell voltage based on the Laplacian operator;

[0040] The voltage neighborhood gradient exponent is obtained by weighted summation of the first-order component and the second-order component.

[0041] Furthermore, methods for generating dynamic thresholds include:

[0042] Calculate the dynamic mean and dynamic standard deviation of the joint criterion for all battery cells at the current moment;

[0043] The sensitivity coefficient used to calculate the threshold is adaptively adjusted based on the consistency of the state of charge of each individual battery cell in the battery module.

[0044] The dynamic threshold is obtained by adding the product of the dynamic mean, the sensitivity coefficient, and the dynamic standard deviation.

[0045] Furthermore, the method also includes:

[0046] If a battery cell is determined to be abnormal and simultaneously meets the following conditions: its calibrated temperature change rate is consistently positive and exceeds the first rate threshold, its voltage neighborhood gradient exponent is significantly higher than its historical statistical baseline, and the average temperature change rate of its physical neighboring battery cells exceeds the second rate threshold, then an early warning for thermal runaway is triggered.

[0047] According to a second aspect of the present invention, an abnormal cell detection system for an electrochemical energy storage system is provided, the system comprising:

[0048] The data acquisition and topology mapping module is used to acquire the operating data of the target battery module. The operating data includes at least the voltage of all battery cells, the total current of the module, and the measured values ​​of temperature sensors arranged in some of the battery cells.

[0049] The temperature field reconstruction and calibration module is used to estimate the first estimated temperature of all battery cells in the target battery module based on the measured value of the temperature sensor by spatial interpolation method, calculate the voltage residual based on the voltage of all battery cells, and correct and smooth the first estimated temperature according to the voltage residual to obtain the calibration temperature.

[0050] The neighborhood gradient analysis module is used to calculate the temperature neighborhood gradient index and the voltage neighborhood gradient index for each battery cell based on the calibration temperature.

[0051] The anomaly diagnosis module is used to construct a joint criterion for determining the anomaly of a battery cell based on the temperature neighborhood gradient index, the voltage neighborhood gradient index, and the time-series change rate of the calibration temperature, and compare it with a preset dynamic threshold. When the joint criterion of any battery cell exceeds the dynamic threshold, the battery cell is determined to be an abnormal cell.

[0052] Furthermore, the system also includes a thermal runaway early warning module for performing the following steps:

[0053] If a battery cell is determined to be abnormal and simultaneously meets the following conditions: its calibrated temperature change rate is consistently positive and exceeds the first rate threshold, its voltage neighborhood gradient exponent is significantly higher than its historical statistical baseline, and the average temperature change rate of its physical neighboring battery cells exceeds the second rate threshold, then an early warning for thermal runaway is triggered.

[0054] The embodiments of the present invention have the following advantages:

[0055] 1) Achieve battery cell temperature reconstruction without increasing hardware costs: By constructing a cell-level temperature field within the battery module, the temperature of battery cells without temperature sensors can be estimated, effectively reducing the dependence on the number of hardware sensors.

[0056] 2) Introducing local hotspot detection and temperature correction based on voltage residuals, as well as exponentially weighted moving average (EWMA), to improve the stability of temperature estimation: Temperature hotspots are locked by voltage residuals, and the hotspots are dynamically weighted to correct the estimated temperature. Finally, the temperature is calibrated by EWMA, which effectively suppresses the influence of operating noise and short-term disturbances and improves the stability of temperature time series.

[0057] 3) Enhanced anomaly identification sensitivity based on spatial arrangement neighborhood gradient analysis: By calculating the temperature neighborhood gradient index and voltage neighborhood gradient index between a single cell and its neighboring cells, the anomaly judgment is expanded from single-point analysis to spatial comparison analysis, which can identify local abnormal cells earlier.

[0058] 4) Combined voltage and temperature criteria to improve the accuracy of anomaly identification: By combining the voltage neighborhood gradient index and the temperature neighborhood gradient index, it is possible to effectively distinguish between normal fluctuations caused by changes in operating conditions and abnormal behavior caused by internal battery anomalies, thereby reducing the false alarm rate.

[0059] 5) Possesses early warning capability for thermal runaway: By jointly analyzing the rate of temperature change and voltage neighborhood gradient anomalies, it can identify potential risks before significant temperature rise or voltage anomalies occur in individual battery cells, providing advance warning for safety control and operation and maintenance decisions.

[0060] 6) Applicable to low sampling frequency and online operation scenarios: The method can be implemented based on 1 to 5 seconds of operation data, without the need for high-frequency sampling or complex electrochemical modeling, with low computational complexity, and is easy to deploy on existing BMS, edge computing devices or cloud platforms. Attached Figure Description

[0061] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0062] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0063] Figure 1 This is a schematic diagram of the logic structure of an abnormal battery cell detection system for an electrochemical energy storage system provided in an embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of the system architecture of an abnormal battery cell detection system for an electrochemical energy storage system provided in an embodiment of the present invention;

[0065] Figure 3A flowchart illustrating an abnormal battery cell detection method for an electrochemical energy storage system provided in an embodiment of the present invention;

[0066] Figure 4 This is a schematic diagram of the data flow in an abnormal battery cell detection method for an electrochemical energy storage system provided in an embodiment of the present invention. Detailed Implementation

[0067] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] With the continuous growth of new energy power generation and the demand for frequency regulation and peak shaving in power systems, the installed capacity of electrochemical energy storage systems in grid-side, user-side, and new energy supporting scenarios is showing a trend of large-scale development. Energy storage systems are typically composed of tens of thousands of battery cells connected in series and parallel, and their operational safety and consistency directly determine the reliability of the system. The battery management system (BMS), as the core monitoring unit, needs to complete the status assessment of all cells at a sampling frequency of seconds. However, due to cost limitations (the cost of a single temperature sensor is about 15-20 yuan) and structural complexity, the density of temperature sensors in actual engineering is generally lower than that of voltage sensors (e.g., a module with 32 cells only has 20 sensors), resulting in the inability to accurately locate the temperature of battery cells.

[0069] Existing technologies lack a way to effectively fuse voltage information and (direct / indirect) temperature information under engineering constraints of low sampling frequency and limited temperature sensor configuration, and to introduce battery spatial neighborhood relationships for collaborative analysis, thereby achieving early, accurate, and engineering-feasible detection of abnormal battery cells.

[0070] The core objective of this invention is to overcome the contradiction between sensor coverage and diagnostic sensitivity in the prior art, specifically solving the following technical problems:

[0071] 1) Temperature monitoring blind spots and hardware dependency issues:

[0072] How can we estimate the temperature of a single sensor unit without increasing the cost of sensor hardware, using easily measurable parameters such as voltage and current?

[0073] 2) Lack of spatial correlation analysis in battery anomaly diagnosis:

[0074] How to transform the fixed physical arrangement structure of battery modules into a calculable neighborhood gradient exponent so that the spatial propagation characteristics of abnormal cells can be quantified.

[0075] 3) The problem of distinguishing between operating condition disturbances and actual anomalies:

[0076] How to design a dynamic threshold mechanism to maintain a low false alarm rate (<1 time / thousand hours) while ensuring a high anomaly detection rate (>95%) under operating conditions such as SOC fluctuations and current surges.

[0077] 4) Problem of insensitivity to early and weak monomeric abnormalities:

[0078] How to achieve early warning during the stage of slight increase in internal resistance (temperature rise rate <0.5°C / min) rather than the stage of accelerated temperature rise.

[0079] 5) High system implementation cost and limited engineering applicability:

[0080] How can we leverage existing hardware at the edge, or deploy low-cost edge computing gateways, to achieve high-precision anomaly detection with lower computational costs without requiring large amounts of historical data for machine learning?

[0081] refer to Figure 1 and Figure 2 This invention discloses an abnormal cell detection system for an electrochemical energy storage system. The system includes: a data acquisition and topology mapping module 1; a temperature field reconstruction and calibration module 2; a neighborhood gradient analysis module 3; an anomaly diagnosis module 4; and an early warning module for thermal runaway 5.

[0082] Data acquisition and preprocessing module: used to acquire battery operating data and perform cleaning, alignment and filtering.

[0083] Cell temperature estimation and calibration module: Built-in algorithms such as local inverse distance weighted (IDW) interpolation, recursive least squares filtering (RLS), and exponentially weighted moving average (EWMA) are used to generate reliable temperature data for all battery cells.

[0084] Neighborhood gradient analysis engine: Calculates the temperature and voltage neighborhood gradient exponents for each cell based on a predefined battery module spatial topology.

[0085] Anomaly diagnosis and early warning module: Based on joint criteria and threshold rules, it performs anomaly judgment, level classification and early warning of thermal runaway risk.

[0086] Data storage and interface module: Used to store historical data, model parameters and diagnostic results, and provides an interface for interaction with host computer or cloud platform.

[0087] The system of this invention can be deployed as an advanced application module in the energy storage battery compartment of an energy storage power station energy management system (EMS). The core analysis algorithm is executed in the edge computing gateway, and the diagnostic results are returned to the EMS and BMS to guide the system operation and control.

[0088] The system can adopt a hierarchical edge computing architecture:

[0089] Sensing layer: Based on the BMS slave and master control modules, it realizes voltage / current acquisition and local filtering;

[0090] Edge computing layer: An embedded industrial gateway based on the ARM Cortex-A core, deploying S2-S5 algorithms written in C / C++ or Python.

[0091] Core board platform: Rockchip RK3568, a domestic platform with strong performance, high cost performance and complete interfaces.

[0092] CPU: ARM Cortex-A53.

[0093] RAM: ≥512MB DDR3 / DDR4.

[0094] Storage: ≥4GB, SQLite can be deployed to meet the needs of storing historical data, model parameters and diagnostic results.

[0095] Operating system: Pre-installed with Linux system (such as Ubuntu Core, Debian, OpenWrt or a vendor-customized system).

[0096] Sampling period: 1-5 seconds, configurable.

[0097] Power supply: 9-48VDC wide voltage input.

[0098] Installation method: DIN rail mounting on the energy storage compartment distribution cabinet.

[0099] Interface layer:

[0100] CAN-FD bus: connects to the BMS master controller.

[0101] Ethernet ports: at least 2, 100 / 1000Mbps, Modbus TCP interface for EMS.

[0102] LTE module: 4G full network compatibility, uploads to the cloud platform via MQTT.

[0103] Corresponding to the abnormal battery cell detection system for an electrochemical energy storage system disclosed above, this invention also discloses a method for detecting abnormal battery cells in an electrochemical energy storage system. The following details a method for detecting abnormal battery cells in an electrochemical energy storage system disclosed in this invention, in conjunction with the abnormal battery cell detection system for an electrochemical energy storage system described above.

[0104] refer to Figure 3 and Figure 4 This invention discloses a method for detecting abnormal battery cells in an electrochemical energy storage system, comprising: collecting operating data of a target battery module, wherein the operating data includes at least the voltage of all battery cells, the total current of the module, and the measured values ​​of temperature sensors arranged in some of the battery cells;

[0105] Based on the measured values ​​of the temperature sensor, the first estimated temperature of all battery cells in the target battery module is estimated by spatial interpolation method, and the voltage residual is calculated based on the voltage of all battery cells. The first estimated temperature is then corrected and smoothed according to the voltage residual to obtain the calibration temperature.

[0106] For each battery cell, its temperature neighborhood gradient exponent is calculated based on the calibration temperature, and its voltage neighborhood gradient exponent is calculated based on its voltage.

[0107] Based on the temperature neighborhood gradient exponent, the voltage neighborhood gradient exponent, and the time-series change rate of the calibration temperature, a joint criterion for determining the abnormality of a battery cell is constructed and compared with a preset dynamic threshold. When the joint criterion of any battery cell exceeds the dynamic threshold, the battery cell is determined to be an abnormal cell.

[0108] S1: Running data acquisition and topology mapping:

[0109] The system collects operational data of all cells in the target battery module within a set time window. The operational data includes at least the cell voltage, the total current of the module, and the temperature of some cells. The collected data is then time-aligned and preprocessed for missing values.

[0110] Based on the Euclidean distance between the geometric center of the battery cell and the temperature sensor, a topological mapping matrix of temperature sensor-battery cell is constructed.

[0111] S2: Single-unit temperature field reconstruction:

[0112] Based on the temperature sensor-cell topology mapping matrix constructed by S1, the temperature of the cell is estimated using the local inverse distance weighted (IDW) interpolation algorithm.

[0113] Temperature hotspots are locked by voltage residuals, and the estimated temperature is dynamically weighted and corrected by the hotspots. Finally, the temperature is calibrated by EWMA to obtain a stable cell temperature.

[0114] S3: Temperature Neighborhood Gradient Index: Calculates the absolute deviation of the target cell temperature from the average temperature of all its neighboring cells, and assigns different weight coefficients to the nearest neighbors (top, bottom, left, right) and diagonal neighbors (top left, bottom left, top right, bottom right) to weight and synthesize the standard deviation of the temperature gradient.

[0115] Voltage Neighborhood Gradient Exponent: Calculates the relative deviation between the target cell voltage and the average cell voltage of the electrical neighborhood (series nodes before and after in the same battery module), and calculates the second-order spatial gradient based on the Laplace operator.

[0116] S4: Combined abnormal diagnosis and risk assessment:

[0117] By combining the temperature neighborhood gradient index, voltage neighborhood gradient index, and the estimated time-series rate of temperature change, a joint criterion is constructed. When the joint criterion of any single entity exceeds a preset dynamic threshold, the single entity is determined to be an abnormal entity, and its abnormality level and type are output.

[0118] S5: Early warning of thermal runaway:

[0119] If a single cell is identified as abnormal, and its estimated temperature change rate remains positive and exceeds the threshold, while the voltage neighborhood gradient exponent increases significantly, an early warning of thermal runaway is triggered.

[0120] Example 1

[0121] A 100MW / 200MWh energy storage power station uses CATL LF280K battery cells. The battery module consists of 32 cells in 4 rows and 8 columns (size 173.93×71.65×207.2mm³), and is equipped with 20 NTC sensors in 4 rows and 5 columns, located on the aluminum bus between the cells.

[0122] S1: Running Data Acquisition and Topology Mapping

[0123] Total current of the acquisition module All individual unit voltages Temperature sensor data The collected data is time-aligned and missing values ​​are handled.

[0124] Constructing a temperature sensor-cell topology mapping matrix ,element Distance weights Let be the Euclidean distance between the geometric center of the i-th battery cell and the k-th temperature sensor (unit: mm; for a 280Ah high-capacity battery, the projected distance along the thickness direction of 207.2mm needs to be considered). Avoid dividing by zero.

[0125] S2: Single-unit temperature field reconstruction

[0126] S2-1: Full-field reconstruction based on local inverse distance weighted (IDW) interpolation

[0127] An isotropic Gaussian kernel weight decay function is used:

[0128]

[0129] in:

[0130] basis functions The Gaussian kernel is used, and the σ parameter is defined in engineering as the spatial interpolation characteristic length, which can be taken as 20~25% of the average spacing of the sensors.

[0131] The set of 3-4 sensors closest to cell i, with truncation distance. The weights of the long-range sensor are reset to zero to reduce the computational load.

[0132] S2-2: Local hotspot detection and temperature correction based on voltage residual

[0133] Engineering practice shows that voltage residual It is a sensitive indicator of increased local internal resistance. Although it cannot accurately invert temperature, it can identify hotspot locations. Meanwhile, due to various interference factors such as baseline drift, operating condition disturbances, and noise misjudgments in the voltage residual, filtering is required. Recursive Least Squares (RLS) is an adaptive filtering algorithm used to dynamically estimate the time-varying baseline of the voltage residual. Compared to moving average (MA) or exponentially weighted moving average (EWMA), RLS has advantages such as fast convergence, forgetting old data, and resistance to sudden changes in operating conditions, making it particularly suitable for engineering scenarios with random current variations in energy storage systems. The specific method is as follows:

[0134] 1) Calculate the average voltage of the battery module ;

[0135] 2) Calculate the residual voltage of the battery module ;

[0136] 3) Recursive least squares filtering (RLS) is used to estimate the residual baseline. Forgetting factor ;

[0137] 4) Abnormal indication: If continuous Instant satisfaction ( If the voltage measurement noise is 2mV, then it is determined that there is a risk of local hotspots in cell i.

[0138] S2-3: Dynamic Weighted Temperature Correction

[0139] For cells identified as having thermal risk, the correction factor is no longer fixed, but dynamically adjusted with the square of the current to adapt to high-rate operating conditions.

[0140]

[0141]

[0142] in:

[0143] Taking CATL's 280Ah lithium iron phosphate battery cell as an example,

[0144] Can be used as a benchmark coefficient:

[0145] when hour, It effectively compensates for the decrease in indication sensitivity caused by increased voltage noise under high current.

[0146] S2-4: Exponentially Weighted Moving Average (EWMA) Smoothing Temperature

[0147] Smoothing the reconstructed temperature using an exponentially weighted moving average (EWMA):

[0148]

[0149] in:

[0150] Smoothing coefficient α is adaptively adjusted:

[0151] steady state (e.g.: ): α=0.25, enhanced smoothing;

[0152] Dynamic (e.g.: ): α=0.5, improves response.

[0153] S3: Calculation of dual-scale neighborhood gradient field

[0154] S3-1: Temperature neighborhood gradient exponent

[0155] To address the thermal coupling between the width and diagonal of energy storage batteries, an extended hybrid 8-neighborhood model is proposed:

[0156]

[0157] in:

[0158] Nearest Neighbor Domain Physical four-neighborhood (up, down, left, right), distance weight ;

[0159] diagonal neighborhood The distance to the four diagonal neighbors is increased by √2, and the distance weight is calculated accordingly. ;

[0160] Can be adopted This refers to the temperature difference during normal operation of the battery module.

[0161] S3-2: Voltage neighborhood gradient exponent

[0162] Voltage neighborhood is defined as electrical connection neighborhood (series nodes in the same battery module), with weights decoupled from physical distance:

[0163]

[0164] in:

[0165] Voltage consistency weight (First-order deviation term):

[0166] Physical meaning: It measures the relative deviation between the voltage of a single cell and the average voltage of its electrical neighborhood (in engineering, this can be achieved by connecting four cells in series, i).

[0167] Detection target: Global consistency disruption, such as a systematic increase in the internal resistance of a certain branch;

[0168] Based on engineering practice, This is the optimal solution.

[0169] Spatial curvature weight (Second-order gradient term):

[0170] Physical meaning: It measures the spatial second-order gradient of the voltage of a single cell in its neighborhood (Laplace operator), reflecting the degree of "bending" of the local voltage field;

[0171] Detection targets: local mutations, such as micro-short circuits within monomers or polarization mutations caused by local overheating;

[0172] Based on engineering practice, This is the optimal solution.

[0173] Discrete Laplace operator :

[0174]

[0175] Electrical neighborhood: This refers to the set of cells that are directly connected in series with cell i in terms of electrical connection. In engineering, four battery cells can be selected and connected in series with cell i.

[0176] Statistical fluctuation baseline of second-order gradient in voltage space :

[0177] Offline statistical calibration can be used (sample size > 10,000 points):

[0178]

[0179] S4: Cell Abnormal Diagnosis

[0180] 1) Anomaly scoring:

[0181]

[0182] The weights β=[0.5,0.3,0.2] can be optimized using Fisher's discrimination method based on offline historical data.

[0183] 2) Dynamic threshold:

[0184]

[0185] Among them, the dynamic mean of the anomaly score

[0186]

[0187] Dynamic standard deviation of outlier ratings :

[0188]

[0189] Adaptive coefficients

[0190]

[0191] in, The basic sensitivity coefficient can take values ​​of... , The average absolute deviation of the SOC of each cell within the battery module from the average value:

[0192]

[0193] This is the SOC consistency penalty coefficient, which can take values ​​of: .

[0194] 3) Abnormal level:

[0195] For example:

[0196] Level 1 alarm ( PCS load reduction by 50%;

[0197] Level 2 alarm ( PCS load reduction by 80%;

[0198] Level 3 alarm ( ): Alarm reminder.

[0199] S5: Multi-level early warning system for thermal runaway

[0200] The following conditions must be tested simultaneously:

[0201] 1) Temperature monotonicity test

[0202]

[0203] 2) Voltage gradient surge

[0204]

[0205] 3) Abnormal thermal diffusion

[0206]

[0207] in: .

[0208] If conditions 1, 2, and 3 are met simultaneously, an early warning of thermal runaway will be triggered.

[0209] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method of detecting abnormal cell of an electrochemical energy storage system, characterized by, The method includes: The operating data of the target battery module is collected, including at least the voltage of all battery cells, the total current of the module, and the measured values ​​of temperature sensors arranged in some of the battery cells. Based on the measured values ​​of the temperature sensor, the first estimated temperature of all battery cells in the target battery module is estimated by spatial interpolation method, and the voltage residual is calculated based on the voltage of all battery cells. The first estimated temperature is then corrected and smoothed according to the voltage residual to obtain the calibration temperature. For each battery cell, its temperature neighborhood gradient exponent is calculated based on the calibration temperature, and its voltage neighborhood gradient exponent is calculated based on its voltage. Based on the temperature neighborhood gradient exponent, the voltage neighborhood gradient exponent, and the time-series change rate of the calibration temperature, a joint criterion for determining the abnormality of a battery cell is constructed and compared with a preset dynamic threshold. When the joint criterion of any battery cell exceeds the dynamic threshold, the battery cell is determined to be an abnormal cell. The first estimated temperature of all individual cells in the target battery module is estimated using a spatial interpolation method, including: Based on the Euclidean distance between the geometric center of the battery cell and the temperature sensor, a weighted mapping matrix from the temperature sensor to the battery cell is constructed. For each battery cell, multiple nearby temperature sensors are selected, and interpolation calculations are performed using a local inverse distance weighting algorithm to obtain the first estimated temperature of the battery cell. The first estimated temperature is corrected based on the voltage residual, specifically including: Calculate the sustained deviation of the voltage residual relative to its baseline obtained by recursive least squares filtering; When the continuous deviation exceeds the preset residual threshold, it is determined that the corresponding battery cell has a thermal risk. The coefficient for temperature correction is dynamically adjusted based on the total current of the module, and the correction coefficient is positively correlated with the square of the absolute value of the current. The first estimated temperature of a battery cell with thermal risk is superimposed with the product of a dynamically adjusted correction factor and the continuous deviation to obtain the second estimated temperature. The temperature neighborhood gradient exponent is calculated as follows: Calculate the deviation between the calibration temperature of the target battery cell and the calibration temperature of its physically adjacent battery cells; Different weighting coefficients are assigned to the nearest or diagonal neighborhoods based on the adjacency relationship, and the deviations are weighted and combined to obtain the temperature neighborhood gradient index. The voltage neighborhood gradient exponent is calculated as follows: Calculate the first-order component, which is the relative deviation of the voltage of the target battery cell from the average voltage of other battery cells in its electrical connection neighborhood. Calculate the second-order component, which is the spatial second-order gradient of the target cell voltage based on the Laplacian operator; The voltage neighborhood gradient exponent is obtained by weighted summation of the first-order component and the second-order component.

2. The method of claim 1, wherein the method further comprises: Temperature smoothing specifically involves: The second estimated temperature is subjected to an exponentially weighted moving average to obtain the calibration temperature, wherein the smoothing coefficient is adaptively adjusted according to the rate of change of current.

3. The method for detecting abnormal battery cells in an electrochemical energy storage system as described in claim 1, characterized in that, Methods for generating dynamic thresholds include: Calculate the dynamic mean and dynamic standard deviation of the joint criterion for all battery cells at the current moment; The sensitivity coefficient used to calculate the threshold is adaptively adjusted based on the consistency of the state of charge of each individual battery cell in the battery module. The dynamic threshold is obtained by adding the product of the dynamic mean, the sensitivity coefficient, and the dynamic standard deviation.

4. The method for detecting abnormal battery cells in an electrochemical energy storage system as described in claim 1, characterized in that, The method further includes: If a battery cell is determined to be abnormal and simultaneously meets the following conditions: its calibrated temperature change rate is consistently positive and exceeds the first rate threshold, its voltage neighborhood gradient exponent is significantly higher than its historical statistical baseline, and the average temperature change rate of its physical neighboring battery cells exceeds the second rate threshold, then an early warning for thermal runaway is triggered.

5. An abnormal cell detection system for an electrochemical energy storage system, characterized in that, The system is used to perform the abnormal battery cell detection method as described in claim 1, specifically including: The data acquisition and topology mapping module is used to acquire the operating data of the target battery module. The operating data includes at least the voltage of all battery cells, the total current of the module, and the measured values ​​of temperature sensors arranged in some of the battery cells. The temperature field reconstruction and calibration module is used to estimate the first estimated temperature of all battery cells in the target battery module based on the measured value of the temperature sensor by spatial interpolation method, calculate the voltage residual based on the voltage of all battery cells, and correct and smooth the first estimated temperature according to the voltage residual to obtain the calibration temperature. The neighborhood gradient analysis module is used to calculate the temperature neighborhood gradient index and the voltage neighborhood gradient index for each battery cell based on the calibration temperature. The anomaly diagnosis module is used to construct a joint criterion for determining the anomaly of a battery cell based on the temperature neighborhood gradient index, the voltage neighborhood gradient index, and the time-series change rate of the calibration temperature, and compare it with a preset dynamic threshold. When the joint criterion of any battery cell exceeds the dynamic threshold, the battery cell is determined to be an abnormal cell.

6. The abnormal cell detection system for an electrochemical energy storage system as described in claim 5, characterized in that, The system also includes a thermal runaway early warning module for performing the following steps: If a battery cell is determined to be abnormal and simultaneously meets the following conditions: its calibrated temperature change rate is consistently positive and exceeds the first rate threshold, its voltage neighborhood gradient exponent is significantly higher than its historical statistical baseline, and the average temperature change rate of its physical neighboring battery cells exceeds the second rate threshold, then an early warning for thermal runaway is triggered.

Images