A lithium ion battery storage cabinet safety protection integrated system

Abstract

The application discloses a lithium ion battery storage cabinet safety protection integrated system and relates to the technical field of lithium battery energy storage safety. The system synchronously collects single-cell surface and tab double-point temperature time sequence data and cabinet internal environment temperature and airflow state background data through a multi-source high-frequency acquisition module. An environmental background compensation module adopts a disturbance compensation algorithm of airflow heat exchange and environmental temperature change coupling, accurately calculates and eliminates cabinet airflow and external temperature change interference, and outputs real cell temperature sequence. A temperature rise curvature fusion identification module solves temperature rise rate and temperature rise curvature through differential operation, constructs a curvature mutation amplitude and duration dual-dimension feature library, establishes a mapping model of curvature features and cell internal short circuit and lithium precipitation thermal runaway incipient state, realizes accurate identification of micro-amplitude implicit thermal anomalies, and a risk grading interlocking module divides three risk grades according to curvature features and outputs grading interlocking control signals to drive an execution mechanism to act.

Description

Technical Field

[0001] This invention relates to the field of lithium battery energy storage safety technology, specifically to an integrated safety protection system for lithium-ion battery storage cabinets. Background Technology

[0002] Lithium-ion batteries are widely used in electric vehicles, energy storage power stations, and many other fields due to their advantages such as high energy density and long cycle life. During the static storage of lithium-ion batteries, their safety is of paramount importance. However, due to differences in battery manufacturing processes, changes in the usage environment, and complex electrochemical reactions inside the battery, lithium-ion batteries may experience early abnormalities such as internal micro-short circuits and lithium plating during storage. If these early abnormalities are not detected and dealt with in time, they will gradually accumulate heat and may eventually lead to thermal runaway safety accidents, resulting in serious consequences such as battery fires and explosions. This will not only cause huge economic losses but may also endanger human lives.

[0003] Most existing battery storage cabinet safety protection systems rely on fixed temperature or temperature rise rate thresholds to determine whether a battery is abnormal. This type of technology has many limitations: First, it performs poorly in resisting environmental interference. The actual environment in which battery storage cabinets are located is complex and variable. Factors such as airflow within the cabinet and fluctuations in ambient temperature can interfere with battery temperature measurements. Traditional systems fail to effectively consider and eliminate these interferences, resulting in inaccurate temperature data and affecting the judgment of battery status. Second, it cannot identify subtle, latent thermal anomalies. Temperature changes caused by early anomalies such as internal micro-short circuits and lithium plating are often very weak. Traditional methods based on fixed thresholds are unable to capture these subtle thermal change signals, easily leading to missed detection of early thermal anomalies and preventing timely discovery of safety hazards. Third, the protection strategy is simplistic. Traditional systems typically only set simple threshold alarms, lacking segmentation and targeted protection measures for different risk levels. They cannot adopt differentiated treatment methods based on the severity of thermal anomalies, making it difficult to achieve precise protection. Finally, it lacks adaptive optimization capabilities. As batteries are used for longer periods, their performance gradually changes. At the same time, different usage environments also affect the battery's state. Traditional systems cannot dynamically adjust the judgment parameters and protection strategies according to these changes, resulting in a decline in the reliability and accuracy of the system's long-term operation, making it difficult to meet the high-precision and high-reliability safety protection requirements of lithium battery storage cabinets. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an integrated safety protection system for lithium-ion battery storage cabinets. This system uses a multi-source high-frequency acquisition module to simultaneously collect time-series temperature data from both the surface and tabs of a single battery cell, as well as ambient temperature and airflow background data within the cabinet. An environmental background compensation module employs a disturbance compensation algorithm that couples airflow heat transfer with environmental temperature changes to accurately calculate and eliminate interference from airflow within the cabinet and external temperature variations, outputting a true battery cell temperature sequence. A temperature rise curvature fusion and identification module uses differential operations to solve for the temperature rise rate and curvature, constructing a dual-dimensional feature library of curvature mutation amplitude and duration. This establishes a mapping model between curvature features and the initial states of short circuits and lithium plating thermal runaway within the battery cell, enabling accurate identification of subtle, latent thermal anomalies. A risk grading and interlocking module classifies risks into three levels based on curvature features and outputs grading interlocking control signals to drive the actuators. A closed-loop self-optimization module uses a sliding window statistical algorithm and an incremental learning algorithm to achieve self-calibration of feature library parameters and online iterative updates of the mapping model.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an integrated safety protection system for lithium-ion battery storage cabinets, the system comprising: a multi-source high-frequency acquisition module, an environmental background compensation module, a temperature rise curvature fusion identification module, a risk classification interlocking module, a closed-loop self-optimization module, and an actuator module; The multi-source high-frequency acquisition module synchronously acquires time-series data of single cell surface temperature and tab temperature, as well as background data of cabinet internal ambient temperature and airflow status. The environmental background compensation module uses an environmental temperature disturbance compensation algorithm to eliminate the interference of airflow disturbance in the cabinet and external environmental temperature changes on the battery cell temperature acquisition data, and outputs a corrected, disturbance-free temperature time series. The temperature rise curvature fusion identification module performs differential operations on the corrected temperature time series, solves the first-order change rate of cell temperature rise and second-order curvature differential features, constructs a two-dimensional feature library of curvature mutation amplitude and duration, and establishes a mapping model between curvature features and the initial state of short circuit and lithium plating thermal runaway in the cell, so as to realize the identification of micro-amplitude hidden thermal anomalies. The risk classification interlocking module matches the corresponding risk level according to the curvature characteristics, generates and outputs a classification interlocking control signal; The closed-loop self-optimization module realizes self-calibration of feature library parameters and online iterative update of the mapping model based on operational feedback data; The actuator module receives hierarchical interlocking control signals and executes corresponding safety protection actions.

[0006] Furthermore, the environmental background compensation module uses an environmental temperature disturbance compensation algorithm. Calculate the temperature measurement interference caused by airflow convection heat transfer within the cabinet and fluctuations in ambient temperature. It is the total temperature measurement interference caused by the combined effect of airflow disturbance in the cabinet and temperature fluctuation in the external environment. It is used to characterize the overall deviation error of environmental factors on the actual temperature acquisition value of the battery cell. The heat transfer coefficient of the cell surface airflow is obtained by fitting the real-time airflow velocity inside the cabinet. The fitting relationship is pre-calibrated through wind tunnel testing. It is the total effective heat exchange area between the cell casing and the two tabs, which is the calibrated value for different specifications of cells; This is the weighted average of the ambient temperatures at multiple nodes inside the cabinet. This is the original temperature value collected by the battery cell; The system sampling interval; The equivalent thermal capacity of a single battery cell is obtained through calibration using a calorimeter. This is the environmental temperature change compensation coefficient; It is obtained by performing real-time first-order differential calculations on the time-series data of ambient temperature.

[0007] Furthermore, the temperature rise curvature fusion identification module performs continuous real-time numerical differentiation operations on the actual temperature time series of the battery cell after environmental background compensation, using the sampling interval as the time base. By solving the first-order differential of the standardized temperature time series data T(t), the first-order rate of change of the battery cell temperature rise is obtained. This reflects the rate of change of cell temperature over time. Based on this temperature rise rate, a second-order differential calculation is performed to solve for the second-order curvature differential characteristic of the cell temperature rise. It is used to characterize the degree of abrupt change and the curvature of the temperature rise trend.

[0008] Furthermore, the temperature rise curvature fusion identification module constructs a two-dimensional feature library with curvature mutation amplitude and mutation duration as the core. Combining the heat generation patterns of the initial stage of thermal runaway such as short circuit and lithium plating in the cell, it presets three types of feature threshold ranges in the feature library: normal operating conditions, micro-amplitude latent thermal anomalies, and manifest thermal faults. Among them, latent thermal anomalies correspond to curvature fluctuation ranges with small amplitude and short duration, while manifest thermal faults correspond to curvature over-limit ranges with large amplitude and long duration. By tracking the instantaneous peak value and continuous over-limit duration of curvature features in real time through a time-series sliding window, the real-time calculated curvature features are matched with the feature library ranges to distinguish the curvature feature morphology corresponding to different thermal anomalies.

[0009] Furthermore, after calculating the cell temperature rise rate and second-order curvature features, the temperature rise curvature fusion identification module uses the curvature mutation amplitude k and duration t as dual-dimensional features as model inputs. Based on a large amount of experimental data on the initial stages of thermal runaway such as short circuits and lithium plating within the cell, a feature mapping classification model is constructed. This model is established using a method of offline training followed by online adaptive matching. The distribution patterns of curvature features corresponding to early thermal anomalies of the cell under different operating conditions are labeled and learned to form a correspondence between features and fault types. During online operation, the difference between real-time features and preset benchmark features within the model is quantified through a feature similarity calculation formula, and classification decision logic is applied. Compare and distinguish. To pre-determine the similarity threshold, the system distinguishes between normal temperature rise fluctuations, normal operating condition disturbances and internal short circuits, and subtle latent thermal anomalies caused by lithium plating. For latent anomaly signals with curvature that deviates slightly from the reference range, have a short duration, and do not show obvious temperature exceedances, the system achieves effective identification through refined feature boundary determination. At the same time, it filters out non-faulty minor fluctuation interference to avoid misjudgment.

[0010] Furthermore, the temperature rise curvature fusion identification module quantifies the difference between real-time features and preset benchmark features within the model using a feature similarity calculation formula, which is: Where d is the similarity between the currently calculated curvature abrupt change amplitude and the curvature baseline value of the corresponding thermal anomaly type in the feature library. , The weighting coefficients for curvature amplitude and duration are determined through cross-validation during offline training. , The normalized value of the real-time feature. , The normalized value is the baseline feature.

[0011] Furthermore, the risk grading interlocking module receives the curvature change amplitude k and duration t, two-dimensional feature parameters, from the temperature rise curvature fusion identification module in real time, solidifies the three-level risk threshold determination logic, and presets the curvature amplitude threshold and duration threshold corresponding to each level. Specifically, the first level low risk corresponds to k being in the micro-fluctuation range and t < t1, where t1 is the preset short-term threshold; the second level medium risk corresponds to k ≥ k1 and t ≥ t1, where k1 is the curvature steep increase threshold; and the third level high risk corresponds to k ≥ k2 and t ≥ t2, where k2 is the curvature extreme value threshold and t2 is the preset long-term threshold.

[0012] Furthermore, the risk-level interlocking module compares the received curvature features with preset thresholds in real time to determine the current risk level, and then outputs corresponding level interlocking control signals. The first-level risk outputs a PWM frequency conversion drive signal to drive the zonal directional pre-cooling unit to operate; the second-level risk outputs a high-level electrical isolation drive signal to control the disconnection of the battery cell branch electrical isolation unit; and the third-level risk synchronously outputs three-way linkage drive signals for fire start, cabinet interlocking, and electrical isolation to ensure coordinated operation of each execution unit.

[0013] Furthermore, the closed-loop self-optimization module collects real-time feedback data from the entire system operation process, including the action feedback signals of the actuators, the thermal anomaly identification results of the temperature rise curvature fusion identification module, the real-time operating status data of the battery cells, and complete records of various abnormal events. Simultaneously, it stores the temperature time-series data and curvature feature parameters after environmental background compensation, performs statistical analysis on historical operating data within a certain period, dynamically adjusts the curvature mutation amplitude thresholds k1 and k2 and the duration thresholds t1 and t2 corresponding to each level of risk in the feature library, corrects feature offsets caused by battery cell aging, environmental temperature changes, and airflow fluctuations, and adopts a lightweight incremental learning algorithm for the mapping model. It incorporates newly added thermal anomaly samples such as internal short circuits and lithium plating, as well as normal operation samples, into the model training in real time, updates the benchmark parameters and classification decision thresholds in the feature similarity calculation formula, optimizes the internal discrimination weights of the model, and realizes closed-loop linkage between feature library parameter self-calibration and online iteration of the mapping model.

[0014] Furthermore, the actuator module includes a zoned directional preheating unit, a cell branch electrical isolation unit, a targeted fire suppression unit, and a full cabinet interlocking unit; the zoned directional preheating unit is a variable frequency axial flow fan, the cell branch electrical isolation unit is a high-voltage solid-state relay, the targeted fire suppression unit is an ultra-fine aerosol fire extinguishing device, and the full cabinet interlocking unit is an electromagnetic safety lock.

[0015] Compared with existing technologies, this integrated safety protection system for lithium-ion battery storage cabinet has the following advantages: This invention employs an environmental background compensation module that uses a disturbance compensation algorithm coupling airflow heat transfer and environmental temperature changes, combined with adaptive Kalman filtering. This accurately eliminates airflow disturbances within the cabinet and interference from external environmental temperature changes, effectively ensuring the authenticity and validity of the cell temperature data and greatly improving the system's reliability in complex environments. For anomaly identification, the temperature rise curvature fusion identification module calculates the temperature rise rate and temperature rise curvature through real-time numerical differentiation, constructing a two-dimensional feature library of curvature mutation amplitude and duration. Furthermore, it establishes a mapping model between curvature features and the initial states of short circuits and lithium plating thermal runaway within the cell. It can accurately identify subtle, hidden thermal anomalies that traditional threshold methods cannot capture. In terms of protection response, the risk-level interlocking module adopts a CPLD programmable hardware logic architecture with internally embedded three-level risk threshold judgment logic. It can quickly determine the risk level and output graded interlocking control signals to achieve graded and precise protection. In addition, the closed-loop self-optimization module relies on a large-capacity data storage unit and uses sliding window statistical algorithms and incremental learning algorithms to achieve self-calibration of feature library parameters and online iterative updates of the mapping model. This enables the system to adapt to cell aging and changes in operating conditions, and maintain high-precision safety protection capabilities over the long term.

[0016] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 A flowchart of an integrated safety protection system for a lithium-ion battery storage cabinet; Figure 2 A flowchart of an environmental background compensation module for an integrated safety protection system for lithium-ion battery storage cabinets; Figure 3 This is a flowchart of an integrated safety protection system for lithium-ion battery storage cabinets. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0020] This invention provides an integrated safety protection system for lithium-ion battery storage cabinets, such as... Figure 1 As shown, the multi-source high-frequency acquisition module synchronously collects time-series temperature data of the single cell surface and electrode dual points, as well as ambient temperature and airflow background data inside the cabinet. The environmental background compensation module adopts a disturbance compensation algorithm that couples airflow heat transfer and environmental temperature change to accurately calculate and eliminate interference from airflow and external temperature change in the cabinet, and outputs the true cell temperature sequence. The temperature rise curvature fusion identification module solves the temperature rise rate and temperature rise curvature through differential operation, constructs a two-dimensional feature library of curvature change amplitude and duration, and establishes a mapping model between curvature features and the initial state of short circuit and lithium plating thermal runaway in the cell, so as to achieve accurate identification of micro-amplitude hidden thermal anomalies. The risk classification interlocking module divides the risk level into three levels according to the curvature features and outputs graded interlocking control signals to drive the actuator. The closed-loop self-optimization module realizes the self-calibration of feature library parameters and online iterative update of mapping model through sliding window statistical algorithm and incremental learning algorithm.

[0021] The multi-source high-frequency acquisition module employs a global clock synchronization triggering mechanism to synchronously acquire fine-grained parameters at the battery cell level and environmental parameters across the entire storage cabinet, eliminating data timing misalignment issues. At the battery cell monitoring level, point-to-point sensor units are deployed for each individual battery cell within the storage cabinet, with temperature acquisition nodes positioned at the center temperature measurement area of ​​the cell casing, the positive electrode tab welding area, and the negative electrode tab welding area. Continuous time-series data on the cell surface temperature and the temperatures of the positive and negative electrode tabs are acquired synchronously, accurately capturing temperature changes in the core heat-generating area of ​​the battery cell. At the cabinet environmental monitoring layer… On the other hand, environmental sensor nodes are deployed in different height zones (upper, middle, and lower) inside the cabinet, as well as in the gaps between the battery cell arrays and at the inlet and outlet positions of the air ducts. These nodes collect background status data such as ambient temperature, airflow velocity, and airflow direction inside the cabinet in real time. All collected analog signals are converted from analog to digital to generate digital time-series data with high-precision timestamps. This data is then stably transmitted to the back-end computing module via a shielded bus, minimizing the impact of electromagnetic interference on data transmission. This provides a high-fidelity, highly synchronized raw data foundation for subsequent environmental interference compensation and temperature rise feature identification, ensuring the accuracy of subsequent algorithm calculations.

[0022] The environmental background compensation module calls the preset corrected environmental temperature disturbance compensation algorithm. The overall temperature measurement interference is calculated for each sampling point, where, It is the total temperature measurement interference caused by the combined effect of airflow disturbance in the cabinet and temperature fluctuation in the external environment. It is used to characterize the overall deviation error of environmental factors on the actual temperature acquisition value of the battery cell. The heat transfer coefficient of the cell surface airflow is obtained by fitting the real-time airflow velocity inside the cabinet. The fitting relationship is pre-calibrated through wind tunnel testing. It is the total effective heat exchange area between the cell casing and the two tabs, which is the calibrated value for different specifications of cells; This is the weighted average of the ambient temperatures at multiple nodes inside the cabinet. This is the original temperature value collected by the battery cell; The system sampling interval; The equivalent thermal capacity of a single battery cell is obtained through calibration using a calorimeter. The environmental temperature change compensation coefficient is determined as follows: Three typical temperature change rates of 0.5℃ / min, 1℃ / min, and 2℃ / min are set in the constant temperature chamber. The temperature difference between the measured value of the non-self-heating battery cell and the actual ambient temperature is collected, and obtained through linear regression fitting. The relationship between the temperature change rate and the value range is as follows: ; By performing real-time first-order differential calculations on environmental temperature time-series data, the instantaneous change trend of environmental temperature can be accurately characterized. The total amount of temperature measurement interference obtained from the calculation The temperature data of the battery cell is deducted point by point in real time from the original temperature data of the corresponding moment, so as to complete the basic decoupling of environmental disturbances and remove the temperature acquisition offset error caused by airflow heat exchange and environmental temperature change. An adaptive Kalman filter algorithm is employed to smooth the decoupled temperature data, automatically filtering out random noise and electromagnetic interference signals from the cabinet's electrical equipment generated during the acquisition process, thus optimizing the smoothness and stability of the temperature time series curve. After the above three-stage processing, the module outputs a true cell temperature time series free from environmental disturbances and noise interference, which is stably transmitted to the temperature rise curvature fusion identification module, ensuring the accuracy of thermal anomaly identification from the data source.

[0023] The temperature rise curvature fusion identification module uses the sampling interval as a time reference to process the standardized cell temperature time series data after environmental compensation. Continuous numerical differentiation operations are performed. For the stable differentiation extraction of discrete temperature data, a 5-point central difference method combined with a moving average smoothing process with a width of 3 sampling points is used. The specific steps are as follows: temperature data from 5 consecutive sampling points... Through formula Calculate the first First-order temperature rise rate at point Then repeat the above 5-point central difference operation on the obtained first-order rate sequence to obtain the... Second-order curvature differential feature of a point This effectively suppresses noise amplification during the discrete data differentiation process.

[0024] The first-order rate of change of cell temperature rise was obtained through calculation. This parameter directly reflects the real-time temperature rise and fall rate of the battery cell; based on the first-order temperature rise rate data, a second-order differential operation is further performed to calculate the second-order curvature differential characteristic of the battery cell temperature rise. This curvature feature can accurately characterize the degree of abrupt change in the temperature rise trend of the battery cell, the inflection point of accelerated heat generation, and the fluctuation characteristics, and can effectively capture micro-heat generation anomalies that cannot be identified by traditional temperature thresholds.

[0025] The method for extracting the amplitude and duration of curvature abrupt changes is as follows: set a curvature reference threshold. (3 times the standard deviation of curvature fluctuation under normal operating conditions), when continuous sampling points Exceed When the curvature change is detected, it is determined to be the starting point of the curvature change; when three consecutive sampling points are detected... Falling back to The following values ​​are considered mutation termination points; mutation amplitude. Take the mutation interval The maximum value, duration of mutation This is the time difference between the end point and the start point.

[0026] With curvature change amplitude Duration of mutation To address these two core dimensions, a standard thermal anomaly feature library adapted to various types of battery cells was constructed. The library construction method involved selecting 100 target model battery cells and conducting accelerated aging tests under normal static conditions, varying degrees of internal short-circuit simulation, and different degrees of lithium plating. Temperature time-series data under each condition were collected, and curvature features were extracted. Cluster analysis was performed on all feature data, dividing them into three feature clusters: normal operating conditions, minor latent thermal anomalies, and overt thermal faults. The central coordinates of each feature cluster represent the baseline features for the corresponding operating condition. In the normal operating condition range, the curvature does not change abruptly or exceed the limit continuously; in the range of slight hidden thermal anomalies, the curvature fluctuates slightly and exceeds the limit for a short time, but there is no obvious absolute temperature rise exceeding the standard; in the range of obvious thermal faults, the curvature rises sharply and exceeds the limit continuously for a long time, accompanied by a rapid increase in temperature. At the same time, a variable-width time-series sliding window mechanism is adopted to track key parameters such as the instantaneous peak value of curvature features, the duration of continuous exceedance, and the fluctuation frequency in real time, and dynamically update the real-time feature dataset to ensure the real-time and continuous nature of feature tracking.

[0027] For different dimensions and The standardization process employs a max-min normalization method: the curvature amplitude is normalized to... The normalization formula for the interval is: The duration will be normalized to The normalization formula for the interval is: ,in , These are the minimum and maximum values ​​of curvature amplitude in the feature library. , For the minimum and maximum durations of all durations.

[0028] like Figure 2 As shown, the temperature rise curvature fusion identification module constructs a mapping classification model between curvature dual-dimensional features and cell thermal anomaly types. It adopts a dual-mode operation of offline batch training and online adaptive matching. In the offline stage, based on massive amounts of test samples of normal operating conditions, internal short circuits, and lithium plating thermal anomalies of cells of different specifications, feature annotation, model training, and weight optimization are completed. In the online operation stage, the weighted feature similarity calculation formula is called: Quantify the difference between real-time curvature features and baseline features in the feature library, among which , The weighting coefficients for curvature amplitude and duration are determined by cross-validation during offline training, with initial values ​​of 0.7 and 0.3 respectively, and are dynamically adjusted through incremental learning. , The normalized value of the real-time feature. , The normalized value of the baseline feature is used as the basis for classification decision logic. Perform intelligent discrimination. The preset similarity threshold is determined by taking half of the minimum Euclidean distance between the normal operating condition feature cluster and the micro-implicit thermal anomaly feature cluster, with a value range of 0.15~0.3. This can accurately distinguish between non-fault interferences such as normal temperature rise fluctuations, airflow disturbances in the duct, and load fluctuations, and micro-implicit thermal anomalies caused by internal short circuits and lithium plating. For implicit anomaly signals with small curvature deviations from the baseline, short durations, and no absolute temperature rise exceeding the standard, accurate identification is achieved through refined feature boundary judgment. At the same time, non-faulty micro-fluctuations are automatically filtered out, significantly reducing the system's false alarm rate and false negative rate.

[0029] The risk grading and interlocking module undertakes the core functions of risk level determination and interlocking signal output. It adopts a CPLD programmable hardware logic architecture to implement hard logic solidified judgment rules. Hard logic solidification refers to the logic flow of the three-level risk determination (the corresponding level is triggered only when both parameters simultaneously meet the threshold conditions), the signal output timing, and the linkage logic of the actuator are implemented through hardware circuits, with a response latency of less than 1ms and strong anti-interference capability. The curvature amplitude thresholds corresponding to each level of risk are... , Duration threshold , The data is stored in a non-volatile Flash memory connected to the CPLD and can be dynamically read and written to and updated through a closed-loop self-optimization module. The hardware judgment logic itself remains unchanged, enabling precise matching and rapid linkage between risk levels and protective actions. The specific implementation logic is as follows: The risk grading interlocking module incorporates a three-level risk grading standard. The threshold calibration method is based on the boundaries of three types of feature clusters in the feature library. The boundary between Level 1 low risk and Level 2 medium risk corresponds to the lower boundary of the micro-amplitude latent thermal anomaly feature cluster, i.e. This represents the minimum curvature amplitude of the micro-amplitude latent thermal anomaly characteristic cluster. Its minimum duration; the boundary between Level 2 medium risk and Level 3 high risk corresponds to the lower boundary of the explicit thermal fault feature cluster, i.e. The minimum curvature amplitude of the characteristic cluster of manifest thermal faults. Given its minimum duration, and satisfying , The hierarchical logic.

[0030] The risk grading interlocking module receives the curvature change amplitude output by the temperature rise curvature fusion identification module in real time. Duration Two-dimensional parameters, threshold comparison and level determination are performed point-by-point: Level 1 Low Risk: The curvature amplitude k is within the normal range of slight fluctuations, and the duration of the sudden change is t < t1. It is determined to be a slight temperature rise disturbance with no risk of thermal anomaly. The module outputs a PWM frequency conversion drive signal. Level 2 Medium Risk: Curvature amplitude k≥k1, and the duration of the sudden change t≥t1, which is judged as a local hidden thermal anomaly with potential thermal runaway risk. The module outputs a high-level electrical isolation drive signal. Level 3 High Risk: Curvature amplitude k≥k2, and sudden change duration t≥t2, judged as an obvious thermal runaway initiation fault, the risk spreads rapidly, the module synchronously outputs three independent linkage drive signals for fire start, cabinet interlock, and electrical isolation, all control signals are transmitted after electrical isolation processing, strong anti-interference ability, and no signal transmission delay.

[0031] The closed-loop self-optimization module is used to solve the problem of system recognition accuracy decay caused by cell aging, environmental condition drift, and equipment wear and tear. It enables lifelong online adaptive iteration of system parameters and recognition model, maintaining high recognition accuracy for a long time without manual calibration. The specific implementation method is as follows: The system collects feedback data in real time during operation, including actuator status feedback signals, thermal anomaly identification results and type labeling, cell operating status data throughout the entire time period, and complete time-series records of various abnormal events. Simultaneously, it stores environmentally compensated cell temperature time-series data and curvature dual-dimensional feature parameters to construct a localized operating database. The labeling method for real thermal anomaly samples is as follows: for events triggering level 2 or higher risks, manual labeling is performed based on subsequent cell disassembly and testing results (internal short circuits are confirmed through internal resistance testing and X-ray scanning; lithium plating is confirmed through post-disassembly electrode observation and capacity testing); for normal operating data that does not trigger risks, it is automatically labeled as a normal sample.

[0032] The module uses a fixed operating cycle as the statistical unit to perform big data statistical analysis and feature mining on historical operating data. For the thermal anomaly feature library, it dynamically and adaptively adjusts the curvature thresholds k1 and k2 and the duration thresholds t1 and t2 corresponding to each level of risk, and automatically corrects the feature benchmark offset caused by cell cycle aging, ambient temperature drift and airflow fluctuation in the duct. For the feature mapping classification model, a lightweight incremental learning algorithm is adopted. The update trigger condition is: when the number of newly labeled samples reaches 50 or there are no new abnormal samples for 30 consecutive days, the update is initiated. The update verification mechanism is: the new samples are divided into training set and validation set in a 7:3 ratio. The recognition accuracy of the updated model on the validation set must be ≥95%, otherwise the original model parameters are retained. The update constraints are: the threshold adjustment range of a single update does not exceed 10% of the original threshold, and the weight coefficient adjustment range does not exceed 5% of the original weight. There is no need to retrain the entire model. Only the newly added internal short circuit, lithium plating heat abnormality test samples and normal operation condition samples are included in the model iteration in real time. The baseline parameters, classification decision threshold and internal discrimination weights in the feature similarity calculation formula are updated online. The entire optimization process is executed silently in the background of the system, without affecting the real-time monitoring and protection functions of the front end. It realizes the closed-loop linkage of feature library parameter self-calibration and online iteration of recognition model, ensuring the stability and accuracy of the system in the long-term operation.

[0033] The actuator module includes a zoned directional preheating unit, a battery cell branch electrical isolation unit, a targeted fire suppression unit, and a full cabinet interlocking unit. The specific implementation method is as follows: Zoned directional pre-heating unit: It adopts a variable frequency airflow drive structure, receives a first-level risk PWM variable frequency drive signal, automatically adjusts the operating power according to the risk level, performs directional airflow circulation heat dissipation for the zone where the thermally abnormal battery cell is located, quickly equalizes the temperature inside the cabinet, suppresses the expansion of slight temperature rise disturbances, and achieves preventive protection. Cell branch electrical isolation unit: It adopts a contactless electrical isolation structure, receives isolation drive signals of level 2 and level 3 risks, quickly disconnects the electrical connection of the corresponding cell branch, completely cuts off the cell energy storage power supply circuit, blocks the energy supply of cells with thermal abnormalities, prevents thermal abnormalities from spreading to adjacent cells, and avoids electrical cascading failures; Targeted fire suppression unit: It adopts a fixed-point release fire suppression structure, responds only to the third-level high-risk linkage signal, and accurately releases the fire suppression medium to the thermally abnormal battery cell area to achieve targeted fire suppression and temperature suppression, quickly terminate the thermal runaway reaction, and minimize the risk of fire spread. Full cabinet locking unit: Triggered synchronously with the targeted fire suppression unit, it executes a fully enclosed electromagnetic locking of the cabinet, isolating the external oxygen supply, and simultaneously physically sealing the cabinet door to prevent accidental contact or misoperation that could lead to personal safety accidents, thus constructing a comprehensive physical safety protection system.

[0034] To verify the technical effect of the present invention, the following comparative experiment was conducted: 200 18650 ternary lithium batteries were selected and randomly divided into an experimental group and a control group, with 100 batteries in each group. The experimental group used the system of the present invention, with a sampling frequency of 10Hz and an initial threshold. ℃ / s 2 , , ℃ / s 2 , The control group used a traditional fixed temperature threshold (55℃ alarm) and temperature rise rate threshold (1℃ / min alarm) system. Different degrees of internal short circuit were simulated using the needle penetration method, and lithium plating anomalies were simulated using low-temperature charging. A total of 50 internal short circuit simulation tests and 50 lithium plating simulation tests were conducted. The results showed that the experimental group had a 96% accuracy rate in identifying minor, latent thermal anomalies, a 2% false alarm rate, and a 2% missed alarm rate; the control group had a 32% accuracy rate in identifying minor, latent thermal anomalies, an 18% false alarm rate, and a 50% missed alarm rate. After six months of continuous operation, the experimental group maintained an accuracy rate above 94%, while the control group's accuracy rate decreased to 25%. like Figure 3 As shown, the specific workflow of the integrated safety protection system for lithium-ion battery storage cabinets provided by this invention is as follows: The multi-source high-frequency acquisition module is triggered synchronously by a global clock, and collects all data in parallel, including the surface temperature of a single cell, the temperature of the tab, the ambient temperature inside the cabinet, and the airflow status. It generates digital time-series data with timestamps and transmits it to the computing unit to complete the standardized acquisition of raw data.

[0035] The environmental background compensation module calls the temperature disturbance compensation algorithm to quantitatively calculate the amount of environmental interference, complete the interference decoupling and deduction, and then uses an adaptive Kalman filter to filter out noise and electromagnetic interference, restore the cell's disturbance-free true temperature time sequence, and eliminate the interference of environmental factors on thermal status monitoring.

[0036] The temperature rise curvature fusion identification module performs first- and second-order differential operations on the corrected temperature data to extract temperature rise rate and curvature features; relying on a two-dimensional feature library and mapping classification model, it quantifies feature similarity and completes intelligent discrimination, accurately distinguishing between normal operating conditions, minor hidden thermal anomalies and obvious thermal faults.

[0037] The risk-level interlocking module performs a three-level risk threshold comparison based on two parameters: curvature amplitude and duration. After determining the current risk level, it outputs corresponding level interlocking control signals such as PWM drive, electrical isolation, and fire linkage in real time, achieving precise matching between risk and command.

[0038] The actuator module receives interlock control signals and triggers the corresponding functional unit actions: Level 1 risk starts zone preheating, Level 2 risk executes electrical isolation of battery cell branch, Level 3 risk synchronously starts fire extinguishing, and the entire cabinet is locked and electrically isolated, completing graded and targeted safety handling.

[0039] The closed-loop self-optimization module collects feedback data from the entire system operation process and builds a local database. Through data statistical analysis, it dynamically calibrates the threshold of the feature library and uses a lightweight incremental learning algorithm to iteratively update the recognition model online, thereby achieving adaptive optimization of system parameters and algorithm model and completing the full closed-loop operation.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An integrated safety protection system for lithium-ion battery storage cabinets, characterized in that, The system includes: a multi-source high-frequency acquisition module, an environmental background compensation module, a temperature rise curvature fusion identification module, a risk classification interlocking module, a closed-loop self-optimization module, and an actuator module; The multi-source high-frequency acquisition module synchronously acquires time-series data of single cell surface temperature and tab temperature, as well as background data of cabinet internal ambient temperature and airflow status. The environmental background compensation module uses an environmental temperature disturbance compensation algorithm to eliminate the interference of airflow disturbance in the cabinet and external environmental temperature changes on the battery cell temperature acquisition data, and outputs a corrected, disturbance-free temperature time series. The temperature rise curvature fusion identification module performs differential operations on the corrected temperature time series, solves the first-order change rate of cell temperature rise and second-order curvature differential features, constructs a two-dimensional feature library of curvature mutation amplitude and duration, and establishes a mapping model between curvature features and the initial state of short circuit and lithium plating thermal runaway in the cell, so as to realize the identification of micro-amplitude hidden thermal anomalies. The risk classification interlocking module matches the corresponding risk level according to the curvature characteristics, generates and outputs a classification interlocking control signal; The closed-loop self-optimization module realizes self-calibration of feature library parameters and online iterative update of the mapping model based on operational feedback data; The actuator module receives hierarchical interlocking control signals and executes corresponding safety protection actions; The environmental background compensation module uses an environmental temperature disturbance compensation algorithm. Calculate the temperature measurement interference caused by airflow convection heat transfer within the cabinet and fluctuations in ambient temperature. It is the total temperature measurement interference caused by the combined effect of airflow disturbance in the cabinet and temperature fluctuation in the external environment. It is used to characterize the overall deviation error of environmental factors on the actual temperature acquisition value of the battery cell. The heat transfer coefficient of the cell surface airflow is obtained by fitting the real-time airflow velocity inside the cabinet. The fitting relationship is pre-calibrated through wind tunnel testing. It is the total effective heat exchange area between the cell casing and the two tabs, which is the calibrated value for different specifications of cells; This is the weighted average of the ambient temperatures at multiple nodes inside the cabinet. This is the original temperature value collected by the battery cell; The system sampling interval; The equivalent thermal capacity of a single battery cell is obtained through calibration using a calorimeter. This is the environmental temperature change compensation coefficient; It is obtained by performing real-time first-order differential calculations on the time-series data of ambient temperature.

2. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The temperature rise curvature fusion identification module performs continuous real-time numerical differentiation operations on the actual temperature time series of the battery cell after environmental background compensation, using the sampling interval as the time base. By solving the first-order differential of the standardized temperature time series data T(t), the first-order rate of change of the battery cell temperature rise is obtained. This reflects the rate of change of cell temperature over time. Based on this temperature rise rate, a second-order differential calculation is performed to solve for the second-order curvature differential characteristic of the cell temperature rise. It is used to characterize the degree of abrupt change and the curvature of the temperature rise trend.

3. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The temperature rise curvature fusion identification module constructs a two-dimensional feature library with curvature mutation amplitude and mutation duration as the core. Combining the heat generation patterns of the initial stage of thermal runaway such as short circuit and lithium plating in the cell, it presets three types of feature threshold ranges in the feature library: normal operating conditions, micro-amplitude latent thermal anomalies, and manifest thermal faults. Among them, latent thermal anomalies correspond to curvature fluctuation ranges with small amplitude and short duration, while manifest thermal faults correspond to curvature over-limit ranges with large amplitude and long duration. The module tracks the instantaneous peak value and continuous over-limit duration of curvature features in real time through a time-series sliding window, and matches the real-time calculated curvature features with the feature library ranges to distinguish the curvature feature morphology corresponding to different thermal anomalies.

4. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 3, characterized in that, The temperature rise curvature fusion identification module, after calculating the cell temperature rise rate and second-order curvature features, uses the curvature mutation amplitude k and duration t as dual-dimensional features as model input. Based on a large amount of experimental and measured data on the initial stages of thermal runaway such as short circuits and lithium plating within the cell, a feature mapping classification model is constructed. This model is established using a method of offline training followed by online adaptive matching. The distribution patterns of curvature features corresponding to early thermal anomalies of the cell under different operating conditions are labeled and learned to form a correspondence between features and fault types. During online operation, the difference between real-time features and preset benchmark features in the model is quantified through a feature similarity calculation formula, and classification decision logic is used. Compare and distinguish. To pre-determine the similarity threshold, the system distinguishes between normal temperature rise fluctuations, normal operating condition disturbances and internal short circuits, and subtle latent thermal anomalies caused by lithium plating. For latent anomaly signals with curvature that deviates slightly from the reference range, have a short duration, and do not show obvious temperature exceedances, the system achieves effective identification through refined feature boundary determination. At the same time, it filters out non-faulty minor fluctuation interference to avoid misjudgment.

5. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 3, characterized in that, The temperature rise curvature fusion identification module quantifies the difference between real-time features and preset benchmark features within the model using a feature similarity calculation formula, which is: Where d is the similarity between the currently calculated curvature abrupt change amplitude and the curvature baseline value of the corresponding thermal anomaly type in the feature library. , The weighting coefficients for curvature amplitude and duration are determined through cross-validation during offline training. , The normalized value of the real-time feature. , The normalized value is the baseline feature.

6. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The risk grading interlocking module receives the curvature change amplitude k and duration t from the temperature rise curvature fusion identification module in real time, solidifies the three-level risk threshold determination logic, and presets the curvature amplitude threshold and duration threshold corresponding to each level. Specifically, the first level low risk corresponds to k being in the range of slight fluctuation and t < t1, where t1 is the preset short-term threshold; the second level medium risk corresponds to k ≥ k1 and t ≥ t1, where k1 is the curvature steep increase threshold; and the third level high risk corresponds to k ≥ k2 and t ≥ t2, where k2 is the curvature extreme value threshold and t2 is the preset long-term threshold.

7. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The risk classification interlocking module compares the received curvature features with preset thresholds in real time to determine the current risk level, and then outputs corresponding classification interlocking control signals. The first-level risk outputs a PWM frequency conversion drive signal to drive the partitioned directional pre-heating unit to operate; the second-level risk outputs a high-level electrical isolation drive signal to control the cell branch electrical isolation unit to disconnect. The three-level risk synchronous output of fire start, cabinet interlock, and electrical isolation three-way linkage drive signals ensures coordinated action of all execution units.

8. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The closed-loop self-optimization module collects real-time feedback data from the entire system operation process, including action feedback signals from the actuators, thermal anomaly identification results from the temperature rise curvature fusion identification module, real-time cell operating status data, and complete records of various abnormal events. Simultaneously, it stores temperature time-series data and curvature feature parameters after environmental background compensation. It performs statistical analysis on historical operating data within a certain period, dynamically adjusting the curvature mutation amplitude thresholds k1 and k2 and duration thresholds t1 and t2 corresponding to each level of risk in the feature library. This corrects feature offsets caused by cell aging, environmental temperature changes, and airflow fluctuations. For the mapping model, a lightweight incremental learning algorithm is used to incorporate newly added internal short-circuit, lithium plating thermal anomaly samples, and normal operation samples into the model training in real time. This updates the baseline parameters and classification decision thresholds in the feature similarity calculation formula, optimizes the internal discrimination weights of the model, and achieves closed-loop linkage between feature library parameter self-calibration and online iteration of the mapping model.

9. The integrated safety protection system for a lithium-ion battery storage cabinet according to claim 1, characterized in that, The actuator module includes a zoned directional preheating unit, a cell branch electrical isolation unit, a targeted fire suppression unit, and a full cabinet interlocking unit; The zoned directional preheating unit is a variable frequency axial flow fan, the battery cell branch electrical isolation unit is a high-voltage solid-state relay, the targeted fire suppression unit is an ultra-fine aerosol fire extinguishing device, and the cabinet interlocking unit is an electromagnetic safety lock.

Images