Distributed pressure monitoring and early warning method and system in battery box

Abstract

The invention discloses a distributed pressure monitoring and early warning method and system in a battery box, and the method comprises the steps: obtaining pressure data, collected by a distributed pressure sensor array, of each point in the battery box, and forming pressure distribution data; acquiring temperature and humidity in the battery box, and compensating the pressure distribution data by adopting an environment compensation algorithm; based on the compensated pressure distribution data, dynamically calculating a pressure normal range threshold value by adopting a 3 sigma principle so as to judge whether the pressure is abnormal or not; if the pressure is abnormal, determining an abnormal position by adopting a weighted centroid positioning algorithm, and identifying a fault type of the pressure abnormality; and triggering graded early warning according to the abnormal position and the fault type. According to the invention, the distributed pressure sensor array is combined with the weighted centroid positioning algorithm, so that the accurate positioning of the abnormal pressure position in the battery box is realized, and the problem that the traditional monitoring only knows the abnormal position but does not know the abnormal position is solved.

Description

Technical Field

[0001] This invention relates to the field of pressure monitoring technology, and in particular to a distributed pressure monitoring and early warning method and system for battery boxes. Background Technology

[0002] With the rapid development of energy storage systems and other fields, batteries, as core energy components, directly determine the reliability of end products and user trust in their safety performance. Under long-term cyclic use, high-temperature storage, or abnormal operating conditions, batteries are prone to problems such as internal gas generation, cell expansion, and separator damage, which can lead to abnormal pressure. If not detected and dealt with in time, these issues may quickly escalate into major safety accidents such as thermal runaway, fire, and explosion.

[0003] Existing battery box pressure monitoring technologies suffer from the problem of ambiguous anomaly localization. Traditional monitoring methods often employ single-point or a small number of discrete sensors, making it difficult to monitor the pressure distribution characteristics within the battery box and accurately pinpoint the specific location of pressure anomalies. This can easily lead to situations where an anomaly is known, but its location is unknown. Summary of the Invention

[0004] To address the problem that existing technologies cannot accurately pinpoint the exact location of pressure anomalies, resulting in a lack of targeted early warning information, this invention provides a distributed pressure monitoring and early warning method and system for battery boxes. This method utilizes a distributed pressure sensor array combined with a weighted centroid positioning algorithm to achieve precise location of pressure anomalies within the battery box, solving the problem of traditional monitoring only detecting the anomaly but not its location. The specific technical solution is as follows: This invention provides a distributed pressure monitoring and early warning method for battery boxes, comprising the following steps: The pressure data of various points inside the battery box is collected by a distributed pressure sensor array to form pressure distribution data. The temperature and humidity inside the battery box are obtained, and an environmental compensation algorithm is used to compensate for the pressure distribution data. Based on the compensated pressure distribution data, the 3σ principle is used to dynamically calculate the normal pressure range threshold to determine whether the pressure is abnormal. If the pressure is abnormal, the weighted centroid localization algorithm is used to determine the location of the abnormality and identify the fault type of the pressure abnormality. A tiered warning is triggered based on the location and type of the anomaly.

[0005] Preferably, the step of dynamically calculating the normal pressure range threshold based on the compensated pressure distribution data using the 3σ principle to determine whether the pressure is abnormal includes: The current normal pressure range threshold is dynamically calculated for each sensor point using the 3σ principle, specifically by calculating the mean and standard deviation of the historical data for that point. If the pressure value at each point in the compensated pressure distribution data is greater than the corresponding normal pressure range threshold, then that point is determined to be a pressure anomaly point.

[0006] Preferably, the step of determining the location of the abnormality by using a weighted centroid localization algorithm if the pressure is abnormal includes: If the pressure anomaly is a single point, an analysis area centered on that point is defined, and the analysis area includes the pressure anomaly point and all its adjacent sensor points. Calculate the pressure gradient of adjacent sensors within the analysis area, and determine the anomaly propagation path based on the pressure gradient; Based on the abnormal propagation path, the centroid coordinates of the abnormal region are calculated using a weighted centroid localization algorithm, and these centroid coordinates are used as the location coordinates of the pressure anomaly source.

[0007] Preferably, a distributed pressure monitoring and early warning method for battery boxes further includes: If the pressure anomaly consists of multiple anomaly points, the DBSCAN clustering algorithm is used to perform cluster analysis based on the spatial location of the multiple anomaly points and the pressure anomaly amplitude to divide different fault areas. For each fault region identified by clustering, a weighted centroid localization algorithm is executed to calculate the location coordinates of each pressure anomaly source.

[0008] Preferably, the location coordinates of the pressure anomaly source are two-dimensional or three-dimensional coordinates.

[0009] Preferably, if the pressure is abnormal, the fault type identified includes: The compensated pressure distribution data is input into the trained pressure pattern recognition model, and the corresponding fault type identifier is output; the fault type includes at least one of short circuit, gas generation, and mechanical impact.

[0010] Preferably, the trained stress pattern recognition model is a hybrid model based on CNN-LSTM, and the construction process includes: Acquire historical data, which includes a sequence of pressure distribution images of the battery under various known fault conditions; Features are extracted from the pressure distribution image sequence, including spatial features, temporal features, and pressure transmission features between sensors; The extracted features and corresponding fault labels are used to train a CNN-LSTM hybrid model so that the model can learn the pressure distribution pattern fingerprint of different fault types.

[0011] Preferably, in the CNN-LSTM hybrid model, CNN is used to extract the spatial features of pressure distribution, and LSTM is used to learn the temporal features of pressure evolution; the pressure transmission features are obtained by calculating the propagation speed and direction of pressure peaks between sensor nodes.

[0012] Preferably, triggering a graded early warning system matching the fault type based on the fault location includes: When the fault type is not identified, but the location of the anomaly is determined: If the abnormal location is a single abnormal point, and the pressure value at that point exceeds the first threshold but does not exceed the second threshold, a level one warning is triggered; If the abnormal location result shows multiple abnormal points, or if the pressure value of a single abnormal point exceeds the second threshold, a level two warning will be triggered. Wherein, the first threshold is less than the second threshold; When the fault type is identified and the location of the anomaly is determined: When a fault type is identified, regardless of the number of abnormal points, a level three warning corresponding to the fault type will be triggered.

[0013] This invention also provides a distributed pressure monitoring and early warning system for a battery box, applied to the aforementioned distributed pressure monitoring and early warning method for a battery box, comprising: The data acquisition unit is used to acquire pressure data from various points inside the battery box collected by the distributed pressure sensor array, forming pressure distribution data. An environmental compensation unit is used to acquire the temperature and humidity inside the battery box and to compensate the pressure distribution data using an environmental compensation algorithm. Anomaly detection unit is used to dynamically calculate the normal pressure range threshold based on the compensated pressure distribution data and the 3σ principle to determine whether the pressure is abnormal. The anomaly analysis unit is used to determine the location of the anomaly and identify the fault type of the pressure anomaly when an anomaly occurs, using a weighted centroid localization algorithm. The early warning unit is used to trigger graded early warnings based on the abnormal location and fault type.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention discloses a distributed pressure monitoring and early warning method and system for battery boxes. By combining a distributed pressure sensor array with a weighted centroid positioning algorithm, the method achieves accurate location of abnormal pressure within the battery box, solving the problem that traditional monitoring only knows the abnormality but not where it is, thus shortening the response time. At the same time, by analyzing the pressure distribution data, the method can identify the fault type, distinguish different fault scenarios, avoid secondary damage caused by blind handling, and improve the efficiency of fault handling. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0016] Figure 1 This is a flowchart of a distributed pressure monitoring and early warning method for a battery box according to the present invention.

[0017] Figure 2 This is a schematic diagram of a distributed pressure monitoring and early warning system inside a battery box according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0019] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0020] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0022] Please refer to the following examples. Figure 1 and Figure 2 .

[0023] This invention provides a distributed pressure monitoring and early warning method for battery boxes, comprising the following steps: Step S1: Obtain pressure data from various points inside the battery box collected by the distributed pressure sensor array to form pressure distribution data; A distributed pressure sensor array deployed inside the battery compartment is activated. This array is laid out in a grid pattern, with sensor nodes distributed in key locations such as between individual battery cells, at the corners of the compartment, and near safety valves, covering the entire battery compartment. Each sensor is pre-assigned a unique ID number, and its two-dimensional (X,Y) / three-dimensional (X,Y,Z) coordinates within the battery compartment are recorded, establishing a sensor ID-physical coordinate mapping table for location calculations.

[0024] The sensor acquisition frequency is set, and raw analog pressure signals from all sensor nodes are acquired synchronously. These raw analog pressure signals are amplified, filtered, and converted into digital pressure values ​​using an analog-to-digital converter. The data is stored in the format of sensor ID, acquisition timestamp, and pressure value. Based on the physical space inside the battery box and the preset physical coordinates of each node, a dynamically updated two-dimensional pressure distribution matrix is ​​constructed. This matrix reflects the real-time pressure field inside the entire battery box, completing the pressure distribution data construction.

[0025] Step S2: Obtain the temperature and humidity inside the battery box, and use an environmental compensation algorithm to compensate for the pressure distribution data; Temperature and humidity sensors were installed at representative locations inside the battery box, on the upper, middle, and lower sides of the box. These sensors, along with the pressure sensors, collected data synchronously using the same timestamp, at the same frequency. The temperature and humidity sensors collected ambient temperature and relative humidity data, which were then stored in association with the corresponding pressure data at the specified timestamps.

[0026] Error curves of the pressure sensor under different temperature and humidity conditions were obtained through experiments in advance: In a constant temperature and humidity chamber, temperature gradients (-40℃, -20℃, 0℃, 25℃, 45℃, 65℃, 85℃) and humidity gradients (20%RH, 40%RH, 60%RH, 80%RH, 100%RH) were set, and the sensor's measured values ​​under standard pressure were collected to establish a three-dimensional database of temperature-humidity-pressure measurement errors.

[0027] A linear compensation model was trained based on a three-dimensional database of temperature, humidity, and pressure measurement errors. in, The pressure value after compensation. The pressure value measured by the sensor is T, the ambient temperature is RH, the ambient humidity is a, b, and c are compensation coefficients obtained through experimental calibration.

[0028] The compensation model is invoked to perform point-by-point compensation calculations on the pressure value of each sensor point in the pressure distribution data matrix obtained in step S1, combined with the temperature and humidity data at the corresponding timestamp. If the compensated pressure value exceeds the sensor's range, it is marked as invalid data, and the compensated data from the previous moment is used as a temporary replacement.

[0029] By establishing a separate historical database for each sensor point, the compensated pressure data for a specific time period is stored and sorted by timestamp. A data update window period is set, and historical data from the previous period is automatically removed and replaced with valid data from the latest period after each period.

[0030] Step S3: Based on the compensated pressure distribution data, the normal pressure range threshold is dynamically calculated using the 3σ principle to determine whether the pressure is abnormal. Preferably, the step of dynamically calculating the normal pressure range threshold based on the compensated pressure distribution data using the 3σ principle to determine whether the pressure is abnormal includes: If the pressure value at each point in the compensated pressure distribution data is greater than the corresponding normal pressure range threshold, then that point is determined to be a pressure anomaly point.

[0031] The current normal pressure range threshold is dynamically calculated for each sensor point using the 3σ principle, specifically by calculating the mean and standard deviation of the historical data for that point. For each sensor point, calculate the mean μ and standard deviation σ of the valid data in its historical database, such as calculating the first... The average of historical data from each sensor and standard deviation The formula is as follows: in, For database The Middle There are 10 valid data points, where n is the number of valid data points.

[0032] The normal pressure range threshold for this sensor point is set to [μ-3σ, μ+3σ], meaning that the normal pressure value must fall within the range of mean ± 3 times the standard deviation. If it exceeds this range, it is considered abnormal.

[0033] Compare the pressure values ​​at each sensor point after compensation in step S2 with the corresponding threshold range: like or If the value is within the threshold range, the sensor point is marked as an abnormal point; if it is within the threshold range, it is marked as a normal point.

[0034] Summarize the IDs, coordinates, and pressure values ​​of all anomalies to form an anomaly list. If the list is empty, return to step S1 to continue data collection; if the list is not empty, proceed to step S4.

[0035] Specifically, by setting a threshold update frequency, the mean μ and standard deviation σ of the historical data of each sensor point are continuously updated and the normal range threshold is updated to ensure that the threshold can adapt to the long-term drift of the sensor points or the gradual change of the environment.

[0036] It should be noted that after an initial determination of an anomaly, multiple consecutive verifications are triggered: for example, if the pressure value still exceeds the updated threshold range in at least 2 out of 3 verifications, then the point is finally confirmed as a valid anomaly; if it exceeds the threshold only once, it is determined to be a momentary interference point and is not included in the anomaly handling.

[0037] Step S4: If the pressure is abnormal, use the weighted centroid localization algorithm to determine the location of the abnormality and identify the fault type of the pressure abnormality. Receive the list of valid outliers output in step S3 and extract the set of valid outliers. , The number of valid outliers, each point Includes parameters Calculate the pressure deviation value: in, The pressure value at point i; In this embodiment, the location coordinates of the pressure anomaly source are either two-dimensional or three-dimensional. When the location coordinates of the pressure anomaly source are set to three-dimensional coordinates, a local coordinate system for the battery box is established. This system uses the lower left corner of the battery box as the origin, the horizontal direction as the X-axis, the vertical direction as the Y-axis, and the depth direction of the box as the Z-axis, to unify the coordinates of all anomaly points. The deviation value is then used as the reference. Weights are used for weighting, and weight normalization is performed. The anomaly location is calculated using a weighted centroid localization algorithm, as shown in the following formula: The weighted centroid coordinates of the pressure anomaly are obtained, i.e., the location of the anomaly core.

[0038] It should be noted that the weighted centroid localization algorithm in this embodiment is suitable for battery boxes with three-dimensional multi-layer structures, such as battery clusters in large energy storage power stations, covering three-dimensional space in horizontal / vertical / depth.

[0039] In other embodiments, when the location coordinates of the pressure anomaly source are two-dimensional coordinates, the weighted centroid localization algorithm formula is: in, , The distance from the sensor to the abnormal area; , Sensor coordinates; An environmental correction factor is used to avoid weight overflow issues when the sensor is too close to the source of pressure anomalies.

[0040] It should be noted that this weighted centroid positioning algorithm formula is applicable to small battery devices with a planar layout, such as portable energy storage power supplies. The weighted centroid positioning algorithm formula can be set according to the size of the energy storage power supply or the application scenario.

[0041] Fault types can be identified by establishing a fault type identification model or a pressure anomaly feature-fault type mapping library. For example, using a pressure anomaly feature-fault type mapping library for identification involves extracting features of the current anomaly point, including the number of anomaly points, distribution pattern, pressure change rate, and location correlation, and matching them with the mapping library. If the feature matching degree is greater than the matching threshold, the corresponding fault type is determined; if the matching degree is less than the matching threshold, the fault type is determined to be unidentified.

[0042] Step S5: Trigger a graded early warning based on the abnormal location and fault type.

[0043] Preferably, triggering a graded early warning system matching the fault type based on the fault location includes: When the fault type is not identified, but the location of the anomaly is determined: If the abnormal location is a single abnormal point, and the pressure value at that point exceeds the first threshold but does not exceed the second threshold, a level one warning is triggered; If the abnormal location result shows multiple abnormal points, or if the pressure value of a single abnormal point exceeds the second threshold, a level two warning will be triggered. Wherein, the first threshold is less than the second threshold; When a Level 1 warning is triggered, the local alarm emits a low-frequency flashing and intermittent beeping sound; it also uploads warning information to the monitoring center, including the coordinates of the abnormal point and pressure data.

[0044] When a Level 2 warning is triggered, the local audible and visual alarm is upgraded to a high-frequency, high-intensity light mode; a current-limiting command is sent to the Battery Management System (BMS) via the CAN bus to forcibly reduce the charging and discharging power.

[0045] When the fault type is identified and the location of the anomaly is determined: When a fault type is identified, regardless of the number of abnormal points, a level three warning corresponding to the fault type will be triggered.

[0046] In addition, when the pressure abnormally exceeds 80% of the safety valve opening pressure, causing a sharp rise in the pressure inside the battery box, a level three warning is triggered.

[0047] When a Level 3 warning is triggered, the system successfully identifies the specific fault type, immediately activates a relay to cut off the battery main circuit, and sends a high-priority emergency alarm to maintenance personnel via a remote communication module (such as 4G / 5G), along with the fault type and abnormal location. Simultaneously, it automatically activates directional fire suppression systems targeting the abnormal area.

[0048] By implementing detailed, tiered early warning rules, the system links early warning levels with risk severity. This avoids the limitations of single-dimensional early warnings while enabling operations and maintenance personnel to quickly assess risk levels, thus balancing security protection with operational efficiency.

[0049] The present invention discloses a distributed pressure monitoring and early warning method and system for battery boxes. By combining a distributed pressure sensor array with a weighted centroid positioning algorithm, the method achieves accurate location of abnormal pressure within the battery box, solving the problem that traditional monitoring only knows the abnormality but not where it is, thus shortening the response time. At the same time, by analyzing the pressure distribution data, the method can identify the fault type, distinguish different fault scenarios, avoid secondary damage caused by blind handling, and improve the efficiency of fault handling.

[0050] Specifically, in a preferred embodiment of this application, the step of determining the abnormal location using a weighted centroid localization algorithm if the pressure is abnormal includes: If the pressure anomaly is a single point, an analysis area centered on that point is defined, and the analysis area includes the pressure anomaly point and all its adjacent sensor points. Calculate the pressure gradient of adjacent sensors within the analysis area, and determine the anomaly propagation path based on the pressure gradient; Based on the abnormal propagation path, the centroid coordinates of the abnormal region are calculated using a weighted centroid localization algorithm, and these centroid coordinates are used as the location coordinates of the pressure anomaly source.

[0051] Specifically, a distributed pressure monitoring and early warning method for battery boxes also includes: If the pressure anomaly consists of multiple anomaly points, the DBSCAN clustering algorithm is used to perform cluster analysis based on the spatial location of the multiple anomaly points and the pressure anomaly amplitude to divide different fault areas. For each fault region identified by clustering, a weighted centroid localization algorithm is executed to calculate the location coordinates of each pressure anomaly source.

[0052] In this embodiment, taking the planar cell layout of a portable energy storage power supply as an example, the specific implementation steps are as follows: When a single pressure anomaly point is identified, denoted as point A, its coordinates are ( , ), define adjacent sensor points as the direct adjacent points of point A in the grid in the four directions of up, down, left and right, and draw a 3×3 local analysis region centered on point A and including point A and the four adjacent points.

[0053] Calculate the pressure gradient between adjacent points within the analysis region, denoted as B / C / D / E and point A: Where d is the sensor spacing, The pressure value at adjacent point j. The pressure value at anomaly point A is due to... Because the pressure at the anomaly source is higher than that of the surrounding area, if... <0, that is < If the pressure gradient direction is from point A to point j, then this direction is the abnormal propagation path.

[0054] Calculate the weights: Calculate the centroid coordinates of the pressure anomaly source ( ,, ): Final coordinates ( ,, The coordinates of the abnormal pressure source are shown below.

[0055] In the process of determining the location of anomalies when there are multiple anomalies, clustering parameters such as neighborhood radius and minimum number of points are set. Points that satisfy the condition that the spatial distance is not greater than the neighborhood radius and the pressure anomaly amplitude ΔP is greater than the anomaly threshold are classified into the same fault area, resulting in multiple independent clusters.

[0056] For each fault region obtained from clustering, repeat the steps for a single anomaly point to obtain the centroid coordinates of the pressure anomaly source corresponding to each fault region.

[0057] In this embodiment, a two-dimensional weighted centroid algorithm is adopted for the two-dimensional cell structure of portable energy storage power supplies. This simplifies the computation and conforms to the anomaly diffusion characteristics in planar space, matching the compact space requirements of small devices. The localization of individual anomalies is more accurate by combining the analysis region and pressure gradient, avoiding errors in single-point localization. Simultaneously, the weights correspond to distance, giving anomalies a higher weighting and making the localization results closer to the actual pressure anomaly source. Furthermore, the distinction between multiple pressure anomaly sources is clearer. DBSCAN clustering can divide multiple independent anomalies into different fault regions, solving the problem of location confusion when multiple cells malfunction simultaneously in small devices, facilitating targeted troubleshooting of each faulty cell by maintenance personnel.

[0058] Specifically, in a preferred embodiment of this application, if the pressure is abnormal, identifying the fault type of the abnormal pressure includes: The compensated pressure distribution data is input into the trained pressure pattern recognition model, and the corresponding fault type identifier is output; the fault type includes at least one of short circuit, gas generation, and mechanical impact.

[0059] The trained stress pattern recognition model is a hybrid model based on CNN-LSTM, and its construction process includes: Acquire historical data, which includes a sequence of pressure distribution images of the battery under various known fault conditions; Features are extracted from the pressure distribution image sequence, including spatial features, temporal features, and pressure transmission features between sensors; Specifically, in the CNN-LSTM hybrid model, CNN is used to extract the spatial features of pressure distribution, and LSTM is used to learn the temporal features of pressure evolution; the pressure transmission features are obtained by calculating the propagation speed and direction of pressure peaks between sensor nodes.

[0060] The spatial feature extraction CNN module uses a lightweight CNN structure with two layers of convolution and pooling: The first layer consists of 32 3×3 convolutional kernels with the ReLU activation function to extract shape features from high-pressure regions; the second layer consists of 64 3×3 convolutional kernels followed by a 2×2 max pooling layer to extract density features; the final output is a 128-dimensional spatial feature vector for each frame.

[0061] The temporal feature extraction LSTM module inputs the spatial feature vectors of each frame output by the CNN into a two-layer LSTM with 64 hidden units in chronological order, learning the stress evolution patterns of different faults. Short circuits are rapid time-series fluctuations that rise and fall sharply; Gas production exhibits a slow, continuous, and gradual upward trend over a period of time. Mechanical shocks are characterized by a pulse-like timing pattern, from an instantaneous peak to a rapid decline.

[0062] The final output is a 64-dimensional temporal feature vector.

[0063] By extracting pressure transmission features and identifying the occurrence time of pressure peaks from each sensor in each image sequence, we can determine the optimal timing for each sensor. Calculate the peak time difference between adjacent sensors. ,get: speed of propagation: , The spacing between adjacent sensors; Direction of dissemination: From The sensor points The sensor quantizes velocity and direction into 32-dimensional transmitted feature vectors, which are then concatenated with the output features of CNN and LSTM to obtain 224-dimensional fused features.

[0064] The extracted features and corresponding fault labels are used to train a CNN-LSTM hybrid model so that the model can learn the pressure distribution pattern fingerprint of different fault types.

[0065] The compensated pressure distribution data is converted into a continuous frame pressure image sequence according to the sampling frequency; the image sequence is input into a trained CNN-LSTM hybrid model, and the model automatically extracts spatial, temporal, and transmission features; the model outputs fault type identifiers such as short circuit / gas generation / mechanical impact / no fault, thus completing the fault type identification of pressure anomalies.

[0066] The model can accurately identify different fault types such as short circuits, gas generation, and mechanical shock. For example, for short circuit faults, the model learns its rapid temporal fluctuation characteristics from sudden rises to falls, as well as spatial features such as the shape and density of high-pressure areas, enabling it to distinguish short circuits from other faults and avoid invalid or erroneous warnings due to misjudgment of fault types, thus effectively reducing false alarm and missed alarm rates. For example, in complex battery operating environments, other disturbances may cause abnormal pressure fluctuations, but the model can combine multi-dimensional information such as temporal and spatial characteristics to accurately determine whether it is a true fault, reducing false alarms caused by disturbances. Furthermore, different fault types and abnormal locations may lead to varying degrees of safety risks. By accurately identifying fault types and abnormal locations, different faults can be graded and warned according to pre-set rules. Utilizing the model's learning of the temporal characteristics of pressure evolution, some potential fault signs can be detected in advance. For example, for gas generation faults, the model can identify its slow and continuous upward trend, issuing warnings before the pressure reaches the danger threshold, reminding relevant personnel to take preventative measures, and improving the timeliness of warnings.

[0067] This invention also provides a distributed pressure monitoring and early warning system for a battery box, applied to the aforementioned distributed pressure monitoring and early warning method for a battery box, comprising: The data acquisition unit is used to acquire pressure data from various points inside the battery box collected by the distributed pressure sensor array, forming pressure distribution data. An environmental compensation unit is used to acquire the temperature and humidity inside the battery box and to compensate the pressure distribution data using an environmental compensation algorithm. Anomaly detection unit is used to dynamically calculate the normal pressure range threshold based on the compensated pressure distribution data and the 3σ principle to determine whether the pressure is abnormal. The anomaly analysis unit is used to determine the location of the anomaly and identify the fault type of the pressure anomaly when an anomaly occurs, using a weighted centroid localization algorithm. The early warning unit is used to trigger graded early warnings based on the abnormal location and fault type.

[0068] The functional explanation of each unit in this embodiment is the same as that of a distributed pressure monitoring and early warning method in a battery box, and the technical effect is the same, so it will not be repeated here.

[0069] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0070] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0071] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0072] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the specification of the present invention.

Claims

1. A distributed pressure monitoring and early warning method for a battery box, characterized in that, Includes the following steps: The pressure data of various points inside the battery box is collected by a distributed pressure sensor array to form pressure distribution data. The temperature and humidity inside the battery box are obtained, and an environmental compensation algorithm is used to compensate for the pressure distribution data. Based on the compensated pressure distribution data, the 3σ principle is used to dynamically calculate the normal pressure range threshold to determine whether the pressure is abnormal. If the pressure is abnormal, the weighted centroid localization algorithm is used to determine the location of the abnormality and identify the fault type of the pressure abnormality. A tiered warning is triggered based on the location and type of the anomaly.

2. The distributed pressure monitoring and early warning method for a battery box according to claim 1, characterized in that, The method of dynamically calculating the normal pressure range threshold based on the compensated pressure distribution data to determine whether the pressure is abnormal includes: The current normal pressure range threshold is dynamically calculated for each sensor point using the 3σ principle, specifically by calculating the mean and standard deviation of the historical data for that point. If the pressure value at each point in the compensated pressure distribution data is greater than the corresponding normal pressure range threshold, then that point is determined to be a pressure anomaly point.

3. The distributed pressure monitoring and early warning method for a battery box according to claim 1, characterized in that, If the pressure is abnormal, the method of determining the location of the abnormality using a weighted centroid localization algorithm includes: If the pressure anomaly is a single point, an analysis area centered on that point is defined, and the analysis area includes the pressure anomaly point and all its adjacent sensor points. Calculate the pressure gradient of adjacent sensors within the analysis area, and determine the anomaly propagation path based on the pressure gradient; Based on the abnormal propagation path, the centroid coordinates of the abnormal region are calculated using a weighted centroid localization algorithm, and these centroid coordinates are used as the location coordinates of the pressure anomaly source.

4. The distributed pressure monitoring and early warning method for a battery box according to claim 3, characterized in that, Also includes: If the pressure anomaly consists of multiple anomaly points, the DBSCAN clustering algorithm is used to perform cluster analysis based on the spatial location of the multiple anomaly points and the pressure anomaly amplitude to divide different fault areas. For each fault region identified by clustering, a weighted centroid localization algorithm is executed to calculate the location coordinates of each pressure anomaly source.

5. A distributed pressure monitoring and early warning method for a battery box according to claim 3 or 4, characterized in that, The location coordinates of the pressure anomaly source are two-dimensional or three-dimensional coordinates.

6. The distributed pressure monitoring and early warning method for a battery box according to claim 4, characterized in that, If the pressure is abnormal, the fault types identified include: The compensated pressure distribution data is input into the trained pressure pattern recognition model, and the corresponding fault type identifier is output; the fault type includes at least one of short circuit, gas generation, and mechanical impact.

7. The distributed pressure monitoring and early warning method for a battery box according to claim 6, characterized in that, The trained stress pattern recognition model is a hybrid model based on CNN-LSTM, and its construction process includes: Acquire historical data, which includes a sequence of pressure distribution images of the battery under various known fault conditions; Features are extracted from the pressure distribution image sequence, including spatial features, temporal features, and pressure transmission features between sensors; The extracted features and corresponding fault labels are used to train a CNN-LSTM hybrid model so that the model can learn the pressure distribution pattern fingerprint of different fault types.

8. The distributed pressure monitoring and early warning method for a battery box according to claim 7, characterized in that, In the CNN-LSTM hybrid model, CNN is used to extract the spatial features of pressure distribution, and LSTM is used to learn the temporal features of pressure evolution; the pressure transmission features are obtained by calculating the propagation speed and direction of pressure peaks between sensor nodes.

9. A distributed pressure monitoring and early warning method for a battery box according to claim 6, characterized in that, The step of triggering a tiered early warning system that matches the fault type based on the location of the anomaly includes: When the fault type is not identified, but the location of the anomaly is determined: If the abnormal location is a single abnormal point, and the pressure value at that point exceeds the first threshold but does not exceed the second threshold, a level one warning is triggered; If the abnormal location result shows multiple abnormal points, or if the pressure value of a single abnormal point exceeds the second threshold, a level two warning will be triggered. Wherein, the first threshold is less than the second threshold; When the fault type is identified and the location of the anomaly is determined: When a fault type is identified, regardless of the number of abnormal points, a level three warning corresponding to the fault type will be triggered.

10. A distributed pressure monitoring and early warning system for a battery box, characterized in that, A distributed pressure monitoring and early warning method for a battery box according to any one of claims 1 to 9, comprising: The data acquisition unit is used to acquire pressure data from various points inside the battery box collected by the distributed pressure sensor array, forming pressure distribution data. An environmental compensation unit is used to acquire the temperature and humidity inside the battery box and to compensate the pressure distribution data using an environmental compensation algorithm. Anomaly detection unit is used to dynamically calculate the normal pressure range threshold based on the compensated pressure distribution data and the 3σ principle to determine whether the pressure is abnormal. The anomaly analysis unit is used to determine the location of the anomaly and identify the fault type of the pressure anomaly when an anomaly occurs, using a weighted centroid localization algorithm. The early warning unit is used to trigger graded early warnings based on the abnormal location and fault type.

Images