A battery pack fire early warning method and system based on multi-source parameters

Abstract

The present application relates to the technical field of fire warning, in particular to a battery pack fire warning method and system based on multi-source parameters, comprising: comparing the change amount of multi-source monitoring parameters of the same multi-source parameter category at the same time before, and combining the distribution distance between different battery pack partitions to calculate the multi-source coupling risk value; when the same battery pack partition transits to other battery pack partitions, analyzing the change gradient direction of multi-source monitoring parameters in different multi-source parameter categories to calculate the partition correlation influence degree; comparing the change amount of numerical value produced when different multi-source parameter categories are subjected to time sequence simulation prediction, weighting and fusing the cross-partition gradient abnormal value and the partition correlation influence degree, and then implementing fire warning response. The present application reduces the hysteresis of warning response and reduces the risk of false alarm or missed alarm.

Description

Technical Field

[0001] This invention relates to the field of fire early warning technology, specifically to a battery pack fire early warning method and system based on multi-source parameters. Background Technology

[0002] Lithium-ion battery thermal runaway is characterized by its suddenness, intensity, and difficulty in extinguishing. Traditional detectors often only sound an alarm after open flames or dense smoke appear, delaying fire suppression efforts. The essence of battery pack fire early warning systems is to race against time before lithium-ion battery thermal runaway occurs, detecting subtle signs such as temperature and characteristic gases before the chain reaction erupts, extinguishing the disaster in its infancy. Once a lithium-ion battery triggers thermal runaway, it transforms from smoking to intense combustion within seconds to minutes, releasing highly toxic fumes and self-sufficiently fueling combustion. Conventional firefighting methods are almost ineffective; therefore, the only effective protection is early warning, detecting anomalies in advance. This not only provides crucial time for evacuation but also automatically triggers circuit cutoffs and suppression measures.

[0003] Traditional battery pack fire early warning methods typically rely on one or a few parameters with corresponding fixed thresholds. When the monitored temperature exceeds the threshold, the system triggers an alarm. However, in real-world scenarios, these multi-source monitored parameters cannot be adjusted in real time for the battery pack containing the lithium batteries, resulting in significant warning lag and a high risk of false alarms or missed alarms. Summary of the Invention

[0004] This invention provides a battery pack fire early warning method and system based on multi-source parameters to solve existing problems.

[0005] The present invention provides a battery pack fire early warning method and system based on multi-source parameters, which adopts the following technical solution: This invention proposes a battery pack fire early warning method based on multi-source parameters, which includes the following steps: Obtain the multi-source monitoring parameters for each battery pack partition within different multi-source parameter categories; By comparing the changes in multi-source monitoring parameters of the same multi-source parameter category before the same time, a multi-source monitoring vector for each battery pack partition at the same time is constructed; the similarity of multi-source monitoring vectors between different battery pack partitions at the same time is analyzed, and the multi-source coupling risk value between different battery pack partitions at each time is calculated by combining the distribution distance between different battery pack partitions. When transitioning from the same battery pack partition to other battery pack partitions, analyze the gradient direction of the changes in multi-source monitoring parameters within different multi-source parameter categories, and calculate the cross-regional gradient anomaly value when transitioning from the same battery pack partition to other battery pack partitions at each time step; perform weighted fusion of multi-source coupling risk value and cross-regional gradient anomaly value, and calculate the partition correlation influence degree when transitioning from the same battery pack partition to other battery pack partitions at each time step. By comparing the changes in values ​​of different multi-source parameter categories during time-series simulation and prediction, the time-series anomaly score of each battery pack partition at different times is calculated. The cross-regional gradient anomaly value, partition correlation influence, and time-series anomaly score are weighted and fused to calculate the fusion risk index when the same battery pack partition transitions to other battery pack partitions at each time point, and fire early warning response is carried out.

[0006] Preferably, the method for obtaining the multi-source monitoring vector is as follows: Within any battery pack partition, the average change of the multi-source monitoring parameters of the same multi-source parameter category within a preset time range before the same moment is taken as the historical risk change rate of the same multi-source parameter category at the same moment; the vector formed by the overall historical risk change rates of all multi-source parameter categories at the same moment is taken as the multi-source monitoring vector of each battery pack partition at the same moment.

[0007] Preferably, the method for obtaining the multi-source coupling risk value is as follows: The similarity mapping results of multi-source monitoring vectors between different battery pack partitions at the same time are used as the multi-source parameter risk similarity between different battery pack partitions; the inverse proportional mapping results of the distance between different battery pack partitions in three-dimensional space are used as the positional coupling correlation between different battery pack partitions; and the product mapping results between multi-source parameter risk similarity and positional coupling correlation are used as the multi-source coupling risk value between different battery pack partitions at the same time.

[0008] Preferably, the method for obtaining the cross-regional gradient outliers is as follows: Take any one battery pack partition as the starting partition and any remaining battery pack partition as the ending partition; at the same time, calculate the transition gradient vector when the starting partition transitions to the ending partition by the direction of change of the multi-source monitoring vector during the transition from the starting partition to the ending partition; combine the direction of the transition gradient vector and the modulus value to calculate the cross-region gradient anomaly value when the starting partition transitions to the ending partition.

[0009] Preferably, the method for obtaining the partition association influence degree is as follows: At the same time, the value after fusing the cross-regional gradient outliers by preset weights is used as the cross-regional gradient weighted outlier during the transition from the starting partition to the ending partition; the mapping result of the product of the cross-regional gradient weighted outlier and the multi-source coupling risk value is used as the partition correlation influence degree during the transition from the starting partition to the ending partition.

[0010] Preferably, the method for obtaining the time series anomaly score is as follows: For the same battery pack partition at the same time, the simulated and predicted values ​​of the multi-source monitoring parameters within each multi-source parameter category are used as the simulated and predicted parameters for each multi-source parameter category; the difference between the multi-source monitoring parameters and the simulated and predicted parameters is used as the time-series anomaly score for the same battery pack partition at the same time.

[0011] Preferably, the method for obtaining the fusion risk index is as follows: By using cross-regional gradient outliers, several high-risk and potential-risk partitions are identified at each time point. When transitioning from the same high-risk partition to each potential-risk partition, the cross-regional gradient outliers, the partition correlation impact, and the time-series outlier score are fused together using a preset fusion weight. This fusion weight is then used as the fusion risk index for the transition from the same high-risk partition to each potential-risk partition at each time point.

[0012] Preferably, the method for obtaining the high-risk partitions and potential risk partitions is as follows: At the same time, the risk classification criteria for cross-regional gradient outliers are determined by classifying battery pack partitions whose values ​​meet the preset threshold conditions as high-risk partitions and those whose values ​​do not meet the preset threshold conditions as potential risk partitions.

[0013] Preferably, the fire early warning response is implemented as follows: At the same time, by combining the fusion risk index when the same high-risk zone transitions to other potential risk zones, several risk response levels are divided, the same high-risk zone is classified, and different levels of early warning and preparedness operations are carried out in conjunction with other potential risk zones.

[0014] The present invention also proposes a battery pack fire early warning system based on multi-source parameters, including a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the steps of the above-mentioned battery pack fire early warning method based on multi-source parameters.

[0015] The beneficial effects of the technical solution of this invention are as follows: By comparing the changes in multi-source monitoring parameters of the same multi-source parameter category before the same time, and combining this with the distribution distance between different battery pack partitions, the multi-source coupling risk value between different battery pack partitions at each time point is calculated; background differences are effectively eliminated, upgrading isolated single-point alarms to spatial perception of the heat spread chain, significantly reducing false alarms and locking the partitions that are about to be affected in advance; when transitioning from the same battery pack partition to other battery pack partitions, the gradient direction of the changes in multi-source monitoring parameters within different multi-source parameter categories is analyzed, and the direction of the transition from the same battery pack partition to other battery pack partitions at each time point is calculated. This invention assesses the impact of battery pack partition transitions; it accurately distinguishes between spontaneous concurrent anomalies and real chain-like propagation, identifying the next affected partition in advance; it compares the changes in values ​​of different multi-source parameter categories during time-series simulation and prediction, calculating the time-series anomaly score for each battery pack partition at different times; it weights and fuses cross-regional gradient anomalies, partition correlation impact, and time-series anomaly scores to calculate the fusion risk index for the transition of the same battery pack partition to other battery pack partitions at each time point, and provides a fire early warning response; it significantly advances the warning window to the very early stage of thermal runaway and effectively suppresses false alarms. This invention partitions the battery pack and constructs spatial propagation correlation relationships through changes in multi-dimensional parameters between different partitions, forming a comprehensive fusion risk criterion covering spatial propagation chains and temporal evolution. It can accurately distinguish between real thermal propagation and global interference at a very early stage, reducing the lag in early warning responses of traditional methods and lowering the risk of false alarms or missed alarms. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of a battery pack fire early warning method based on multi-source parameters according to the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a battery pack fire early warning method and system based on multi-source parameters proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for a battery pack fire early warning method and system based on multi-source parameters provided by the present invention.

[0021] Please see Figure 1 The diagram illustrates a flowchart of a battery pack fire early warning method based on multi-source parameters according to an embodiment of the present invention. The method includes the following steps: Step S001: Obtain the multi-source monitoring parameters for each battery pack partition in different multi-source parameter categories.

[0022] It should be noted that traditional battery pack fire early warning methods typically rely on one or a few parameters with corresponding fixed thresholds. When the monitored temperature exceeds the threshold, the system triggers an alarm. However, in real-world scenarios, these multi-source monitored parameters cannot be adjusted in real time for the battery pack containing the lithium batteries, resulting in significant warning lag and a high risk of false alarms or missed alarms.

[0023] It should be noted beforehand that for large lithium-ion battery packs in energy storage power stations, these packs are typically designed with multiple independent modules in their physical structure. Adjacent battery pack modules have pre-reserved heat insulation gaps filled with high-temperature resistant insulation cotton. Each battery pack module contains several cells and is equipped with an independent BMS slave board, fire extinguishing nozzles, pressure relief valves, and several types of environmental monitoring sensors. The data detected by each type of environmental monitoring sensor serves as a multi-source monitoring parameter for a multi-source parameter category. For the convenience of subsequent explanations, this embodiment uses... Each battery pack is partitioned. This embodiment will be illustrated using a multi-source parameter category as an example, where the example is... , This example is used for illustration; no specific limitations are set in this embodiment. , It depends on the specific implementation situation.

[0024] In one specific implementation of this invention, the process of obtaining multi-source monitoring parameters is as follows: extract the multi-source monitoring parameters of each battery pack partition in each multi-source parameter category from the database of the monitoring platform.

[0025] It should be noted that this embodiment uses voltage, temperature, and carbon dioxide concentration as parameters. This embodiment uses an environmental monitoring sensor as an example, a data sampling frequency of 1Hz as an example, and a heat insulation gap reserved between adjacent battery packs as an example. This embodiment does not impose specific limitations, and the environmental monitoring sensor, data sampling frequency, and heat insulation gap can be determined according to the specific implementation.

[0026] Thus, the multi-source monitoring parameters for each battery pack partition in different multi-source parameter categories are obtained using the above method.

[0027] Step S002: Compare the changes in multi-source monitoring parameters of the same multi-source parameter category before the same time, and construct the multi-source monitoring vector of each battery pack partition at the same time; analyze the similarity of the multi-source monitoring vectors of different battery pack partitions at the same time, and calculate the multi-source coupling risk value between different battery pack partitions at each time by combining the distribution distance between different battery pack partitions.

[0028] It should be noted that by quantifying the multi-source coupling risk value, the fire warning of the battery pack no longer only detects single-point anomalies, but also captures the sprawling coupling characteristics of thermal runaway in the spatial dimension. Constructing multi-source monitoring vectors based on changes prior to the same moment can effectively eliminate background differences between zones, highlighting dynamic acceleration trends such as temperature rise rate, pressure drop rate, and gas concentration slope, thereby fully quantifying the degree of multi-physics field coordinated anomalies in each zone. Furthermore, by analyzing the similarity of vectors between different zones and fusing their distribution distance, it is possible to distinguish between proximity synchronous anomalies caused by local thermal spread and global load fluctuation interference. The multi-source coupling risk value calculated in this way can accurately pinpoint areas with similar anomaly patterns and geographical proximity, upgrading isolated alarms to the perception and prediction of thermal spread chains, significantly suppressing false alarms and predicting zones about to be affected in advance.

[0029] Preferably, in some implementations of the present invention, the method for obtaining the multi-source monitoring vector is as follows: within any battery pack partition, the average change of multi-source monitoring parameters of the same multi-source parameter category within a preset time range before the same moment is taken as the historical risk change rate of the same multi-source parameter category at the same moment; the vector formed by the overall historical risk change rates of all multi-source parameter categories at the same moment is taken as the multi-source monitoring vector of each battery pack partition at the same moment. As an example, the multi-source monitoring vector can be obtained through the following process: Taking any battery pack partition and any time point as an example, the reference historical duration for each multi-source parameter category is preset. Before each multi-source parameter category at that time Within seconds, the mean of the absolute values ​​of the differences between all adjacent multi-source monitoring parameters is used as the historical risk change rate for each multi-source parameter category at that moment; the vector formed by arranging the historical risk change rates of all multi-source parameter categories in ascending order is used as the multi-source monitoring vector for each battery pack partition at that moment. In this embodiment, [the following is used as an example]. This example uses seconds as an example; no specific limitations are set in this embodiment. It depends on the specific implementation. Additionally, different multi-source parameter categories... It can be preset to different values ​​or the same value, depending on the implementation situation.

[0030] Preferably, in some implementations of the present invention, the method for obtaining the multi-source coupling risk value is as follows: The similarity mapping result of the multi-source monitoring vectors of different battery pack partitions at the same time is used as the multi-source parameter risk similarity between different battery pack partitions; the inverse proportional mapping result of the distance between different battery pack partitions in three-dimensional space is used as the positional coupling correlation between different battery pack partitions; the product mapping result between the multi-source parameter risk similarity and the positional coupling correlation is used as the multi-source coupling risk value between different battery pack partitions at the same time. As an example, the multi-source coupling risk value can be obtained through the following process: The cosine similarity of the multi-source monitoring vectors between different battery pack partitions at this time is used as the multi-source parameter risk similarity between the same battery pack partition; the inverse proportional value of the center distance between different battery pack partitions is used as the positional coupling correlation between different battery pack partitions; and the product of the multi-source parameter risk similarity and the positional coupling correlation is used as the multi-source coupling risk value between different battery pack partitions at this time.

[0031] It should be noted that the embodiments adopt The model is used to represent the inverse proportional relationship. As the input to the model, the implementer can choose the inverse proportional function according to the actual situation; the acquisition of cosine similarity is a well-known technique, and will not be described in detail in this embodiment.

[0032] Thus, the multi-source coupling risk value between different battery pack partitions at each time point is obtained using the above method.

[0033] Step S003: When transitioning from the same battery pack partition to other battery pack partitions, analyze the gradient direction of the changes in multi-source monitoring parameters within different multi-source parameter categories, and calculate the cross-regional gradient anomaly value when transitioning from the same battery pack partition to other battery pack partitions at each time step; perform weighted fusion of multi-source coupling risk value and cross-regional gradient anomaly value, and calculate the partition correlation influence degree when transitioning from the same battery pack partition to other battery pack partitions at each time step.

[0034] It should be noted that in space thermal spread analysis, the multi-source coupling risk value focuses on assessing whether two zones are connected in thermal runaway states through anomalous pattern similarity and spatial proximity, but it cannot determine whether there is directional energy and mass transfer. Analyzing the gradient direction of cross-zone parameter changes and calculating gradient anomalies is precisely to capture the spatial surge and anomalous transport trends of physical quantities such as temperature and gas concentration from the source zone to the target zone, thereby confirming the driving force and propagation path of the spread. Weighting and fusing these two factors into a zone correlation impact degree preserves the static evidence of "pattern synchronization" while introducing the dynamic characteristics of "directional transmission," thus enabling a more accurate distinction between spontaneously occurring similar anomalies and real chain spread, and allowing for early identification of the next zone that heat and gas jets will affect, providing a more reliable basis for containment control.

[0035] Preferably, in some implementations of the present invention, the method for obtaining cross-regional gradient anomalies is as follows: Any battery pack partition is taken as the starting partition, and the remaining battery pack partition is taken as the ending partition; at the same time, the transition gradient vector during the transition from the starting partition to the ending partition is calculated by using the direction of change of the multi-source monitoring vector as the transition occurs; the cross-regional gradient anomaly during the transition is calculated by combining the direction of the transition gradient vector and its modulus value. As an example, cross-regional gradient anomalies can be obtained through the following process: Any one battery pack partition is taken as the starting partition, and any remaining adjacent battery pack partition is taken as the ending partition. At the same time, based on the multi-source monitoring vectors of the starting and ending partitions and the direction from the starting partition to the ending partition, the gradient vector from the starting partition to the ending partition is obtained and used as the transition gradient vector when the starting partition transitions to the ending partition. The transition gradient vector contains multiple scalars, and the number of scalars is equal to the number of multi-source parameter categories.

[0036] It should be noted that obtaining the gradient vector based on direction and vector is a well-known existing method, and will not be described in detail in this embodiment.

[0037] Furthermore, the normalized values ​​of all scalars within the transition gradient vector are used as the baseline scalars; the absolute value of the sum of all the baseline scalar values ​​is used as the cross-region gradient outlier value during the transition from the starting region to the ending region.

[0038] It should be noted that this embodiment uses... Normalization is performed on the function as an example, and the implementer can choose the normalization function according to the actual situation.

[0039] Preferably, in some implementations of the present invention, the method for obtaining the partition association influence degree is as follows: at the same time, the value obtained by fusing the cross-regional gradient outliers with preset weights is used as the cross-regional gradient weighted outlier during the transition from the starting partition to the ending partition; the mapping result of the product of the cross-regional gradient weighted outlier and the multi-source coupling risk value is used as the partition association influence degree during the transition from the starting partition to the ending partition. As an example, it can be implemented through the following process: As an example, at the same time, the partition correlation influence can be calculated using the following formula. :

[0040] In the formula, This indicates the degree of partition association impact during the transition from the starting partition to the ending partition; This represents the preset weights, which are used in this embodiment. Let's take an example to illustrate; This indicates anomalies in the gradient across regions during the transition from the starting region to the ending region. This represents the risk value of multi-source coupling between the starting and ending partitions; This represents the cross-region gradient weighted outlier during the transition from the starting partition to the ending partition.

[0041] Thus, the partition association impact degree when the same battery pack partition transitions to other battery pack partitions at each time point is obtained using the above method.

[0042] Step S004: Compare the changes in values ​​of different multi-source parameter categories during time series simulation and prediction, calculate the time series anomaly score of each battery pack partition at different times; weight and fuse the cross-regional gradient anomaly value, partition correlation influence degree and time series anomaly score, calculate the fusion risk index when the same battery pack partition transitions to other battery pack partitions at each time, and conduct fire early warning response.

[0043] It should be noted that spatial spread analysis can only identify existing anomalous correlations and cannot predict the future deterioration trend of a partition itself. Thermal runaway evolves extremely rapidly, and by the time gradients or patterns become synchronized, the critical point is often near, resulting in insufficient early warning windows. Therefore, this problem is addressed by using a quantitative time-series anomaly score. By predicting future changes in parameters such as temperature and gas through time-series simulation and comparing them with a safety baseline, early warning judgments can be pushed forward from the current state to the impending state, preemptively capturing early self-heating in the originating partition. A weighted fusion of cross-regional gradient anomalies, partition correlation impact, and time-series anomaly scores is constructed to construct a comprehensive risk criterion covering spread direction, spatial correlation, and future trends. Correlation impact measures the spread proximity and pattern synchronization between partitions, gradient anomalies strengthen evidence of directional transmission, and time-series scores supplement the possibility of the target partition itself becoming uncontrollable. The synergy of these three factors can effectively distinguish between isolated self-heating, harmless heat diffusion, and true chain spread. The highest alarm is triggered only when adjacent zones have both directional gradients, strong correlations, and a rapidly deteriorating future trend. Thus, by combining spatial spread chain and temporal evolution prediction, the warning time is significantly moved forward and false alarms are suppressed, achieving a very early and highly reliable graded fire warning response.

[0044] Preferably, in some implementations of the present invention, the method for obtaining the timing anomaly score is as follows: For the same battery pack partition at the same time, the simulated predicted values ​​of the multi-source monitoring parameters within each multi-source parameter category are used as the simulated predicted parameters for each multi-source parameter category; the difference between the multi-source monitoring parameters and the simulated predicted parameters is used as the timing anomaly score for the same battery pack partition at the same time. As an example, the timing anomaly score can be obtained through the following process: At the same time within the same battery pack partition, the simulated and predicted values ​​of the multi-source monitoring parameters within each multi-source parameter category are used as the simulated and predicted parameters for each multi-source parameter category. The absolute value of the difference between the multi-source monitoring parameters and the simulated and predicted parameters for the same multi-source parameter category is used as the time-series anomaly score for the same multi-source parameter category. The average of the time-series anomaly scores for the same multi-source parameter category is used as the time-series anomaly score for the same battery pack partition at the same time.

[0045] It should be noted that in this embodiment, the historical data sequence consisting of all multi-source monitoring parameters in each multi-source parameter category before the same time in the same battery pack partition is used as input data and input to the Exponential Smoothing prediction algorithm. The Exponential Smoothing prediction algorithm is used to perform simulation prediction processing to obtain the corresponding simulated prediction value. The process of performing simulation prediction is a well-known part of the Exponential Smoothing prediction algorithm and will not be described in detail in this embodiment.

[0046] Preferably, in some implementations of the present invention, the method for obtaining the fusion risk index is as follows: Based on cross-regional gradient outliers, several high-risk partitions and potential risk partitions are divided at each time step; when transitioning from the same high-risk partition to each potential risk partition, the cross-regional gradient outlier, the partition correlation influence, and the time-series outlier score are fused using a preset fusion weight, and the calculated value is used as the fusion risk index for the transition from the same high-risk partition to each potential risk partition at each time step. As an example, the fusion risk index can be obtained through the following process: Preferably, in some implementations of the present invention, the method for obtaining high-risk partitions and potential-risk partitions is as follows: at the same time, based on the criteria for classifying cross-regional gradient outlier risks, battery pack partitions whose values ​​meet a preset threshold condition are designated as high-risk partitions, and battery pack partitions whose values ​​do not meet the preset threshold condition are designated as potential-risk partitions. As an example, high-risk partitions and potential-risk partitions can be obtained through the following process: At any given moment, taking any battery pack partition as an example, the normalized value of the mean of the cross-regional gradient anomalies during the transition from that battery pack partition to other battery packs is taken as the partition risk anomaly value for that battery pack partition; if the partition risk anomaly value is greater than a preset threshold... If the battery pack partition is classified as a high-risk partition, then the partition risk anomaly value is considered to be less than or equal to a preset threshold. Therefore, the battery pack is partitioned as a potential risk partition. At each time point, there are multiple high-risk partitions and potential risk partitions. Furthermore, this embodiment uses... This example is used for illustration; no specific limitations are set in this embodiment. It depends on the specific implementation situation.

[0047] Furthermore, when transitioning from the same high-risk partition to each potential risk partition, the normalized value of the product of the cross-regional gradient weighted outlier, the partition correlation impact, and the time-series outlier score is used as the fusion risk index at each time point when transitioning from the same high-risk partition to each potential risk partition.

[0048] Preferably, in some implementations of the embodiments of the present invention, the fire early warning response process is as follows: At the same time, by comprehensively considering the fusion risk index during the transition from the same high-risk zone to other potential risk zones, several risk response levels are divided, the same high-risk zone is classified, and other potential risk zones are linked to carry out early warning and preparedness operations of different degrees. As an example, the fire early warning response can be implemented through the following specific process: At any given moment, taking any high-risk partition as an example, the average of the fusion risk indices during the transition from that high-risk partition to other potential risk partitions is used as the risk level determination value for that high-risk partition. .

[0049] Furthermore, three risk level thresholds are preset. , , In this embodiment, , , This example is used for illustration; no specific limitations are set in this embodiment. , , It depends on the specific implementation situation and must meet the following requirements. .

[0050] Furthermore, if Greater than and less than The system issues an audible and visual alert and highlights the affected partition on the human-machine interface, displaying the words "Observation Level." Simultaneously, it activates the partition-related prediction logic, increases its data sampling frequency, and enters a high-alert state.

[0051] Furthermore, if Greater than and less than The system drives the alarm to sound and displays the warning level. The linkage control module pre-pressurizes the fire extinguishing nozzle pipeline of the zone to keep it ready to start spraying at any time, and issues a preparation command to the nozzles in the downstream spread zone to open the pipeline valves and keep them in a dry state of preparation.

[0052] Furthermore, if Greater than The system displays the word "trigger level" and immediately triggers the fire suppression nozzles in that zone to spray and suppress the fire. At the same time, based on the spread path predicted by the built-in intelligent module, it automatically triggers the nozzles in the downstream spread zone to spray synchronously or with a delay to prevent heat spread. The pressure relief valve opens according to the internal pressure interlock.

[0053] It should be noted that each battery pack section in this embodiment is equipped with an independent BMS subsystem, which is responsible for the acquisition and status monitoring of the cell voltage and current of the corresponding section; each section has a fixed-point fire extinguishing nozzle at the top and a one-way pressure relief valve on the side, and the nozzle and pressure relief valve are independently controlled by the section BMS.

[0054] Through the above steps, a battery pack fire early warning method based on multi-source parameters is completed.

[0055] Another embodiment of the present invention provides a battery pack fire early warning system based on multi-source parameters. The system includes a memory and a processor. When the processor executes a computer program stored in the memory, it performs the above-described method steps S001 to S004.

[0056] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 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 this application, and should all be included within the protection scope of this application.

Claims

1. A battery pack fire early warning method based on multi-source parameters, characterized in that, The method includes the following steps: Obtain the multi-source monitoring parameters for each battery pack partition within different multi-source parameter categories; By comparing the changes in multi-source monitoring parameters of the same multi-source parameter category before the same time, a multi-source monitoring vector for each battery pack partition at the same time is constructed; the similarity of multi-source monitoring vectors between different battery pack partitions at the same time is analyzed, and the multi-source coupling risk value between different battery pack partitions at each time is calculated by combining the distribution distance between different battery pack partitions. When transitioning from the same battery pack partition to other battery pack partitions, analyze the gradient direction of the changes in multi-source monitoring parameters within different multi-source parameter categories, and calculate the cross-regional gradient anomaly value when transitioning from the same battery pack partition to other battery pack partitions at each time step; perform weighted fusion of multi-source coupling risk value and cross-regional gradient anomaly value, and calculate the partition correlation influence degree when transitioning from the same battery pack partition to other battery pack partitions at each time step. By comparing the changes in values ​​of different multi-source parameter categories during time-series simulation and prediction, the time-series anomaly score of each battery pack partition at different times is calculated. The cross-regional gradient anomaly value, partition correlation influence, and time-series anomaly score are weighted and fused to calculate the fusion risk index when the same battery pack partition transitions to other battery pack partitions at each time point, and fire early warning response is carried out.

2. The battery pack fire early warning method based on multi-source parameters according to claim 1, characterized in that, The method for obtaining the multi-source monitoring vector is as follows: Within any battery pack partition, the average change of the multi-source monitoring parameters of the same multi-source parameter category within a preset time range before the same moment is taken as the historical risk change rate of the same multi-source parameter category at the same moment; the vector formed by the overall historical risk change rates of all multi-source parameter categories at the same moment is taken as the multi-source monitoring vector of each battery pack partition at the same moment.

3. The battery pack fire early warning method based on multi-source parameters according to claim 1, characterized in that, The method for obtaining the multi-source coupling risk value is as follows: The similarity mapping results of multi-source monitoring vectors between different battery pack partitions at the same time are used as the risk similarity of multi-source parameters between different battery pack partitions; The inverse proportional mapping result of the distance between different battery pack partitions in three-dimensional space is used as the positional coupling correlation between different battery pack partitions; The product mapping result between the risk similarity of multi-source parameters and the location coupling correlation is used as the multi-source coupling risk value between different battery pack partitions at the same time.

4. The battery pack fire early warning method based on multi-source parameters according to claim 1, characterized in that, The method for obtaining cross-regional gradient outliers is as follows: Take any one battery pack partition as the starting partition and any remaining battery pack partition as the ending partition; at the same time, calculate the transition gradient vector when the starting partition transitions to the ending partition by the direction of change of the multi-source monitoring vector during the transition from the starting partition to the ending partition; combine the direction of the transition gradient vector and the modulus value to calculate the cross-region gradient anomaly value when the starting partition transitions to the ending partition.

5. The battery pack fire early warning method based on multi-source parameters according to claim 4, characterized in that, The method for obtaining the partition association influence is as follows: At the same time, the value after fusing the cross-regional gradient outliers by preset weights is used as the cross-regional gradient weighted outlier during the transition from the starting partition to the ending partition; the mapping result of the product of the cross-regional gradient weighted outlier and the multi-source coupling risk value is used as the partition correlation influence degree during the transition from the starting partition to the ending partition.

6. The battery pack fire early warning method based on multi-source parameters according to claim 1, characterized in that, The method for obtaining the time series anomaly score is as follows: For the same battery pack partition at the same time, the simulated and predicted values ​​of the multi-source monitoring parameters within each multi-source parameter category are used as the simulated and predicted parameters for each multi-source parameter category; the difference between the multi-source monitoring parameters and the simulated and predicted parameters is used as the time-series anomaly score for the same battery pack partition at the same time.

7. The battery pack fire early warning method based on multi-source parameters according to claim 1, characterized in that, The method for obtaining the fusion risk index is as follows: By using cross-regional gradient outliers, several high-risk and potential-risk zones are identified at each time point. When transitioning from the same high-risk partition to each potential risk partition, the cross-regional gradient outlier, partition correlation impact, and time-series outlier score are fused together using a preset fusion weight and used as the fusion risk index for the transition from the same high-risk partition to each potential risk partition at each time point.

8. The battery pack fire early warning method based on multi-source parameters according to claim 7, characterized in that, The method for obtaining the high-risk and potential-risk zones is as follows: At the same time, the risk classification criteria for cross-regional gradient outliers are determined by classifying battery pack partitions whose values ​​meet the preset threshold conditions as high-risk partitions and those whose values ​​do not meet the preset threshold conditions as potential risk partitions.

9. The battery pack fire early warning method based on multi-source parameters according to claim 7, characterized in that, The implementation process of the fire early warning response is as follows: At the same time, by combining the fusion risk index when the same high-risk zone transitions to other potential risk zones, several risk response levels are divided, the same high-risk zone is classified, and different levels of early warning and preparedness operations are carried out in conjunction with other potential risk zones.

10. A battery pack fire early warning system based on multi-source parameters, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of a battery pack fire early warning method based on multi-source parameters as described in any one of claims 1-9.

Images