A quantitative grading evaluation method for fire disaster hazard of lithium battery system

Abstract

This invention discloses a quantitative classification and assessment method for the hazard of fires in lithium battery systems. This invention relates to the field of fire assessment technology, and obtains a multi-dimensional evaluation index set characterizing the degree of fire hazard. The multi-dimensional evaluation index set includes the heat release rate (QHR) characterizing the fire's heat release capacity, the smoke toxicity (TOX) quantifying the toxicity of combustion products, the explosion potential (EXP) reflecting the risk of flammable gas accumulation and sudden release, and the fire spread rate (SPD) characterizing the fire's expansion trend. This invention constructs a multi-dimensional fire hazard assessment index including heat release rate, smoke toxicity, explosion potential, and fire spread rate, and combines this with the analytic hierarchy process (AHP) to determine the relative weights of each hazard-causing factor. This achieves quantitative modeling and comprehensive assessment of the hazard degree of fires in lithium battery systems, more objectively reflecting the true hazard characteristics during the fire development process, and improving the scientific rigor and consistency of fire risk assessment results.

Description

Technical Field

[0001] This invention relates to the field of fire assessment technology, specifically a quantitative classification assessment method for the hazard of fires in lithium battery systems. Background Technology

[0002] Most methods for assessing the hazards of lithium battery fires rely on qualitative analysis or single-indicator judgments, typically focusing on phenomena such as whether thermal runaway has occurred, whether open flames have been generated, or whether an explosion has been triggered. These methods lack a systematic and quantitative comprehensive assessment mechanism for the multidimensional hazard factors during the fire process. In practical applications, some assessment methods rely solely on single parameters such as heat release rate, surface temperature, or smoke generation to make risk judgments. This makes it difficult to fully reflect the differences in the comprehensive hazards of lithium battery fires under different operating conditions, structural forms, and evolution stages, resulting in a lack of comparability in risk levels between different battery systems and different fire scenarios. Existing methods for assessing fire hazards fail to adequately consider the coupling relationships between key hazard-causing pathways such as smoke toxicity, the explosion potential of combustible gases, and the speed of fire spread. They also lack weighting and objective quantitative support based on historical accident data and experimental samples. To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a quantitative classification and assessment method for the fire hazards of lithium battery systems.

[0004] To achieve the above objectives, the present invention provides a quantitative classification and assessment method for the hazard of fires in lithium battery systems, comprising: S1. Obtain a multi-dimensional evaluation index set that characterizes the degree of fire hazard. The multi-dimensional evaluation index set includes the heat release rate (QHR) for characterizing the heat release capacity of a fire, the smoke toxicity (TOX) for quantifying the toxicity of combustion products, the explosion potential (EXP) for reflecting the risk of accumulation and sudden release of combustible gases, and the fire spread rate (SPD) for characterizing the fire spread trend. S2. Based on historical lithium battery fire accident samples and standardized experimental data, a hierarchical evaluation structure model is constructed to conduct pairwise comparative analysis of the relative importance of each fire hazard index and obtain the index weight vector W. S3. Perform interval mapping and standardization on the collected raw data of fire hazard indicators, and combine the scores of each indicator with the indicator weights to construct a fire hazard quantitative scoring model and output the fire hazard score value HRS. S4. Based on the Fire Hazard Score (HRS), establish a mapping relationship between the score and the hazard level, determine the range of the target fire sample, and output the fire hazard level identifier.

[0005] Through multi-channel data acquisition, data were collected on heat release rate (QHR), smoke toxicity (TOX), explosion potential (EXP), and fire spread rate (SPD), including: The heat release rate QHR was acquired by using an ARC heat release analysis device to conduct a combustion test on a lithium battery pack under controlled thermal runaway conditions, measuring the heat release rate curve per unit time, and extracting the heat release rate QHR. The collection of TOX produced by the smoke: The concentration of the target toxic components CO, HF and HCl in the gas released by the combustion of the battery is analyzed by non-dispersive infrared spectroscopy, and the TOX produced by the smoke is calculated according to the toxicity equivalent conversion formula. The explosive potential (EXP) is collected by setting up a sealed chamber and arranging a high-frequency pressure sensor array to record the pressure during the battery explosion, calculating the maximum casing rupture pressure, and evaluating the explosive potential (EXP) based on the maximum casing rupture pressure. The fire spread rate (SPD) was collected by deploying thermocouple arrays at multiple points inside and around the lithium battery module, recording the time it takes for the combustion front to reach each monitoring point, and calculating the fire spread rate (SPD) by the distance between the points and the time difference.

[0006] The process of determining the indicator weight vector W includes: A hierarchical structure model of the analytic hierarchy process is constructed, in which the target layer is used to characterize the evaluation target of the hazard of lithium battery fire, the criteria layer sets the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP and the fire spread rate index SPD, and the index layer is used to carry the statistical characteristic parameters corresponding to each criterion layer. Based on historical lithium battery fire accident database and standardized fire test samples, the relative importance of each indicator in the criterion layer is compared and evaluated. A pairwise comparison matrix is ​​constructed according to the preset scaling rules, so that the matrix elements reflect the relative influence of the corresponding indicators in fire catastrophicity. The pairwise comparison matrix is ​​subjected to eigenvector solving, and the eigenvector corresponding to the largest eigenvalue is extracted as the initial weight parameters of each index. The pairwise comparison matrix is ​​then subjected to a consistency check. When the consistency check result meets the preset consistency condition, the weight parameters are confirmed to be valid.

[0007] The weight parameters that pass the consistency test are processed to form a weight vector W, where the weight wi corresponding to the i-th index is obtained according to the following relationship: Wherein, the weight parameter wi represents the final weight coefficient corresponding to the i-th catastrophicity evaluation index; the initial weight parameter ai represents the importance component of the i-th index obtained by solving the maximum eigenvalue of the pairwise comparison matrix; and an represents the summation of the initial weight parameters of all indicators.

[0008] The calculation of the weighted contribution value includes: For the heat release rate index QHR, smoke toxicity index TOX, explosion potential index EXP, and fire spread rate index SPD, their original statistical characteristic values ​​are extracted from fire test samples. The original statistical characteristic values ​​are physical statistics. The original statistical characteristic values ​​of each indicator are processed by interval mapping to map them to a unified indicator score interval, resulting in the corresponding standardized indicator scores S_QHR, S_TOX, S_EXP and S_SPD. The weight vector W obtained after the consistency test of the analytic hierarchy process is weighted and calculated with the corresponding index scores to form a weighted contribution value.

[0009] The fire hazard quantification scoring model includes a scoring network structure model MOD consisting of an input layer, multiple feature processing layers, and a scoring output layer, where: The input layer contains four input nodes, which correspond to the weighted contribution values ​​of the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD, respectively. The feature processing layer includes a first processing layer, a second processing layer, and a third processing layer. The first processing layer contains 64 nodes, the second processing layer contains 32 nodes, and the third processing layer contains 16 nodes. Each processing layer uses the ReLU activation function to perform non-linear mapping on the input data. A BatchNormalization layer is set after the second processing layer to improve the stability of model training; a Dropout layer with a Dropout rate of 0.2 is set after the third processing layer to prevent the scoring model from overfitting. The scoring output layer contains one output node, which is used to output the fire hazard score (HRS).

[0010] The weighted feature vector undergoes linear transformation and nonlinear mapping layer by layer through the first, second, and third feature processing layers. After the second feature processing layer, normalization is applied to stabilize the feature distribution, and after the third feature processing layer, random inactivation is used to suppress feature co-adaptation. Finally, the fire hazard score is output by the scoring output layer, calculated using the following formula: The comprehensive feature input vector X consists of the weighted contribution values ​​corresponding to the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD; W1, W2, and W3 represent the weight matrices corresponding to each feature processing layer, and b1, b2, and b3 are bias terms; BN represents the batch normalization process; f2, f3, and f0 all represent nonlinear mapping operations using the ReLU activation function.

[0011] Output the fire hazard level, including: Based on the Fire Hazard Score (HRS), a grade mapping function (GRF) is constructed. The GRF divides the range according to the statistical distribution characteristics of HRS values ​​in historical lithium battery fire samples, and sets multiple range thresholds, including: low risk level (L1, 0.00≤HRS<0.25), medium risk level (L2, 0.25≤HRS<0.50), second-high risk level (L3, 0.50≤HRS<0.75), and high risk level (L4, 0.75≤HRS≤1.00). The fire score (HRS) of the target fire sample is input into the rating function (GRF) for interval judgment, and the corresponding rating label (Label) is output.

[0012] This invention provides a quantitative classification and assessment method for the hazard of fires in lithium battery systems. Compared with existing technologies, it has the following advantages: This invention constructs a multidimensional fire hazard assessment index that includes heat release rate, smoke toxicity, explosion potential, and fire spread rate, and combines the analytic hierarchy process (AHP) to determine the relative weights of each hazard factor. This enables quantitative modeling and comprehensive assessment of the degree of fire hazard in lithium battery systems, which can more objectively reflect the true hazard characteristics in the fire development process and improve the scientificity and consistency of fire risk assessment results. This invention introduces a quantitative scoring model and a level mapping mechanism for fire hazards, transforming complex multi-source fire experimental data into unified fire hazard scores and clear hazard level identifiers. This transforms the risk of lithium battery fires from a state of being difficult to compare into a standardized result that can be graded and compared. This facilitates the rapid identification of high-risk objects in battery design evaluation, transportation and storage management, and safety supervision scenarios, effectively reducing the risk of personal injury and property damage caused by lithium battery fires. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the principle framework of the present invention. Detailed Implementation

[0014] 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 embodiments of the present invention, and not all embodiments. 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.

[0015] Please see Figure 1 This application provides a quantitative classification and assessment method for the hazard of fires in lithium battery systems, including: S1. Obtain a multi-dimensional evaluation index set that characterizes the degree of fire hazard. The multi-dimensional evaluation index set includes the heat release rate (QHR) for characterizing the heat release capacity of a fire, the smoke toxicity (TOX) for quantifying the toxicity of combustion products, the explosion potential (EXP) for reflecting the risk of accumulation and sudden release of combustible gases, and the fire spread rate (SPD) for characterizing the fire spread trend. S2. Based on historical lithium battery fire accident samples and standardized experimental data, a hierarchical evaluation structure model is constructed to conduct pairwise comparative analysis of the relative importance of each fire hazard index and obtain the index weight vector W. S3. Perform interval mapping and standardization on the collected raw data of fire hazard indicators, and combine the scores of each indicator with the indicator weights to construct a fire hazard quantitative scoring model and output the fire hazard score value HRS. S4. Based on the Fire Hazard Score (HRS), establish a mapping relationship between the score and the hazard level, determine the range of the target fire sample, and output the fire hazard level identifier.

[0016] Through multi-channel data acquisition, data were collected on heat release rate (QHR), smoke toxicity (TOX), explosion potential (EXP), and fire spread rate (SPD), including: The heat release rate QHR was acquired by using an ARC heat release analysis device to conduct a combustion test on a lithium battery pack under controlled thermal runaway conditions, measuring the heat release rate curve per unit time, and extracting the heat release rate QHR. The collection of TOX produced by the smoke: The concentration of the target toxic components CO, HF and HCl in the gas released by the combustion of the battery is analyzed by non-dispersive infrared spectroscopy, and the TOX produced by the smoke is calculated according to the toxicity equivalent conversion formula. The explosive potential (EXP) is collected by setting up a sealed chamber and arranging a high-frequency pressure sensor array to record the pressure during the battery explosion, calculating the maximum casing rupture pressure, and evaluating the explosive potential (EXP) based on the maximum casing rupture pressure. The fire spread rate (SPD) was collected by deploying thermocouple arrays at multiple points inside and around the lithium battery module, recording the time it takes for the combustion front to reach each monitoring point, and calculating the fire spread rate (SPD) by the distance between the points and the time difference.

[0017] In this embodiment, a multi-channel synchronous acquisition process for fire physical quantities is constructed to quantitatively collect key catastrophic parameters during a lithium battery system fire. Firstly, under controlled thermal runaway experimental conditions, the lithium battery pack under test is placed in an ARC (Anaerobic Thermal Release Analysis) device. Thermal runaway is induced through programmed temperature increases or external triggering. The heat flow measurement unit built into the ARC device continuously monitors the heat released per unit time during combustion, generating a curve of the heat release rate over time. The heat release rate index (QHR), characterizing the fire's thermal energy release capability, is extracted from this curve. Simultaneously, a gas sampling pipeline is introduced into the combustion gas acquisition channel to sample the flue gas released during lithium battery combustion in real time. Non-dispersive infrared detection methods are used to detect the volume fraction or mass concentration of typical toxic gas components such as carbon monoxide, hydrogen fluoride, and hydrogen chloride. The toxicity equivalents of various toxic components are then comprehensively converted according to a preset toxicity equivalent conversion rule. The method yields the smoke toxicity index TOX, used to quantify the toxicity level of combustion products. Furthermore, a closed pressure-resistant test chamber is installed outside the battery combustion test chamber, with a multi-point high-frequency pressure sensor array arranged on the inner wall of the chamber. This array is used to synchronously record the pressure changes inside the chamber when the battery casing ruptures or experiences a violent energy release event. By performing peak analysis on the pressure time series data, the maximum casing rupture pressure is obtained as a characterization of explosion intensity, and the explosion potential index EXP corresponding to the accumulation and sudden release of combustible gases during a lithium battery fire is assessed accordingly. In addition, multiple thermocouple sensor arrays are deployed at key locations inside the lithium battery module and along the edge of the module in a predetermined direction to monitor the temperature changes of each monitoring point over time in real time. When the flame or high-temperature front reaches different monitoring points sequentially, the corresponding arrival time is recorded. The propulsion speed of the flame front is calculated by the spatial distance and time difference between adjacent monitoring points, thereby obtaining the fire spread rate index SPD, which characterizes the fire expansion trend.

[0018] The process of determining the indicator weight vector W includes: A hierarchical structure model of the analytic hierarchy process is constructed, in which the target layer is used to characterize the evaluation target of the hazard of lithium battery fire, the criteria layer sets the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP and the fire spread rate index SPD, and the index layer is used to carry the statistical characteristic parameters corresponding to each criterion layer. Based on historical lithium battery fire accident database and standardized fire test samples, the relative importance of each indicator in the criterion layer is compared and evaluated. A pairwise comparison matrix is ​​constructed according to the preset scaling rules, so that the matrix elements reflect the relative influence of the corresponding indicators in fire catastrophicity. The pairwise comparison matrix is ​​subjected to eigenvector solving, and the eigenvector corresponding to the largest eigenvalue is extracted as the initial weight parameters of each index. The pairwise comparison matrix is ​​then subjected to a consistency check. When the consistency check result meets the preset consistency condition, the weight parameters are confirmed to be valid.

[0019] In this embodiment, the determination of the index weight vector W is achieved through a standardized process of the analytic hierarchy process (AHP). First, a hierarchical evaluation structure model is constructed based on the thermal runaway and combustion hazard mechanism of lithium battery fires. The comprehensive evaluation target of the hazard caused by lithium battery fires is set as the target layer. Under the target layer, a criterion layer is set to characterize the core influencing factors of fire hazard. The criterion layer sequentially includes the heat release rate (QHR), smoke toxicity (TOX), explosion potential (EXP), and fire spread rate (SPD). Corresponding index layers are constructed under each criterion layer index to carry the characteristic parameters obtained from fire experiments and accident sample statistics, thus realizing a step-by-step mapping of the evaluation system from abstract targets to quantifiable parameters. Based on this, a historical lithium battery fire accident database and standardized fire experiment samples are used to evaluate the hazard indicators in the criterion layer under different fire scenarios. The impact of the degree of hazard on the fire is statistically analyzed and evaluated with expert assistance. According to the preset relative importance scaling rules, the importance of any two indicators is compared pairwise to construct a pairwise comparison matrix that reflects the relative contribution of each indicator to the fire hazard assessment. Each element in the matrix can characterize the relative influence intensity of the corresponding indicator in the fire hazard assessment. Subsequently, the pairwise comparison matrix is ​​processed by eigenvector solving. By calculating the maximum eigenvalue of the matrix and extracting its corresponding eigenvector, the eigenvector is used as the initial weight parameter set for each criterion layer indicator. At the same time, the pairwise comparison matrix is ​​subjected to consistency verification. When the consistency verification result meets the preset consistency judgment condition, it is confirmed that the initial weight parameters are logically reasonable and stable, and they are used as the effective input basis for the subsequent construction of weight vector W and the calculation of the fire hazard quantification scoring model.

[0020] The weight parameters that pass the consistency test are processed to form a weight vector W, where the weight wi corresponding to the i-th index is obtained according to the following relationship: Wherein, the weight parameter wi represents the final weight coefficient corresponding to the i-th catastrophicity evaluation index; the initial weight parameter ai represents the importance component of the i-th index obtained by solving the maximum eigenvalue of the pairwise comparison matrix; and an represents the summation of the initial weight parameters of all indicators.

[0021] After completing the consistency check of the pairwise comparison matrix and confirming the validity of each initial weight parameter, the initial weight parameters are further normalized to form a weight vector W for quantifying fire hazard. Specifically, the process is as follows: First, the importance components corresponding to each hazard assessment index obtained by solving for the maximum eigenvalue are used as the initial weight parameter set, where the initial weight parameter corresponding to the i-th index is denoted as ai. This initial weight parameter reflects the relative hazard contribution intensity of the index relative to other indicators in the hierarchical analysis structure. Then, the initial weight parameter set is summed to obtain the normalized baseline. The value an is the sum of the initial weight parameters of all indicators participating in the fire hazard assessment, used to standardize the weights of each indicator. Based on this, according to the preset weight normalization rule, each initial weight parameter ai is proportionalized to calculate the corresponding final weight parameter wi, so that the weights of each indicator fall into a uniform numerical range and meet the requirements of weight comparability. In the weight vector W formed by the above processing, each weight parameter wi not only maintains the ranking relationship of the relative importance of each indicator in the original analytic hierarchy process, but also eliminates the influence of different dimensions and scale differences on the subsequent calculation of fire hazard score.

[0022] The calculation of the weighted contribution value includes: For the heat release rate index QHR, smoke toxicity index TOX, explosion potential index EXP, and fire spread rate index SPD, their original statistical characteristic values ​​are extracted from fire test samples. The original statistical characteristic values ​​are physical statistics. The original statistical characteristic values ​​of each indicator are processed by interval mapping to map them to a unified indicator score interval, resulting in the corresponding standardized indicator scores S_QHR, S_TOX, S_EXP and S_SPD. The weight vector W obtained after the consistency test of the analytic hierarchy process is weighted and calculated with the corresponding index scores to form a weighted contribution value.

[0023] In this embodiment, the calculation process of the weighted contribution value is performed after the indicator weight vector W is determined. This process is used to uniformly transform fire hazard indicators of different physical dimensions into quantitative inputs that can directly participate in the comprehensive scoring calculation. The specific process is as follows: First, for the heat release rate indicator QHR, smoke toxicity indicator TOX, explosion potential indicator EXP, and fire spread rate indicator SPD, corresponding original statistical feature values ​​are extracted from controlled lithium battery fire experimental samples or standardized accident samples. These original statistical feature values ​​are measured or statistical quantities that directly reflect the physical process of the fire, used to characterize the actual intensity level of each hazard factor under specific fire conditions. Subsequently, the original statistical feature values ​​of each indicator are mapped according to a preset indicator interval mapping rule. Then, standardization processing is performed by mapping the original physical quantities to a unified index score range, making the indicators of different dimensions and orders of magnitude comparable at the numerical level. This yields the standardized scores S_QHR for the heat release rate, S_TOX for the smoke toxicity index, S_EXP for the explosion potential index, and S_SPD for the fire spread rate index. After obtaining the standardized scores, the weight vector W determined by the analytic hierarchy process (AHP) consistency test is used as the weighting benchmark. The scores of each standardized index are weighted according to the one-to-one correspondence between the indicators, forming a weighted contribution value that reflects the degree of contribution of each hazard index to the overall fire hazard. The fire hazard quantification scoring model includes a scoring network structure model MOD consisting of an input layer, multiple feature processing layers, and a scoring output layer, where: The input layer contains four input nodes, which correspond to the weighted contribution values ​​of the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD, respectively. The feature processing layer includes a first processing layer, a second processing layer, and a third processing layer. The first processing layer contains 64 nodes, the second processing layer contains 32 nodes, and the third processing layer contains 16 nodes. Each processing layer uses the ReLU activation function to perform non-linear mapping on the input data. A BatchNormalization layer is set after the second processing layer to improve the stability of model training; a Dropout layer with a Dropout rate of 0.2 is set after the third processing layer to prevent the scoring model from overfitting. The scoring output layer contains one output node, which is used to output the fire hazard score (HRS).

[0024] In this embodiment, the fire hazard quantification scoring model is constructed using a hierarchical neural network structure to perform nonlinear fusion and comprehensive discrimination of the aforementioned weighted contribution values. The specific implementation process is as follows: First, the weighted contribution values ​​corresponding to the heat release rate (QHR), smoke toxicity (TOX), explosion potential (EXP), and fire spread rate (SPD) indicators obtained from the weighted contribution value calculation step are used as model input features and input into the input layer of the scoring network structure model MOD. The input layer has four input nodes to carry the comprehensive contribution information of each hazard index. Subsequently, the feature vector output from the input layer is sequentially fed into the first feature processing layer, the second feature processing layer, and the third feature processing layer for layer-by-layer processing. The first feature processing layer has 64 processing nodes for preliminary linear combination and nonlinear mapping of the input features, and the second feature processing layer has 32 processing nodes for further extraction of higher-order correlations between each hazard index. The third feature processing layer has 16 processing nodes to compress feature dimensions and enhance the ability to express key disaster-causing modes. Each feature processing layer uses the ReLU activation function to achieve nonlinear mapping, thereby enhancing the model's ability to fit complex fire-causing mechanisms. After the output of the second feature processing layer, BatchNormalization is introduced to standardize the distribution of intermediate features, reducing the impact of sample differences on model training stability. After the output of the third feature processing layer, a Dropout mechanism is introduced to deactivate feature nodes at a random deactivation ratio of 0.2, suppressing the model's over-reliance on local features and reducing the risk of overfitting in the scoring model. The feature vectors processed and regularized are then input to the scoring output layer, which has one output node to output the Fire Hazard Score (HRS), representing the comprehensive disaster-causing severity of lithium battery fires.

[0025] This invention constructs a multidimensional fire hazard assessment index that includes heat release rate, smoke toxicity, explosion potential, and fire spread rate, and combines the analytic hierarchy process (AHP) to determine the relative weights of each hazard factor. This enables quantitative modeling and comprehensive assessment of the fire hazard level of lithium battery systems, which can more objectively reflect the true hazard characteristics in the fire development process and improve the scientificity and consistency of fire risk assessment results.

[0026] The weighted feature vector undergoes linear transformation and nonlinear mapping layer by layer through the first, second, and third feature processing layers. After the second feature processing layer, normalization is applied to stabilize the feature distribution, and after the third feature processing layer, random inactivation is used to suppress feature co-adaptation. Finally, the fire hazard score is output by the scoring output layer, calculated using the following formula: The comprehensive feature input vector X consists of the weighted contribution values ​​corresponding to the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD; W1, W2, and W3 represent the weight matrices corresponding to each feature processing layer, and b1, b2, and b3 are bias terms; BN represents the batch normalization process; f2, f3, and f0 all represent nonlinear mapping operations using the ReLU activation function.

[0027] Output the fire hazard level, including: Based on the Fire Hazard Score (HRS), a grade mapping function (GRF) is constructed. The GRF divides the range according to the statistical distribution characteristics of HRS values ​​in historical lithium battery fire samples, and sets multiple range thresholds, including: low risk level (L1, 0.00≤HRS<0.25), medium risk level (L2, 0.25≤HRS<0.50), second-high risk level (L3, 0.50≤HRS<0.75), and high risk level (L4, 0.75≤HRS≤1.00). The fire score (HRS) of the target fire sample is input into the rating function (GRF) for interval judgment, and the corresponding rating label (Label) is output.

[0028] Beneficial effects: This invention introduces a quantitative scoring model and a level mapping mechanism for fire hazards, transforming complex multi-source fire experimental data into unified fire hazard scores and clear hazard level identifiers. This transforms the risk of lithium battery fires from a state of being difficult to compare into a standardized result that can be graded and compared. This facilitates the rapid identification of high-risk objects in battery design evaluation, transportation and storage management, and safety supervision scenarios, effectively reducing the risk of personal injury and property damage caused by lithium battery fires.

[0029] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0030] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0031] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A quantitative classification and assessment method for the hazard of fires in lithium battery systems, characterized in that, include: S1. Obtain a multi-dimensional evaluation index set that characterizes the degree of fire hazard. The multi-dimensional evaluation index set includes the heat release rate (QHR) for characterizing the heat release capacity of a fire, the smoke toxicity (TOX) for quantifying the toxicity of combustion products, the explosion potential (EXP) for reflecting the risk of accumulation and sudden release of combustible gases, and the fire spread rate (SPD) for characterizing the fire spread trend. S2. Based on historical lithium battery fire accident samples and standardized experimental data, a hierarchical evaluation structure model is constructed to conduct pairwise comparative analysis of the relative importance of each fire hazard index and obtain the index weight vector W. S3. Perform interval mapping and standardization on the collected raw data of fire hazard indicators, and combine the scores of each indicator with the indicator weights to construct a fire hazard quantitative scoring model and output the fire hazard score value HRS. S4. Based on the Fire Hazard Score (HRS), establish a mapping relationship between the score and the hazard level, determine the range of the target fire sample, and output the fire hazard level identifier.

2. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, Through multi-channel data acquisition, data were collected on heat release rate (QHR), smoke toxicity (TOX), explosion potential (EXP), and fire spread rate (SPD), including: The heat release rate QHR was acquired by using an ARC heat release analysis device to conduct a combustion test on a lithium battery pack under controlled thermal runaway conditions, measuring the heat release rate curve per unit time, and extracting the heat release rate QHR. The collection of TOX produced by the smoke: The concentration of the target toxic components CO, HF and HCl in the gas released by the combustion of the battery is analyzed by non-dispersive infrared spectroscopy, and the TOX produced by the smoke is calculated according to the toxicity equivalent conversion formula. The explosive potential (EXP) is collected by setting up a sealed chamber and arranging a high-frequency pressure sensor array to record the pressure during the battery explosion, calculating the maximum casing rupture pressure, and evaluating the explosive potential (EXP) based on the maximum casing rupture pressure. The fire spread rate (SPD) was collected by deploying thermocouple arrays at multiple points inside and around the lithium battery module, recording the time it takes for the combustion front to reach each monitoring point, and calculating the fire spread rate (SPD) by the distance between the points and the time difference.

3. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, The process of determining the indicator weight vector W includes: A hierarchical structure model of the analytic hierarchy process is constructed, in which the target layer is used to characterize the evaluation target of the hazard of lithium battery fire, the criteria layer sets the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP and the fire spread rate index SPD, and the index layer is used to carry the statistical characteristic parameters corresponding to each criterion layer. Based on historical lithium battery fire accident database and standardized fire test samples, the relative importance of each indicator in the criterion layer is compared and evaluated. A pairwise comparison matrix is ​​constructed according to the preset scaling rules, so that the matrix elements reflect the relative influence of the corresponding indicators in fire catastrophicity. The pairwise comparison matrix is ​​subjected to eigenvector solving, and the eigenvector corresponding to the largest eigenvalue is extracted as the initial weight parameters of each index. The pairwise comparison matrix is ​​then subjected to a consistency check. When the consistency check result meets the preset consistency condition, the weight parameters are confirmed to be valid.

4. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, The weight parameters that pass the consistency test are processed to form a weight vector W, where the weight wi corresponding to the i-th index is obtained according to the following relationship: Wherein, the weight parameter wi represents the final weight coefficient corresponding to the i-th catastrophicity evaluation index; the initial weight parameter ai represents the importance component of the i-th index obtained by solving the maximum eigenvalue of the pairwise comparison matrix; and an represents the summation of the initial weight parameters of all indicators.

5. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, The calculation of the weighted contribution value includes: For the heat release rate index QHR, smoke toxicity index TOX, explosion potential index EXP, and fire spread rate index SPD, their original statistical characteristic values ​​are extracted from fire test samples. The original statistical characteristic values ​​are physical statistics. The original statistical characteristic values ​​of each indicator are processed by interval mapping to map them to a unified indicator score interval, resulting in the corresponding standardized indicator scores S_QHR, S_TOX, S_EXP and S_SPD. The weight vector W obtained after the consistency test of the analytic hierarchy process is weighted and calculated with the corresponding index scores to form a weighted contribution value.

6. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, The fire hazard quantification scoring model includes a scoring network structure model MOD consisting of an input layer, multiple feature processing layers, and a scoring output layer, where: The input layer contains four input nodes, which correspond to the weighted contribution values ​​of the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD, respectively. The feature processing layer includes a first processing layer, a second processing layer, and a third processing layer. The first processing layer contains 64 nodes, the second processing layer contains 32 nodes, and the third processing layer contains 16 nodes. Each processing layer uses the ReLU activation function to perform non-linear mapping on the input data. A BatchNormalization layer is set after the second processing layer to improve the stability of model training; a Dropout layer with a Dropout rate of 0.2 is set after the third processing layer to prevent the scoring model from overfitting. The scoring output layer contains one output node, which is used to output the fire hazard score (HRS).

7. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, The weighted feature vector undergoes linear transformation and nonlinear mapping layer by layer through the first, second, and third feature processing layers. After the second feature processing layer, normalization is applied to stabilize the feature distribution, and after the third feature processing layer, random inactivation is used to suppress feature co-adaptation. Finally, the fire hazard score is output by the scoring output layer, calculated using the following formula: The comprehensive feature input vector X consists of the weighted contribution values ​​corresponding to the heat release rate index QHR, the smoke toxicity index TOX, the explosion potential index EXP, and the fire spread rate index SPD; W1, W2, and W3 represent the weight matrices corresponding to each feature processing layer, and b1, b2, and b3 are bias terms; BN represents the batch normalization process; f2, f3, and f0 all represent nonlinear mapping operations using the ReLU activation function.

8. The quantitative classification and assessment method for the fire hazard of a lithium battery system according to claim 1, characterized in that, Output the fire hazard level, including: Based on the Fire Hazard Score (HRS), a grade mapping function (GRF) is constructed. The GRF divides the range according to the statistical distribution characteristics of HRS values ​​in historical lithium battery fire samples, and sets multiple range thresholds, including: low risk level (L1, 0.00≤HRS<0.25), medium risk level (L2, 0.25≤HRS<0.50), second-high risk level (L3, 0.50≤HRS<0.75), and high risk level (L4, 0.75≤HRS≤1.00). The fire score (HRS) of the target fire sample is input into the rating function (GRF) for interval judgment, and the corresponding rating label (Label) is output.

Images