Energy Storage Thermal Runaway Damage Assessment Method and Device Based on Energy Storage Safety Risk and Firefighting Efficiency

Abstract

This invention discloses a method and device for assessing thermal runaway losses in energy storage based on energy storage safety risks and firefighting effectiveness. The method first constructs a multi-level Bayesian network based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems. After initialization with multi-source prior knowledge, a conditional probability table is generated using a gradient reinforcement algorithm. Limited key objective information from on-site investigation is injected as evidence, and a variable elimination algorithm is used to estimate the probability of thermal runaway in a single compartment. Then, combined with the structural characteristics of the energy storage system, a multi-level loss reduction index system is constructed. Relevant loss reduction coefficients are calculated through various models, and after integration, the thermal runaway loss ratio is obtained according to the differences in protection configuration. Finally, the overall thermal runaway probability is calculated, coupled and integrated with the loss ratio, and the expected loss ratio of thermal runaway at the site level is output to complete the loss assessment. This invention solves the problems of existing technologies, such as the lack of tools for quantifying thermal runaway probability, the absence of a protection-loss correlation model in firefighting loss assessment, and the separation of accident probability and loss assessment in premium pricing.

Description

Technical Field

[0001] This invention relates to the field of electrochemical energy storage system safety technology, specifically to a method and device for assessing thermal runaway damage in energy storage based on energy storage safety risks and fire-fighting efficiency. Background Technology

[0002] Electrochemical energy storage, as a core regulating unit of new power systems, is widely used in scenarios such as grid peak shaving and distributed energy support. To reduce the operational risks of energy storage projects, the insurance industry has gradually launched dedicated energy storage insurance products, with annual premiums reaching billions of yuan. Risk assessment and loss determination technologies have become core tools supporting the implementation of insurance business.

[0003] However, existing technologies have significant bottlenecks: on the one hand, risk assessment in the underwriting stage relies heavily on subjective experience or single parameters, lacking quantitative models based on system structure and failure mechanisms, making it difficult to quickly and accurately assess the probability of thermal runaway with limited information; on the other hand, fire damage assessment only involves qualitative verification of equipment configuration, without establishing a quantitative correlation between detection and fire extinguishing systems and loss reduction, resulting in a disconnect between the assessment of "accident probability" and "loss severity," failing to provide objective technical basis for premium pricing and loss sharing, and restricting the standardization and large-scale development of energy storage insurance.

[0004] Therefore, there is an urgent need for a thermal runaway loss assessment method for energy storage based on energy storage safety risks and fire protection efficiency, in order to solve the problems of existing technologies such as the lack of tools for quantifying the probability of thermal runaway, the lack of protection-loss correlation models in fire protection loss assessment, and the separation of accident probability and loss assessment in premium pricing. Summary of the Invention

[0005] To address these issues, this invention provides a method and apparatus for assessing thermal runaway in energy storage based on energy storage safety risks and fire-fighting effectiveness. This method solves problems such as the lack of quantitative tools for thermal runaway probability under limited information during the underwriting stage of energy storage insurance, the absence of a protection-loss correlation model in fire-fighting loss assessment, and the separation of accident probability and loss assessment in premium pricing.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the damage caused by thermal runaway in energy storage based on energy storage safety risks and fire-fighting efficiency, characterized in that it includes: Based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems, a multi-level Bayesian network is constructed. The network parameters of the multi-level Bayesian network are initialized based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network. For the non-root nodes in the initialized multi-level Bayesian network, a conditional probability table is generated using a gradient reinforcement algorithm. Based on the conditional probability table, limited key objective information obtained from on-site investigation is injected as evidence through an interface. Based on this evidence, the posterior probability of thermal runaway is calculated using a variable elimination algorithm to obtain the probability of thermal runaway in a single compartment. Based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability, a multi-level loss reduction index system is constructed, including inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. The inter-compartment isolation loss reduction coefficient is calculated using an inter-compartment isolation performance evaluation model. The comprehensive detection loss reduction coefficient for the compartment and pack levels is calculated using a detection performance weighted scoring model. The comprehensive fire suppression loss reduction coefficient for the compartment and pack levels is calculated using a fire suppression performance weighted scoring model. Based on the comprehensive detection loss reduction coefficient and the comprehensive fire suppression loss reduction coefficient, the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient are calculated. The inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient are integrated using a series model. Differential calculation methods are adopted based on whether pack-level and compartment-level protection are provided to calculate the thermal runaway loss ratio. Based on the single-compartment thermal runaway probability, the overall thermal runaway probability is calculated; the overall thermal runaway probability and the thermal runaway loss ratio are dynamically integrated through a coupling algorithm to output the expected loss ratio of thermal runaway at the site level, thus completing the energy storage thermal runaway loss assessment.

[0007] As a preferred solution for energy storage thermal runaway damage assessment method based on energy storage safety risks and fire protection efficiency, the multi-level Bayesian network includes: a data acquisition and environmental parameter layer, a primary fault layer, a secondary fault propagation layer, a safety boundary layer, and a fault assessment layer; the multi-source prior knowledge includes: domain expert experience, historical fault statistics, equipment reliability models, and simulation data.

[0008] As a preferred embodiment of the energy storage thermal runaway loss assessment method based on energy storage safety risks and fire-fighting efficiency, the formula for calculating the conditional probability is:

[0009] In the formula, This represents the conditional probability that the node is in an abnormal state. and These are the preset minimum and maximum probability baseline values ​​for the level to which the node belongs; It is a nonlinear gradient factor; This represents the number of parent nodes that are in an abnormal state. This represents the total number of parent nodes; This is the sum of the influence weights of the parent node in the abnormal state; This represents the maximum sum of the weights of the parent nodes. The target is a non-root node; , , , These are all parameters configured differently based on node type and importance.

[0010] As a preferred solution for energy storage thermal runaway loss assessment based on energy storage safety risks and fire-fighting efficiency, in the process of calculating the posterior probability of thermal runaway through the variable elimination algorithm, based on Bayes' theorem, the joint probability distribution of the network is decomposed into the product of the conditional probabilities of each node, and non-query variables are eliminated through marginal summation to obtain unnormalized probability factors; the posterior probability of thermal runaway is obtained by normalizing the unnormalized probability factors. The formula for calculating the posterior probability of thermal runaway is:

[0011]

[0012] In the formula, For the collection of evidence E Under the condition of querying variables Q Values q The posterior probability; For querying variables; The set of evidence to be injected; This is an unnormalized probability factor containing only query variables and evidence; This is a normalization constant; For querying variables Q Possible values ​​for .

[0013] As a preferred solution for energy storage thermal runaway loss assessment methods based on energy storage safety risks and fire-fighting efficiency, in the process of calculating the inter-cabin isolation loss reduction coefficient, when the actual isolation distance... ≥ When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0014] In the formula, This is the inter-cabin isolation reduction factor; This represents the total number of battery compartments across the entire station. when ≤ < When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0015]

[0016] when hour: =0 In the formula, To normalize the distance progress; Recommended safe distance according to national standards; The minimum safe distance is mandated by national standards; This refers to the actual isolation distance.

[0017] As a preferred method for assessing thermal runaway damage in energy storage based on energy storage safety risks and fire-fighting efficiency, the calculation formula for the comprehensive detection and damage reduction coefficient for both the cabin and pack levels is as follows:

[0018] In the formula, This is the overall performance coefficient for detection; The coefficient of performance is the response speed. The reliability performance coefficient; The life performance coefficient; The formula for calculating the overall fire suppression loss reduction factor for cabin and pack classes is as follows:

[0019] In the formula, This is the comprehensive coefficient for fire extinguishing performance; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the loss reduction coefficient of the cabin class system is as follows:

[0020] In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the cabin-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is as follows:

[0021] In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system.

[0022] As a preferred embodiment of the energy storage thermal runaway loss assessment method based on energy storage safety risks and fire-fighting efficiency, the expression for the thermal runaway loss ratio calculated through the series model is as follows: In the formula, This represents the proportion of thermal runaway losses.

[0023] As a preferred embodiment of the energy storage thermal runaway loss assessment method based on energy storage safety risks and fire fighting efficiency, the formula for calculating the probability of thermal runaway across the entire station is as follows:

[0024]

[0025] In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. For the first k Probability of thermal runaway in each battery compartment; The formula for calculating the expected loss ratio of the site-level thermal runaway is as follows:

[0026] In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

[0027] This invention also provides an energy storage thermal runaway damage assessment device based on energy storage safety risks and fire-fighting efficiency, employing the above-mentioned energy storage thermal runaway damage assessment method based on energy storage safety risks and fire-fighting efficiency, including: The single-compartment thermal runaway probability assessment module is used to construct a multi-level Bayesian network based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems; initialize the network parameters of the multi-level Bayesian network based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; generate a conditional probability table for the non-root nodes in the initialized multi-level Bayesian network using a gradient reinforcement algorithm; inject limited key objective information obtained from on-site investigation as evidence based on the conditional probability table; and calculate the posterior probability of thermal runaway using a variable elimination algorithm based on the evidence to obtain the single-compartment thermal runaway probability. The thermal runaway loss ratio acquisition module is used to construct a multi-level loss reduction index system based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability. This system includes inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. It calculates the inter-compartment isolation loss reduction coefficient using an inter-compartment isolation performance evaluation model; it calculates the comprehensive detection loss reduction coefficient for both compartment and pack levels using a detection performance weighted scoring model; it calculates the comprehensive fire suppression loss reduction coefficient for both compartment and pack levels using a fire suppression performance weighted scoring model; based on the comprehensive detection loss reduction coefficient and the comprehensive fire suppression loss reduction coefficient, it calculates the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient; and it integrates the inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient using a series model, employing differentiated calculation methods based on whether pack-level and compartment-level protection are provided, to calculate the thermal runaway loss ratio. The station-level thermal runaway expected loss ratio acquisition module is used to calculate the station-wide thermal runaway probability based on the single-compartment thermal runaway probability; the station-wide thermal runaway probability and the thermal runaway loss ratio are dynamically integrated through a coupling algorithm to output the station-level thermal runaway expected loss ratio, thus completing the energy storage thermal runaway loss assessment.

[0028] As a preferred solution for an energy storage thermal runaway damage assessment device based on energy storage safety risks and fire protection efficiency, the multi-level Bayesian network in the single-compartment thermal runaway probability assessment module includes: a data acquisition and environmental parameter layer, a primary fault layer, a secondary fault propagation layer, a safety boundary layer, and a fault assessment layer; the multi-source prior knowledge includes: domain expert experience, historical fault statistics, equipment reliability models, and simulation data.

[0029] As a preferred embodiment of an energy storage thermal runaway damage assessment device based on energy storage safety risks and fire-fighting efficiency, the calculation formula for the conditional probability in the single-compartment thermal runaway probability assessment module is as follows:

[0030]

[0031] In the formula, This represents the conditional probability that the node is in an abnormal state. and These are the preset minimum and maximum probability baseline values ​​for the level to which the node belongs; It is a nonlinear gradient factor; This represents the number of parent nodes that are in an abnormal state. This represents the total number of parent nodes; This is the sum of the influence weights of the parent node in the abnormal state; This represents the maximum sum of the weights of the parent nodes. The target is a non-root node; , , , These are all parameters configured differently based on node type and importance.

[0032] As a preferred solution for an energy storage thermal runaway damage assessment device based on energy storage safety risks and fire-fighting efficiency, in the single-compartment thermal runaway probability assessment module, during the process of calculating the posterior probability of thermal runaway through the variable elimination algorithm, based on Bayes' theorem, the network joint probability distribution is decomposed into the product of the conditional probabilities of each node, and non-query variables are eliminated through marginal summation to obtain unnormalized probability factors; the posterior probability of thermal runaway is obtained by normalizing the unnormalized probability factors. The formula for calculating the posterior probability of thermal runaway is:

[0033]

[0034] In the formula, For the collection of evidence E Under the condition of querying variables Q Values q The posterior probability; For querying variables; The set of evidence to be injected; This is an unnormalized probability factor containing only query variables and evidence; This is a normalization constant; For querying variables Q Possible values ​​for .

[0035] As a preferred solution for an energy storage thermal runaway loss assessment device based on energy storage safety risks and fire-fighting efficiency, in the thermal runaway loss ratio acquisition module, during the calculation of the inter-cabin isolation loss reduction coefficient, when the actual isolation distance... ≥ When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0036] In the formula, This is the inter-cabin isolation reduction factor; This represents the total number of battery compartments across the entire station. when ≤ < When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0037]

[0038] when hour: =0 In the formula, To normalize the distance progress; Recommended safe distance according to national standards; The minimum safe distance is mandated by national standards; This refers to the actual isolation distance.

[0039] As a preferred embodiment of an energy storage thermal runaway loss assessment device based on energy storage safety risks and fire-fighting efficiency, the calculation formula for the comprehensive detection and loss reduction coefficient of the cabin level and Pack level in the thermal runaway loss ratio acquisition module is as follows:

[0040] In the formula, This is the overall performance coefficient for detection; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the overall fire suppression loss reduction factor for cabin and pack classes is as follows:

[0041] In the formula, This is the comprehensive coefficient for fire extinguishing performance; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the loss reduction coefficient of the cabin class system is as follows:

[0042] In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the cabin-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is as follows:

[0043] In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system.

[0044] As a preferred embodiment of an energy storage thermal runaway loss assessment device based on energy storage safety risks and fire-fighting efficiency, the expression for the thermal runaway loss ratio obtained by the thermal runaway loss ratio acquisition module through the series model is as follows:

[0045] In the formula, This represents the proportion of thermal runaway losses.

[0046] As a preferred embodiment of an energy storage thermal runaway loss assessment device based on energy storage safety risks and fire-fighting efficiency, the calculation formula for the overall thermal runaway probability in the site-level thermal runaway expected loss ratio acquisition module is as follows:

[0047]

[0048] In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. For the first k Probability of thermal runaway in each battery compartment; The formula for calculating the expected loss ratio of the site-level thermal runaway is as follows:

[0049] In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

[0050] The present invention has the following advantages: First, it achieves accurate quantification of thermal runaway probability under limited information. Through multi-level Bayesian networks and gradient reinforcement algorithms, it solves the risk assessment problem when information is insufficient during the underwriting stage.

[0051] Second, a quantitative correlation between fire protection and loss reduction has been established. Based on a multi-dimensional loss reduction model, the actual loss reduction value of different protection configurations can be accurately calculated.

[0052] Third, it has established a connection between probability and loss, providing a unified quantitative benchmark for standardized underwriting and scientific pricing of energy storage insurance, and improving the engineering practicality and industry adaptability of the technical solution. Attached Figure Description

[0053] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0054] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0055] Figure 1 This is a flowchart illustrating the energy storage thermal runaway loss assessment method based on energy storage safety risks and fire-fighting efficiency provided in Embodiment 1 of the present invention. Figure 2 This is a schematic diagram of the Bayesian thermal runaway probability network structure in the energy storage thermal runaway loss assessment method based on energy storage safety risk and fire protection efficiency provided in Embodiment 1 of the present invention. Figure 3 This is a schematic diagram of the inter-compartment isolation loss function setting in the energy storage thermal runaway loss assessment method based on energy storage safety risk and fire protection efficiency provided in Embodiment 1 of the present invention. Figure 4 This is a schematic diagram of the architecture of the energy storage thermal runaway damage assessment device based on energy storage safety risks and fire protection efficiency provided in Embodiment 2 of the present invention. Detailed Implementation

[0056] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0057] See Figure 1 Embodiment 1 of the present invention provides a method for assessing the thermal runaway of energy storage based on energy storage safety risks and fire-fighting efficiency, comprising the following steps: S1. Based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems, a multi-level Bayesian network is constructed; the network parameters of the multi-level Bayesian network are initialized based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; for the non-root nodes in the initialized multi-level Bayesian network, a conditional probability table is generated through a gradient reinforcement algorithm; based on the conditional probability table, limited key objective information obtained from on-site investigation is injected as evidence through an interface; based on the evidence, the posterior probability of thermal runaway is calculated through a variable elimination algorithm to obtain the single-compartment thermal runaway probability; S2. Based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability, a multi-level loss reduction index system is constructed, including inter-compartment isolation performance, compartment-level detection and fire extinguishing capabilities, and pack-level detection and fire extinguishing capabilities. The inter-compartment isolation loss reduction coefficient is calculated using an inter-compartment isolation performance evaluation model. The comprehensive detection loss reduction coefficient for the compartment and pack levels is calculated using a detection performance weighted scoring model. The comprehensive fire extinguishing loss reduction coefficient for the compartment and pack levels is calculated using a fire extinguishing performance weighted scoring model. Based on the comprehensive detection loss reduction coefficient and the comprehensive fire extinguishing loss reduction coefficient, the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient are calculated. The inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient are integrated using a series model. Differential calculation methods are adopted based on whether pack-level and compartment-level protection are provided to calculate the thermal runaway loss ratio. S3. Based on the single-compartment thermal runaway probability, calculate the overall thermal runaway probability; dynamically integrate the overall thermal runaway probability and the thermal runaway loss ratio through a coupling algorithm, output the expected loss ratio of thermal runaway at the site level, and complete the energy storage thermal runaway loss assessment.

[0058] In this embodiment, in step S1, a multi-level Bayesian network is constructed based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems; the network parameters of the multi-level Bayesian network are initialized based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; for the non-root nodes in the initialized multi-level Bayesian network, a conditional probability table is generated using a gradient reinforcement algorithm; based on the conditional probability table, limited key objective information obtained from on-site investigation is injected as evidence through an interface; based on the evidence, the posterior probability of thermal runaway is calculated using a variable elimination algorithm to obtain the single-compartment thermal runaway probability.

[0059] Specifically, based on first principles of thermal runaway evolution in electrochemical energy storage systems, and integrating battery failure mechanisms, system operation logic, and domain expert knowledge, a causal model of the fault propagation path is constructed, and a Bayesian network is built. , where nodes This indicates multiple environmental conditions or fault types that lead to thermal runaway (such as "low temperature environment," "DC internal resistance exceeding the limit," "inconsistent SOC"), etc. Indicates the fault propagation path, parameters This is a conditional probability table.

[0060] Among them, such as Figure 2 As shown, the nodes are divided into the following physical levels: Level 1 (Data Acquisition and Environmental Parameters Layer): This layer contains 14 root nodes, representing the most basic fault causes that trigger thermal runaway.

[0061] Node list: Manufacturing process, low temperature environment, high temperature environment, temperature difference, SOC assessment inaccuracy, voltage equalization failure, BCU failure, NTC failure, voltage sensor failure, cooling system failure, uneven cooling system, voltage acquisition failure, current acquisition failure, temperature acquisition failure.

[0062] Level 2 (Battery Fault Layer): This level contains 2 nodes, representing abnormalities in key battery parameters caused by fundamental faults.

[0063] Node list: DC internal resistance exceeding limit, SOH exceeding limit.

[0064] Level 3 (Secondary Fault Transmission Layer): This layer contains 2 nodes, representing an imbalance in the operational state at the system level.

[0065] Node list: Inconsistent SOC, inconsistent voltage.

[0066] Level 4 (Safety Boundary Layer): This level contains 5 nodes and is a direct precursor to thermal runaway, representing a breach of critical safety boundaries.

[0067] Node list: SOC exceeding limit, charge / discharge cutoff voltage exceeding limit, cluster current exceeding limit, leakage current exceeding limit, maximum cell temperature exceeding limit.

[0068] Level 5 (Fault Assessment Layer): This layer is the final output node of the network.

[0069] Node list: Thermal runaway.

[0070] In this embodiment, after constructing the Bayesian network, domain expert experience, historical fault statistics, equipment reliability models, and simulation data are integrated to build initial weight rules and basic probability ranges for each non-root node. The weight rules are used to quantify the influence of different parent nodes on the state of child nodes. For example, an initial rule is set for the node "DC internal resistance exceeding limit": {"manufacturing process": 0.2, "low temperature environment": 0.3, "high temperature environment": 0.1, "temperature difference": 0.4}, to reflect the relative importance of each factor. Simultaneously, based on the failure frequency statistics of the node's level, a priori probability range for each level is set to ensure that the network probability magnitude conforms to engineering practice.

[0071] After initializing the Bayesian network, an initial conditional probability table is generated using a gradient reinforcement algorithm. The algorithm iterates through all possible combinations of parent node states for each non-root node, and for any given combination, calculates the total number of parent nodes in the "True" state. and the sum of their corresponding weights By introducing a nonlinear gradient factor function to dynamically calculate the conditional probability that the child node is "True" under this parent node combination, the nonlinear gradient factor function is defined as follows:

[0072] Child nodes under parent node combination The conditional probability of "True" is defined as follows:

[0073] In the above formula, This represents the conditional probability that the node is in an abnormal state. and These are the preset minimum and maximum probability baseline values ​​for the level to which the node belongs; It is a nonlinear gradient factor; This represents the number of parent nodes that are in an abnormal state. This represents the total number of parent nodes; This is the sum of the influence weights of the parent node in the abnormal state; This represents the maximum sum of the weights of the parent nodes. The target is a non-root node; , , , These are all parameters configured differently based on node type and importance.

[0074] In this embodiment, after generating the conditional probability table using the gradient reinforcement algorithm, real-time evidence is injected through an interface to trigger probability updates. In engineering practice and simulation analysis, operators can input known or hypothetical system states as deterministic evidence into the model through a specified interface. This process is called "evidence injection," and its purpose is to combine the model's prior general assessment with specific safety assessment and analysis results, focusing on quantitative analysis under specific failure scenarios. After injecting evidence, the model needs to recalculate the probability distribution of all nodes in the network. The core objective is to solve the probability distribution under given evidence conditions. Below, among them Represents a specific observed state of a node; query variable. That is, the posterior probability of the occurrence of the top-level event "thermal runaway". To achieve this goal, this invention employs a variable elimination algorithm for efficient and accurate probabilistic reasoning. The specific reasoning process is as follows: T1. The algorithm is derived based on Bayes' theorem and the joint probability distribution of the network. The joint probability distribution of a Bayesian network can be decomposed into the product of the conditional probabilities of each node: In the formula, For the first in the network One node; Its parent node set; Define the conditional probability table (CPT).

[0075] T2. By systematically summing and integrating the non-query variables, these variables are gradually "eliminated." Mathematically, this "elimination" process is equivalent to marginally summing the probabilities corresponding to all possible states of the intermediate variables and calculating the posterior probability. The core is to apply Bayes' theorem and perform marginalization:

[0076] In the formula, For network variables other than query variables and evidence variables All non-query variables, i.e., intermediate variables, except for those in the query; The normalization constant is usually denoted as . .

[0077] T3. To efficiently calculate the above formula and avoid [further details needed], The algorithm exhaustively sums all possible combinations, and the variable elimination algorithm follows a predefined order. Stepwise eliminate non-query variables. For a variable to be eliminated... The algorithm performs the following marginal summation operation:

[0078] In the formula, For the current and variable The set of all relevant probability factors. This operation multiplies the factor pairs. All possible states Summing, thus generating a new factor. This factor no longer depends on the variable. ,Right now It has been "eliminated".

[0079] T4. Repeat step T3 until all non-query variables are resolved. They are eliminated one by one. Ultimately, the program will obtain only information about the query variable. and evidence Unnormalized probability factor .

[0080] The posterior probability is calculated through normalization:

[0081]

[0082] In the formula, For the collection of evidence E Under the condition of querying variables Q Values q The posterior probability; For querying variables; The set of evidence to be injected; This is an unnormalized probability factor containing only query variables and evidence; This is a normalization constant; For querying variables Q Possible values ​​for .

[0083] By following the steps above, computational complexity can be significantly reduced while maintaining computational accuracy, avoiding an exhaustive traversal of all possible combinations of node states.

[0084] In this embodiment, in step S2, based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability, a multi-level loss reduction index system is constructed, including inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. The inter-compartment isolation loss reduction coefficient is calculated using an inter-compartment isolation performance evaluation model. The combined detection loss reduction coefficient for the compartment and pack levels is calculated using a detection performance weighted scoring model. The combined fire suppression loss reduction coefficient for the compartment and pack levels is calculated using a fire suppression performance weighted scoring model. Based on the combined detection loss reduction coefficient and the combined fire suppression loss reduction coefficient, the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient are calculated. The inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient are integrated using a series model. Differential calculation methods are adopted based on whether pack-level and compartment-level protection are provided to calculate the thermal runaway loss ratio.

[0085] Specifically, based on the physical structural characteristics of the energy storage system, a fire-fighting performance evaluation index system with three dimensions is established. The three dimensions are compartment isolation performance, compartment-level detection and fire suppression capability, and pack-level detection and fire suppression capability.

[0086] The inter-compartment isolation performance dimension aims to quantify the effect of physical isolation between battery compartments on suppressing fire spread. The assessment is based on the safety distance requirements specified in the national standard GB51048-2014. A nonlinear mathematical model of distance-loss reduction is established to map the actual isolation distance to the degree of loss reduction at the isolation level. This coefficient characterizes the degree of loss reduction in the event of a thermal runaway fire in a single battery compartment, preventing the spread to adjacent compartments due to inter-compartmental protection. Its calculation comprehensively considers the actual distance, the safety distance threshold corresponding to the battery type, and the attenuation pattern when the distance is insufficient, thus transforming the static design parameter of spatial layout into a dynamic and quantifiable risk mitigation indicator.

[0087] The compartment-level detection and fire suppression capability dimension is used to assess the overall fire early warning and initial emergency response capabilities of the battery compartment. A three-pronged scoring system integrating response performance, reliability, and timeliness is constructed to quantify the effectiveness of various detectors (including heat, smoke, and gas detectors) and fire suppression equipment (including gas, dry powder, and sprinkler systems) installed within the compartment. Based on a comparison of actual equipment parameters (including response speed, failure rate, and lifespan) with national standards, the percentage of fire damage reduction at the compartment level is calculated. This coefficient reflects the comprehensive ability of the compartment-level fire suppression system to promptly detect and suppress fires within the compartment and prevent the spread of disasters.

[0088] The Pack-level detection and extinguishing capability dimension employs differentiated loss calculation models based on whether independent detection and extinguishing systems are set up at the Pack level, to more accurately reflect the risk level under different protection strategies: When no Pack-level protection is set up, fire risk mainly relies on compartment-level and higher-level protection measures for control; when Pack-level protection is set up, the system has a more refined initial fire suppression capability, taking into account the scale effect brought about by the internal structure of the battery pack and the number of cells. A scale correction factor is introduced, incorporating the number of cells in the Pack into the evaluation model, and adjusting the evaluation weights for the characteristic that the Pack interior is usually mainly for temperature detection. The scoring of each device is still based on the comparison between the actual parameters of the device and the national standard requirements, and finally outputs the Pack-level fire loss reduction percentage. This coefficient characterizes whether the built-in detection and fire suppression devices can effectively suppress the spread of the fault between packs when thermal runaway occurs inside the pack, thereby achieving a more basic and precise protection effect. The closer the value is to 1, the stronger the ability of the pack-level protection system to control the fire within the pack and prevent it from spreading to the compartment level, and the lower the proportion of losses suffered by the entire station due to thermal runaway of that pack.

[0089] In this embodiment, the process of calculating the inter-cabin isolation reduction coefficient using the inter-cabin isolation performance evaluation model is as follows: When the actual isolation distance between battery compartments is greater than or equal to the recommended distance specified by national standards or material specifications, it is determined that physical isolation can completely prevent the spread of fire. In this case, the loss is strictly limited to the single battery compartment where the fire initially started. The degree of reduction in actual loss compared to the total site damage can be quantified as follows:

[0090] In the formula, This is the inter-cabin isolation reduction factor; This represents the total number of battery compartments across the entire station. When the actual isolation distance falls between the national standard's mandatory minimum safe distance and the recommended distance, the isolation effect is deemed incomplete, and there is a certain risk of fire spreading. In this situation, the loss reduction calculation involves the product of two parts:

[0091]

[0092] When the actual isolation distance is less than the national standard's mandatory minimum safe distance, the isolation effect is determined to be 0, and the loss reduction in this case... =0.

[0093] In the formula, To normalize the distance progress; Recommended safe distance according to national standards; The minimum safe distance is mandated by national standards; This refers to the actual isolation distance.

[0094] As shown in Table 1, the "Design Code for Electrochemical Energy Storage Power Stations" (GB 51048-2014) clearly states that different electrochemical energy storage batteries have different fire hazard classifications. Lead-acid batteries, lithium-ion batteries, and flow batteries belong to Class E fire hazard, while sodium-sulfur batteries belong to Class A fire hazard. The code also stipulates that when designing energy storage power stations, the fire hazard category must be determined based on the battery type, and the corresponding legally mandated safety distances must be strictly enforced. For example, for outdoor lithium-ion battery installations with a fire hazard classification of Class E, the fire separation distance between them and Class C, D, and E production buildings with fire resistance ratings of Class I and II should be no less than 10 meters; while for outdoor sodium-sulfur battery installations with a fire hazard classification of Class A, the fire separation distance between them and similar buildings must be increased to 12 meters. These specific distance values ​​constitute the threshold reference benchmarks in the compartment isolation performance model.

[0095]

[0096] Table 1. Design Specifications for Electrochemical Energy Storage Power Stations specify safe isolation distances for different battery types. Based on the specifications in the table, set the parameters. In the formula rice, The selection should be based on the specific battery materials used at the actual site. For example, when lithium-ion batteries are used at the site, the strictest safety distance from various buildings is 12 meters. Set to 12 meters.

[0097] When the actual distance equals When p=0, the attenuation factor A value of 0 means that the isolation is completely ineffective, and the loss reduction is 0.

[0098] When the actual distance equals When p=1, the attenuation factor is 1, meaning the isolation is completely effective. The formula degenerates into the "completely effective isolation" case, and the degree of reduction in total loss compared to total loss is:

[0099] When the actual distance is arrive Between, the model through This quadratic term is used to simulate the accelerated process of risk decay. This means that as the distance increases from near the minimum safe distance, the improvement in protective effectiveness is slow in the early stages and accelerates in the later stages. This is consistent with the decay law of fire heat radiation and flying fire risk with increasing distance, reflecting the increasing marginal benefits of engineering protection.

[0100] when Set to 5 meters, When set to 12 meters, the isolation distance and loss reduction rate Relationship such as Figure 3 As shown.

[0101] In this embodiment, the comprehensive detection loss reduction coefficient for both cabin-level and Pack-level detection is calculated using a detection performance weighted scoring model. The specific calculation process is as follows: A comprehensive quantitative evaluation model is established to assess the detection capabilities of fire suppression systems in energy storage power stations. This model is based on detector performance coefficients. Relative response speed performance coefficient ( Reliability performance coefficient and life performance coefficient The weighted scoring formula is as follows:

[0102] In the formula, This is the overall performance coefficient for detection; The coefficient of performance is the response speed. The reliability performance coefficient; This represents the lifespan performance coefficient.

[0103] In the detector damage assessment model, the weighting is based on the priorities of fire safety engineering and actual risk prevention and control needs. Response speed performance coefficient. It is given the highest weight because the timeliness of early fire detection is crucial: the earlier the detector detects a fire, the more time it buys for firefighting operations, thus significantly reducing the scale of damage. Lifetime performance coefficient The lowest weight is assigned because while the long-term durability of the detector affects maintenance costs and service life, its contribution to mitigating damage in a single fire event is relatively indirect compared to immediate response and reliable operation. As long as performance meets standards during the critical service period, the marginal impact on lifespan is weak. This weighting ensures the model focuses on the two core functions of "rapid response" and "reliable operation," while also considering long-term economic efficiency, aligning with the principle of "time priority and reliability as the foundation" in fire safety.

[0104] Among them, the response speed performance coefficient The calculation formula is:

[0105] In the formula, k The detectors include three types: gas, temperature, and smoke detectors. For type k The detector's actual response time / response threshold; For type k Detector performance standard reference values; For importance weights.

[0106] The output of the above formula is a comprehensive score between 0 and 1. The higher the score, the more significant the overall improvement in the actual average response performance of the entire detection system compared to the national standard requirements, meaning a faster response and better performance.

[0107] Referring to national standards, gas detectors and temperature detectors and The response time of the corresponding detector is the ratio of the response threshold of the smoke detector under different smoke flow rates.

[0108] For gas detectors, the GB15322 standard for combustible gas detectors stipulates that the response time of carbon monoxide detectors should not exceed 60 seconds, and that of other combustible gas detectors should not exceed 30 seconds.

[0109] For heat detectors, according to GB4716-2024 "Point-type Heat Detectors", the upper limit of their response time is 145 seconds at a standard heating rate of 30℃ / min.

[0110] For smoke detectors, according to GB4715-2024 "Point-type Smoke Detectors", the ratio of the response threshold of the detector at different airflow velocities of 0.2 m / s and 1.0 m / s should not be greater than 1.6.

[0111] Based on these three factors, the corresponding response time performance references for different types of gas detectors can be determined. .

[0112] In this embodiment, the lifetime performance coefficient The calculation formula is:

[0113] In the formula, For type The actual lifespan of the detector; For type Detector performance standard reference values.

[0114] The output of the above formula is a comprehensive score between 0 and 1. The higher the score, the more significant the improvement in the actual lifespan of the entire detection system compared to the national standard requirements.

[0115] According to the national standard GB 29837-2013 "Maintenance and Scrapping of Fire Detection and Alarm Products", the original text of the standard stipulates that the service life of fire detection and alarm products generally does not exceed 12 years, and it is clearly stated that the service life of gas-sensitive elements in combustible gas detectors does not exceed 5 years.

[0116] For gas detectors, based on national standards and industry experience (such as catalytic, electrochemical, and semiconductor types, whose lifespan is mostly 3-5 years), the standard lifespan is set at 5 years.

[0117] Regarding smoke detectors, State-owned Factory 262, a major manufacturer of large-scale automatic fire alarm systems in my country, conducted batch and continuous operation tests on the FJ-2701 smoke detector. Theoretical calculations showed that the average mean time between failures (MTBF) of the FJ-2701 was 34 years. The Swiss company CER-BERUS, using the F910 fire detector as an example, considered the MTBF to be 30 years. Based on this, the standard lifespan of smoke detectors is set at 30 years.

[0118] The standard does not specify a separate lifespan for temperature detectors, and there is no authoritative research report in the industry. Therefore, the standard lifespan of fire detection and alarm products in the national standard is set at 12 years.

[0119] Based on these three factors, the lifespan performance references for different types of gas detectors can be determined. .

[0120] In this embodiment, the reliability performance coefficient The calculation process is as follows: Based on the product lifecycle failure rate theory (bathtub curve), this describes the operational... t Time-dependent component failure period with fixed failure rate λ The failure model below is a negative exponential model:

[0121] In the formula, λAs a failure rate, it is defined as the total number of components that fail within a unit of test time.

[0122]

[0123] Mean Time To Life (MTTF) is defined as the average lifespan of each product.

[0124] When the testing period is extended to infinity, continuing until the lifespan of all products reaches the end, the number of failures equals the total number of products under test. At this point:

[0125] Therefore, it can be approximated as follows: That is, the failure rate is the reciprocal of the average lifespan. At this point, it has been running... t Failure rate of time-related components It can be converted into:

[0126] Therefore, the reliability of a component can be determined by... To measure this, the only variable in this function that needs to be benchmarked is lifespan. Therefore, by benchmarking it against the standard lifespan, The formula can be written as follows:

[0127] Finally, for To ensure the accuracy and scientific validity of the fire damage assessment model for different installation scenarios, this invention is strictly based on the propagation characteristics and response timing differences of different physical signals in specific structures of energy storage systems.

[0128] Specifically, within the relatively open space of the battery compartment, flammable gases released in the early stages of thermal runaway can spread rapidly, and their detection signals typically appear close to or slightly earlier than a significant temperature rise, while smoke formation and diffusion are relatively delayed. Therefore, to prioritize the detection of the earliest signs of fire, the detector weights within the compartment are configured as follows: gas detector: heat detector: smoke detector = 5:4:1, highlighting the paramount importance of gas detection in this scenario.

[0129] Within the compact and relatively sealed battery pack, the localized temperature rise caused by thermal runaway is often extremely rapid and direct. The generation and transmission of temperature signals are slightly faster than the accumulation and diffusion of gas within the limited space of the pack. Therefore, to more accurately reflect the fire development characteristics inside the pack, the weighting configuration is adjusted to: gas detector: heat detector: smoke detector = 4:5:1, to emphasize the priority of temperature detection in this environment.

[0130] In this embodiment, the comprehensive fire extinguishing loss reduction coefficient for both compartment and pack levels is calculated using a fire extinguishing performance weighted scoring model. The specific process is as follows: A comprehensive quantitative evaluation model is established for the performance of the fire suppression system in the fire protection system of an energy storage power station. This model is based on a fire suppression performance coefficient. Relative response speed performance coefficient Reliability performance coefficient and life performance coefficient The weighted scoring formula is as follows:

[0131] In the formula, This is the comprehensive coefficient for fire extinguishing performance; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. Response speed performance coefficient With reliability performance coefficient These two aspects are given equal and highest weight (40% each), stemming from the dual fundamental requirements of "effective intervention" in fire suppression systems: first, timeliness—the system must quickly activate and release the extinguishing agent within the golden timeframe of the initial stage of a fire; any delay will lead to the expansion of the fire and a multiplication of losses; second, certainty—the fire suppression action must be executed reliably and completely, and the system itself must not experience malfunctions such as accidental activation, activation failure, or interruption of spraying. Both are indispensable and together constitute the necessary and sufficient conditions for effective fire suppression, thus requiring equal consideration.

[0132] Life performance coefficient The relatively low weight (20%) is because fire suppression systems, as "one-time use in critical moments" safety devices, have their core value in ensuring functional integrity throughout their design lifespan. As long as performance meets standards within the critical service life, the remaining lifespan has no direct impact on the effectiveness of handling a single fire incident; however, its weight is still retained to reflect the potential risk of degradation in response speed and reliability caused by equipment aging, thereby urging regular replacement and maintenance.

[0133] In this embodiment, The calculation formula is:

[0134] In the formula, These are performance standard reference values ​​for fire suppression systems. The actual response time of the fire suppression system / response threshold; For importance weights.

[0135] For gaseous fire suppression systems, their response and extinguishing performance shall be in accordance with GB 25972-2024 "Gaseous Fire Suppression Systems and Components". This standard stipulates that when using total flooding method for fire suppression, the system nozzles shall bring the protected area to the design extinguishing concentration within 30 seconds after the spraying ends.

[0136] For dry powder fire extinguishing systems, their extinguishing performance is governed by GB 16668, "General Technical Conditions for Dry Powder Fire Extinguishing Systems and Components." This standard specifies clear requirements for total flooding extinguishing performance: during Class A surface fire or Class B fire extinguishing tests, the system should extinguish the fire within 30 seconds after the extinguishing agent is sprayed. For localized fire extinguishing and specific Class D metal fires (such as sodium fires, magnesium fires, and triethylaluminum fires), the standard also requires extinguishing open flames within 30 seconds after spraying and provides specific regulations on subsequent reignition suppression time ranging from minutes to hours.

[0137] For automatic sprinkler systems, their response performance is based on GB 5135.1-2019 "Automatic Sprinkler Systems - Part 1: Sprinkler Heads". This standard uses the response time coefficient (RTI) as the core parameter to measure the sprinkler head's response sensitivity, and the RTI of a standard response sprinkler head should be less than 350. RTI describes the rate at which the sprinkler head's temperature-sensing element responds to a stream of hot smoke from a fire. The lower the RTI value, the faster the sprinkler head responds.

[0138] Based on these three factors, the corresponding response performance references for different fire extinguishers can be determined. .

[0139] In this embodiment, the lifetime performance coefficient The calculation formula is:

[0140] In the formula, For type The actual lifespan of a fire extinguisher; For type Reference values ​​for fire extinguisher performance standards.

[0141] As shown in Table 2, the standard lifespan of various fire extinguishing systems is specified in GB50140-2005 "Code for Design of Fire Extinguisher Configuration in Buildings":

[0142] Table 2 Design Specifications for Fire Extinguisher Configuration in Buildings Based on this standard, the lifespan performance references for different types of gas detectors can be determined. .

[0143] In this embodiment, the reliability performance coefficient The calculation formula is:

[0144] Finally, for To ensure the effectiveness and consistency of the fire damage assessment model in evaluating different levels of protection systems, this invention strictly determines the settings based on the comprehensive performance and reliability of different fire extinguishing media, as well as the safety logic of their linkage with detection signals.

[0145] Specifically, gaseous extinguishing agents are considered the optimal choice due to their high extinguishing efficiency, good electrical insulation, cleanliness with no residue, and minimal impact on battery systems. While water mist extinguishing is superior to dry powder in terms of cooling and asphyxiation effects, it poses a significant short-circuit risk in electrical environments, drastically reducing its reliability. Dry powder extinguishing presents cleaning challenges and the risk of reignition. Therefore, when evaluating the overall effectiveness of a fire extinguishing system, gaseous extinguishing agents, which offer the best damage reduction performance, are given the highest weight, while the other two are given equal weight.

[0146] Both compartment-level and pack-level fire suppression systems employ the same 6:2:2 weighting configuration. The core reason for this is that, despite differences in the internal spatial structure, thermal runaway propagation speed, and signal propagation timing between the battery compartment and the battery pack, the ranking of fire suppression performance is fundamentally determined by the inherent characteristics of the medium itself, not by the structural hierarchy of the protected space. Gas, water mist, and dry powder extinguishing agents exhibit similar performance advantages or disadvantages in both scenarios. Therefore, when quantifying their damage reduction effects, compartments and packs follow the same set of material-defined performance levels, and the weighting configuration need not be differentiated.

[0147] In this embodiment, the loss reduction coefficients for the compartment-level system and the fire extinguishing-level system are calculated based on the detection loss reduction coefficient and the fire extinguishing loss reduction coefficient. The specific calculation process is as follows: The formula for calculating the loss reduction coefficient of the cabin class system is as follows:

[0148] In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the cabin-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is:

[0149] In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system.

[0150] Among them, regarding and Based on the battery safety incident database compiled by the Electric Power Research Institute (EPRI) from 2011 to the present, approximately 55.1% of recorded incidents involved open flames. Considering that the residual value is significantly lower in fires compared to non-fires, we assume a 60% loss rate for incidents without open flames and a 95% loss rate for incidents with open flames. For pack-class batteries, due to their smaller components, the losses under the same fire intensity are more severe; therefore, we assume an 80% loss rate for incidents without open flames and a 100% loss rate for incidents with open flames. From this, we can calculate... and as follows:

[0151]

[0152] for The formula, the basic reduction effect multiplied by the right-hand side includes... , , , The item indicates that a detection / firefighting system with certain performance capabilities will be able to bring about greater damage reduction effects. (Molecular) The actual performance of the detectors and fire suppression systems was quantified. , The positive enhancement of the protective effect; among which, the parameters and As a key weight, it reflects the relative importance of the detection system and the fire suppression system in the overall assessment. Considering that during the damage reduction process, the detectors only play an identification role, while the fire suppression system is responsible for truly performing fire suppression and damage reduction functions, therefore, it will be considered as a key weight. Set to 0.4, Set to 0.6, therefore, when only the detection system ( The degree of reduction (=0) is:

[0153] The degree of damage reduction will be less than that of fire suppression systems alone:

[0154] This demonstrates the difference in loss reduction effectiveness between the two systems; denominator It is a normalization coefficient that ensures the factor is 1 when the performance index is optimal. At this point, it represents the basic risk that still exists even when the system has reached its best protection. The factor is at its minimum when both systems are completely ineffective. This structure ensures that the output value always varies within a reasonable range.

[0155] for The formula calculates the percentage of value that the entire battery compartment (containing numerous packs) can ultimately retain if the pack level is equipped with an independent detection and extinguishing system and successfully activates in a fire incident.

[0156] Due to the basic reduction effect This can represent the proportion of a pack lost when it is completely out of control and burned to the ground in a fire, divided by... This means that under ideal control, the loss from this accident is evenly distributed across every pack in the entire cabin, making the "loss per pack" extremely small.

[0157] Molecular part Based on the control effectiveness coefficient of Pack-level detection and fire suppression devices, in extreme cases, if the Pack-level detection or fire suppression system performs extremely poorly (i.e., and If all values ​​are small, close to 0 or lower, then it will be close to or less than 0. This means that the control effect is somewhat reduced.

[0158] When the system is fully effective (i.e.) and If both are 1), then =1, reaching the maximum value.

[0159] therefore The percentage of damage not under control when Pack-level detection and firefighting measures are effective ( and The larger the value, the smaller it becomes. The smaller the losses, the less likely they are to be contained.

[0160] To ensure logical rigor, when When =1, add to the control effect coefficient This transforms the original expression into:

[0161] To ensure the system remains fully effective, the system's loss reduction effect is set to result in only a very small portion of the loss. The loss is not an absolute zero.

[0162] Calculate in sequence and Afterwards, with Combined, the final loss for the entire site can be assessed as follows:

[0163] In the formula, This represents the proportion of thermal runaway losses.

[0164] In this embodiment, in step S3, the overall thermal runaway probability is calculated based on the single-compartment thermal runaway probability; the overall thermal runaway probability and the thermal runaway loss ratio are dynamically integrated through a coupling algorithm to output the expected loss ratio of thermal runaway at the site level, thus completing the energy storage thermal runaway loss assessment.

[0165] Specifically, the probability of thermal runaway assessed in step S1 is coupled with the fire protection facility loss reduction efficiency quantified in step S2 to calculate a key indicator that can be directly used for insurance pricing and underwriting decisions: the expected loss ratio of station-level thermal runaway.

[0166] The formula for calculating the expected loss ratio of station-level thermal runaway is as follows:

[0167] In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

[0168] Given that the probability of thermal runaway in each compartment is independent, the overall safety of the station is ensured. The probability is calculated by multiplying the probabilities of single-compartment safety by the total probability, while the complement of all station safety events is the probability of thermal runaway occurring within the station, i.e., .

[0169]

[0170]

[0171] In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. No. k The probability of thermal runaway in the battery compartment.

[0172] In engineering applications, if only single-compartment charge / discharge data is available during the underwriting phase, it is assumed that the thermal runaway probability of all compartments in the entire station is the same as that of the sampled compartment. If the battery compartments with the highest and lowest charge / discharge frequencies in the entire station can be obtained, based on the assumption that higher charge / discharge frequencies lead to more severe battery aging and degradation, and thus a higher probability of thermal runaway, the probability of each compartment can be considered... The probability is uniformly distributed between the highest and lowest probabilities, which can be expressed by the formula:

[0173] In the formula, This represents the minimum probability of thermal runaway in the battery compartment. This represents the maximum probability of thermal runaway in the battery compartment.

[0174] In one possible implementation, a verification example is provided as follows: Based on the charge and discharge data of a certain station's single compartment (including cluster current, individual cell voltage, individual cell temperature, etc.), various faults were analyzed and calculated. The station has five major faults: DC internal resistance exceeding the limit, SOC exceeding the limit, SOH exceeding the limit, voltage inconsistency, and uneven cooling of the cooling system.

[0175] M1. Evidence Injection and Network State Initialization: Inject the following 5 pieces of evidence into the model and fix its state as "True". DC internal resistance exceeding limit = True; :SOH exceeding limit = True; Voltage inconsistency = True; Uneven cooling in the cooling system = True; :SOC out of bounds=True.

[0176] M2, Recalculation of the conditional probability of the parent node's influence based on injected evidence: Based on the Bayesian influence network framework, the probabilities of the child nodes of the nodes where these pieces of evidence are located are all affected by the injected evidence.

[0177] Taking the probability of node “leakage current exceeding the limit” as an example, among its four parent nodes, two were confirmed as True by evidence (DC internal resistance exceeding the limit, SOH exceeding the limit), while the other two (manufacturing process, low temperature environment) did not show obvious fault evidence.

[0178] The initial weighting rule for the parent node, based on engineering and expert experience, is {"DC internal resistance exceeding limit": 0.2,"SOH exceeding limit": 0.4,"Manufacturing process": 0.25,"Low temperature environment": 0.15}; the initial probability range for events occurring at this node is 0.001%-0.09%; nonlinear gradient factor. Set to (0.6, 1.2, 0.4, 0.5).

[0179] Based on the input evidence, the algorithm comprehensively calculates the number of parent nodes in the "True" state. =2, the sum of its corresponding weights (DC internal resistance exceeding limit) + 0.4 (SOH exceeding limit) = 0.6 The gradient factor can be calculated based on the nonlinear gradient factor function. :

[0180] Child nodes under parent node combination The conditional probability of "True" can be calculated as follows:

[0181] M3, marginalize unknown parent node: In practical Bayesian network inference, all possible states of the manufacturing process and cryogenic environment need to be considered and weighted averaged. As shown in Table 3, since "manufacturing process" and "cryo-environment" are root nodes, their prior probabilities are obtained from the code. As shown in Table 4, in this site context, the initial probabilities of "manufacturing process" and "cryo-environment" are set to 0.003%, therefore:

[0182]

[0183] Therefore, the contributions of the four possibilities of these two nodes to the probability that the leakage current is True are: Scenario 1: Manufacturing process = False, Low temperature environment = False Joint probability = .

[0184] Conditional probability .

[0185] This probability contribution value .

[0186] Scenario 2: Manufacturing process = False, Low temperature environment = True Joint probability = 99.997%*0.003%≈0.003%.

[0187] in this case, .

[0188] Conditional probability .

[0189] This probability contribution value .

[0190] Scenario 3: Manufacturing process = True, Low temperature environment = False Joint probability = 0.003% * 99.997% ≈ 0.003%.

[0191] in this case, .

[0192] Conditional probability .

[0193] This probability contribution value

[0194] Scenario 4: Manufacturing process = True, Low temperature environment = True Joint probability =0.003%*0.003%≈0.00000009%.

[0195] in this case, .

[0196] This probability contribution value .

[0197] Adding the contributions of the four combinations, the probability of a node experiencing "leakage current exceeding the limit" given the evidence is approximately:

[0198] Based on the above reasoning mechanism and the following pre-set theoretical boundary, the remaining nodes are deduced in turn. Finally, the probability of thermal runaway after injecting the above fault evidence is calculated to be 0.1289%.

[0199]

[0200] Table 3 Initial weight settings for each level

[0201] Table 4 Preset probability boundaries for each level M4. Fire damage assessment based on configuration parameters: The entire station has 80 compartments, each compartment contains 10 clusters, and each cluster contains 20 packs. The battery material is lithium-ion batteries. Based on its parameters, the performance scores of each part and the final loss ratio are calculated as shown in Table 5.

[0202] Table 5 Fire Protection Configuration Parameters for Stations Based on these parameters, the property loss rate of the station was calculated to be 6.71%.

[0203] M5. Comprehensive thermal runaway probability and loss assessment: Based on a single-compartment thermal runaway probability of 0.1289%, and assuming, based on sampling calculations, that the thermal runaway probability of all compartments in the entire station is equal to that of the single-compartment thermal runaway, which is also 0.1289%, the overall thermal runaway probability of the entire station is calculated as follows:

[0204] Based on the fact that the property damage rate in the event of thermal runaway at this station is 6.71%, the estimated potential loss rate for the entire station under current thermal runaway conditions is: =0.659%.

[0205] The application scenarios of this invention are as follows: In the context of energy storage project insurance underwriting, this invention can complete the quantitative assessment of thermal runaway risk based on limited site information, assisting insurance companies in completing risk screening for insured projects.

[0206] In the scenario of optimizing fire protection configuration for energy storage power stations, this invention can calculate the impact of losses for different protection schemes, providing a reference for the selection of appropriate safety protection measures for the site.

[0207] In the scenario of energy storage accident damage assessment and claims settlement, this invention can output the loss ratio based on the actual equipment status and protection conditions to support the reasonable determination of the claim amount.

[0208] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0209] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0210] Example 2 See Figure 4 Embodiment 2 of the present invention also provides an energy storage thermal runaway damage assessment device based on energy storage safety risks and fire-fighting efficiency, comprising: The single-compartment thermal runaway probability assessment module 001 is used to construct a multi-level Bayesian network based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems; initialize the network parameters of the multi-level Bayesian network based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; generate a conditional probability table for the non-root nodes in the initialized multi-level Bayesian network using a gradient reinforcement algorithm; inject limited key objective information obtained from on-site investigation as evidence based on the conditional probability table; and calculate the posterior probability of thermal runaway using a variable elimination algorithm based on the evidence to obtain the single-compartment thermal runaway probability. The thermal runaway loss ratio acquisition module 002 is used to construct a multi-level loss reduction index system based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability. This system includes inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. It calculates the inter-compartment isolation loss reduction coefficient using an inter-compartment isolation performance evaluation model; it calculates the comprehensive detection loss reduction coefficient for both compartment and pack levels using a detection performance weighted scoring model; it calculates the comprehensive fire suppression loss reduction coefficient for both compartment and pack levels using a fire suppression performance weighted scoring model; based on the comprehensive detection loss reduction coefficient and the comprehensive fire suppression loss reduction coefficient, it calculates the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient; and it integrates the inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient using a series model, employing differentiated calculation methods based on whether pack-level and compartment-level protection are provided, to calculate the thermal runaway loss ratio. The station-level thermal runaway expected loss ratio acquisition module 003 is used to calculate the station-wide thermal runaway probability based on the single-compartment thermal runaway probability; dynamically integrate the station-wide thermal runaway probability and the thermal runaway loss ratio through a coupling algorithm, output the station-level thermal runaway expected loss ratio, and complete the energy storage thermal runaway loss assessment.

[0211] In this embodiment, the multi-level Bayesian network in the single-compartment thermal runaway probability assessment module 001 includes: a data acquisition and environmental parameter layer, a primary fault layer, a secondary fault propagation layer, a safety boundary layer, and a fault assessment layer; the multi-source prior knowledge includes: domain expert experience, historical fault statistics, equipment reliability models, and simulation data. In this embodiment, the formula for calculating the conditional probability in the single-compartment thermal runaway probability assessment module 001 is as follows:

[0212] In the formula, This represents the conditional probability that the node is in an abnormal state. and These are the preset minimum and maximum probability baseline values ​​for the level to which the node belongs; It is a nonlinear gradient factor; This represents the number of parent nodes that are in an abnormal state. This represents the total number of parent nodes; This is the sum of the influence weights of the parent node in the abnormal state; This represents the maximum sum of the weights of the parent nodes. The target is a non-root node; , , , These are all parameters configured differently based on node type and importance.

[0213] In this embodiment, in the single-compartment thermal runaway probability assessment module 001, during the process of calculating the posterior probability of thermal runaway through the variable elimination algorithm, based on Bayes' theorem, the joint probability distribution of the network is decomposed into the product of the conditional probabilities of each node, and non-query variables are eliminated through marginal summation to obtain unnormalized probability factors; the posterior probability of thermal runaway is obtained by normalizing the unnormalized probability factors. The formula for calculating the posterior probability of thermal runaway is:

[0214]

[0215] In the formula, For the collection of evidence E Under the condition of querying variables Q Values q The posterior probability; For querying variables; The set of evidence to be injected; This is an unnormalized probability factor containing only query variables and evidence; This is a normalization constant; For querying variables Q Possible values ​​for .

[0216] In this embodiment, in the thermal runaway loss ratio acquisition module 002, during the calculation of the inter-cabin isolation loss reduction coefficient, when the actual isolation distance... ≥ When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0217] In the formula, This is the inter-cabin isolation reduction factor; This represents the total number of battery compartments across the entire station. when ≤ < When the inter-cabin isolation reduction factor is calculated, the formula is as follows:

[0218]

[0219] when hour: =0 In the formula, To normalize the distance progress; Recommended safe distance according to national standards; The minimum safe distance is mandated by national standards; This refers to the actual isolation distance.

[0220] In this embodiment, the calculation formula for the detection loss reduction comprehensive coefficient of the cabin level and the Pack level in the thermal runaway loss ratio acquisition module 002 is as follows:

[0221] In the formula, This is the overall performance coefficient for detection; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the overall fire suppression loss reduction factor for cabin and pack classes is as follows:

[0222] In the formula, This is the comprehensive coefficient for fire extinguishing performance; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the loss reduction coefficient of the cabin class system is as follows:

[0223] In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the cabin-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is as follows:

[0224] In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system.

[0225] In this embodiment, the expression for obtaining the thermal runaway loss ratio in the thermal runaway loss ratio module 002 through the series model is as follows:

[0226] In the formula, This represents the proportion of thermal runaway losses.

[0227] In this embodiment, the calculation formula for the probability of thermal runaway across the entire station in the station-level thermal runaway expected loss ratio acquisition module 003 is as follows:

[0228]

[0229] In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. For the first k Probability of thermal runaway in each battery compartment; The formula for calculating the expected loss ratio of the site-level thermal runaway is as follows:

[0230] In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

[0231] It should be noted that the information interaction and execution process between the modules of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0232] Example 3 Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, wherein the computer-readable storage medium stores program code for a method for assessing the damage of thermal runaway of energy storage based on energy storage safety risks and fire protection efficiency. The program code includes instructions for executing the method for assessing the damage of thermal runaway of energy storage based on energy storage safety risks and fire protection efficiency as described in Embodiment 1 or any possible implementation thereof.

[0233] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0234] Example 4 Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor; The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can call the program instructions to execute the energy storage thermal runaway loss assessment method based on energy storage safety risks and fire protection efficiency in Embodiment 1 or any possible implementation thereof.

[0235] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0236] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and 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 system. 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 via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0237] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0238] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for determining the loss of energy storage thermal runaway based on energy storage safety risk and fire efficiency, characterized in that, include: Based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems, a multi-level Bayesian network is constructed. The network parameters of the multi-level Bayesian network are initialized based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; for the non-root nodes in the initialized multi-level Bayesian network, a conditional probability table is generated through a gradient reinforcement algorithm. Based on the conditional probability table, limited key objective information obtained from on-site investigation is injected through the interface as evidence; based on the evidence, the posterior probability of thermal runaway is calculated through the variable elimination algorithm to obtain the probability of thermal runaway in a single compartment. Based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability, a multi-level loss reduction index system is constructed, including inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. The inter-compartment isolation loss reduction coefficient is calculated using an inter-compartment isolation performance evaluation model. The comprehensive detection loss reduction coefficient for the compartment and pack levels is calculated using a detection performance weighted scoring model. The comprehensive fire suppression loss reduction coefficient for the compartment and pack levels is calculated using a fire suppression performance weighted scoring model. Based on the comprehensive detection loss reduction coefficient and the comprehensive fire suppression loss reduction coefficient, the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient are calculated. The inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient are integrated using a series model. Differential calculation methods are adopted based on whether pack-level and compartment-level protection are provided to calculate the thermal runaway loss ratio. Based on the single-compartment thermal runaway probability, the overall thermal runaway probability is calculated; the overall thermal runaway probability and the thermal runaway loss ratio are dynamically integrated through a coupling algorithm to output the expected loss ratio of thermal runaway at the site level, thus completing the energy storage thermal runaway loss assessment.

2. The energy storage thermal runaway damage assessment method based on energy storage safety risk and fire efficiency according to claim 1, characterized in that, The multi-level Bayesian network includes: a data acquisition and environmental parameter layer, a ontological fault layer, a secondary fault propagation layer, a security boundary layer, and a fault assessment layer; the multi-source prior knowledge includes: domain expert experience, historical fault statistics, equipment reliability models, and simulation data.

3. The energy storage thermal runaway damage assessment method based on energy storage safety risk and fire efficiency according to claim 2, characterized in that, The formula for calculating the conditional probability is: ； ； wherein, is the conditional probability that the node is in an abnormal state; are preset minimum and maximum probability reference values of the level where the node is located, respectively; is a nonlinear gradient factor; is the number of parent nodes in an abnormal state; is the total number of parent nodes; is the sum of the influence weights of the parent nodes in an abnormal state; is the maximum value of the total weight of the parent nodes; is the target non-root node; are parameters configured differently according to the node type and importance.​​​​ 4. The energy storage thermal runaway damage assessment method based on energy storage safety risk and fire efficiency according to claim 3, characterized in that, In the process of calculating the posterior probability of thermal runaway using the variable elimination algorithm, based on Bayes' theorem, the joint probability distribution of the network is decomposed into the product of the conditional probabilities of each node, and non-query variables are eliminated through marginal summation to obtain unnormalized probability factors; the posterior probability of thermal runaway is obtained by normalizing the unnormalized probability factors. The formula for calculating the posterior probability of thermal runaway is: ； ； where is the set of evidence E is the query variable is the posterior probability that the query variable q takes the value is the query variable; is the set of injected evidence; is the unnormalized probability factor containing only the query variable and the evidence; is the normalization constant; is the possible values of the query variable .

5. The energy storage thermal runaway loss assessment method based on energy storage safety risk and fire protection efficiency according to claim 4, characterized in that, In the process of calculating the cabin separation impairment factor, when the actual separation distance is less than the design separation distance, the formula for calculating the cabin separation impairment factor is: ； wherein is the inter-bay isolation degradation factor; is the total number of battery bays across the station. When ≤ The cabin isolation reduction factor is calculated by the following formula: ； ； When time: =0； wherein, is the normalized distance progress; is the national standard recommended safety distance; is the national standard mandatory minimum safety distance; is the actual isolation distance.

6. The energy storage thermal runaway damage assessment method based on energy storage safety risk and fire efficiency according to claim 5, characterized in that, The formula for calculating the overall detection loss reduction coefficient for both cabin and pack classes is as follows: ； wherein is a coefficient of detection performance; is a coefficient of response speed performance; is a coefficient of reliability performance; is a coefficient of life performance; The formula for calculating the overall fire suppression loss reduction factor for cabin and pack classes is as follows: ； In the formula, is a fire extinguishing performance comprehensive coefficient; is a response speed performance coefficient; is a reliability performance coefficient; is a life performance coefficient; The formula for calculating the loss reduction coefficient of the cabin class system is as follows: ； In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the compartment-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is as follows: ； In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system.

7. The energy storage thermal runaway loss assessment method based on energy storage safety risk and fire protection efficiency according to claim 6, characterized in that, The expression for the thermal runaway loss ratio calculated using the aforementioned series model is as follows: ； In the formula, This represents the proportion of thermal runaway losses.

8. The energy storage thermal runaway loss assessment method based on energy storage safety risk and fire protection efficiency according to claim 7, characterized in that, The formula for calculating the probability of thermal runaway at the entire station is as follows: ； ； In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. For the first k Probability of thermal runaway in each battery compartment; The formula for calculating the expected loss ratio of the site-level thermal runaway is as follows: ； In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

9. A thermal runaway loss assessment device for energy storage based on energy storage safety risks and fire-fighting efficiency, employing the thermal runaway loss assessment method for energy storage based on energy storage safety risks and fire-fighting efficiency as described in any one of claims 1-8, characterized in that, include: The single-compartment thermal runaway probability assessment module is used to construct a multi-level Bayesian network based on the thermal runaway evolution mechanism and engineering experience of electrochemical energy storage systems. The network parameters of the multi-level Bayesian network are initialized based on multi-source prior knowledge to obtain the initialized multi-level Bayesian network; for the non-root nodes in the initialized multi-level Bayesian network, a conditional probability table is generated through a gradient reinforcement algorithm. Based on the conditional probability table, limited key objective information obtained from on-site investigation is injected through the interface as evidence; based on the evidence, the posterior probability of thermal runaway is calculated through the variable elimination algorithm to obtain the probability of thermal runaway in a single compartment. The thermal runaway loss ratio acquisition module is used to construct a multi-level loss reduction index system based on the structural characteristics of the energy storage system corresponding to the single-compartment thermal runaway probability. This system includes inter-compartment isolation performance, compartment-level detection and fire suppression capabilities, and pack-level detection and fire suppression capabilities. It calculates the inter-compartment isolation loss reduction coefficient using an inter-compartment isolation performance evaluation model; it calculates the comprehensive detection loss reduction coefficient for both compartment and pack levels using a detection performance weighted scoring model; it calculates the comprehensive fire suppression loss reduction coefficient for both compartment and pack levels using a fire suppression performance weighted scoring model; based on the comprehensive detection loss reduction coefficient and the comprehensive fire suppression loss reduction coefficient, it calculates the compartment-level system loss reduction coefficient and the pack-level system loss reduction coefficient; and it integrates the inter-compartment isolation loss reduction coefficient, the compartment-level system loss reduction coefficient, and the pack-level system loss reduction coefficient using a series model, employing differentiated calculation methods based on whether pack-level and compartment-level protection are provided, to calculate the thermal runaway loss ratio. The station-level thermal runaway expected loss ratio acquisition module is used to calculate the station-wide thermal runaway probability based on the single-compartment thermal runaway probability; the station-wide thermal runaway probability and the thermal runaway loss ratio are dynamically integrated through a coupling algorithm to output the station-level thermal runaway expected loss ratio, thus completing the energy storage thermal runaway loss assessment.

10. The energy storage thermal runaway damage assessment device based on energy storage safety risk and fire-fighting efficiency according to claim 9, characterized in that, In the single-compartment thermal runaway probability assessment module, the multi-level Bayesian network includes: a data acquisition and environmental parameter layer, a primary fault layer, a secondary fault propagation layer, a safety boundary layer, and a fault assessment layer; the multi-source prior knowledge includes: domain expert experience, historical fault statistics, equipment reliability models, and simulation data. In the single-compartment thermal runaway probability assessment module, the formula for calculating the conditional probability is: ； ； In the formula, This represents the conditional probability that the node is in an abnormal state. and These are the preset minimum and maximum probability baseline values ​​for the level to which the node belongs; It is a nonlinear gradient factor; This represents the number of parent nodes that are in an abnormal state. This represents the total number of parent nodes; This is the sum of the influence weights of the parent nodes in the abnormal state; This represents the maximum sum of the weights of the parent nodes. The target is a non-root node; , , , These are all parameters configured differently based on node type and importance; In the single-compartment thermal runaway probability assessment module, during the process of calculating the posterior probability of thermal runaway using the variable elimination algorithm, based on Bayes' theorem, the joint probability distribution of the network is decomposed into the product of the conditional probabilities of each node, and non-query variables are eliminated through marginal summation to obtain unnormalized probability factors; the posterior probability of thermal runaway is obtained by normalizing the unnormalized probability factors. The formula for calculating the posterior probability of thermal runaway is: ； ； In the formula, For the collection of evidence E Under the condition of querying variables Values q The posterior probability; For querying variables; The set of evidence to be injected; This is an unnormalized probability factor containing only query variables and evidence; This is the normalization constant; For querying variables Q Possible values; In the thermal runaway loss ratio acquisition module, during the calculation of the inter-cabin isolation loss reduction coefficient, when the actual isolation distance... When the inter-cabin isolation reduction factor is calculated, the formula is as follows: ； In the formula, This is the inter-cabin isolation reduction factor; This represents the total number of battery compartments across the entire station. when When the inter-cabin isolation reduction factor is calculated, the formula is as follows: ； ； when hour: =0 in the formula, To normalize the distance progress; Recommended safe distance according to national standards; The minimum safe distance is mandated by national standards; This refers to the actual isolation distance; In the thermal runaway loss ratio acquisition module, the calculation formula for the detection loss reduction comprehensive coefficient for the cabin level and the Pack level is as follows: ； In the formula, This is the overall performance coefficient for detection; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the overall fire suppression loss reduction factor for cabin and pack classes is as follows: ； In the formula, This is the comprehensive coefficient for fire extinguishing performance; The coefficient of performance is the response speed. The reliability performance coefficient; This refers to the life performance coefficient. The formula for calculating the loss reduction coefficient of the cabin class system is as follows: ； In the formula, For cabin-level system loss reduction coefficient; This refers to the reduction ratio for cabin classes that only meet basic standards. This refers to the actual performance coefficient of the cabin-level detection system; This refers to the actual performance coefficient of the compartment-level fire suppression system; , These are the baseline weights for detection and fire suppression systems, respectively. The formula for calculating the loss reduction coefficient of the Pack-level system is as follows: ； In the formula, The loss reduction factor for the Pack-level system; The Pack level only has the loss reduction ratio under the condition of meeting the basic standard; This represents the number of packs within the battery cluster. The number of battery clusters contained in a single battery compartment; This refers to the actual performance coefficient of a Pack-level detection system; This represents the actual performance coefficient of a Pack-level fire suppression system. In the thermal runaway loss ratio acquisition module, the expression for calculating the thermal runaway loss ratio using the series model is as follows: ； In the formula, This represents the proportion of thermal runaway losses. In the station-level thermal runaway expected loss ratio acquisition module, the calculation formula for the station-wide thermal runaway probability is as follows: ； ； In the formula, The probability of thermal runaway at the entire site; For the overall site security probability; n This represents the total number of battery compartments at the entire station. For the first k Probability of thermal runaway in each battery compartment; The formula for calculating the expected loss ratio of the site-level thermal runaway is as follows: ； In the formula, The expected loss ratio for thermal runaway at the station level; This is the result of the calculation of the total station loss ratio when thermal runaway occurs.

Images