Hydrogen leakage detection and risk response method based on dynamic Bayesian fault tree

Abstract

The invention discloses a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree. The method comprises the following steps: S1, analyzing a hydrogen leakage accident occurrence mode of a mobile hydrogen energy device; s2, clarifying an accident leakage position and a leakage reason based on an existing hydrogen accident event library and current main leakage components; s3, establishing a hydrogen safety fault tree of the mobile hydrogen energy device in four typical scenes to obtain the occurrence rate of each fault of the whole system; s4, constructing a dynamic Bayesian (DBN) model integrating the four types of sub-fault trees, and introducing a Leaky Noise-or gate model to adjust probability parameters; s5, adopting diagnostic reasoning to obtain a time sequence curve of a top event T (hydrogen leakage) and a posterior probability of a basic event under each fault; and S6, performing risk response at all levels on the third-level safety barrier (maintenance, shutdown and evacuation) according to a probability result. According to the invention, a multi-level safety barrier is established by adopting dynamic risk analysis and risk response, dynamic alarm of the system is realized, and response decisions of all levels of the system are given.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen leakage, and in particular to a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree. Background Art

[0002] With the rapid development of industry, global demand for energy continues to increase. The extraction, transportation, storage, and use of energy play a key role in driving the progress of modern society. Hydrogen, due to its environmental friendliness and high calorific value, has become one of the most promising clean energy sources. However, hydrogen leakage is a major safety concern for mobile hydrogen energy systems. During the operation of hydrogen power generation vehicles, hydrogen leakage can be caused by a variety of factors, including human error, damage to pipes and valves, damage to hydrogen supply lines, and the release of temperature-activated safety relief devices due to overheating or overpressure in gas cylinders. When leaked hydrogen reaches its flammability limit and encounters sufficient ignition energy, it can cause a combustion and explosion accident. Therefore, hydrogen leakage reaching the flammability limit should be used as the top event in the fault tree for analysis. By constructing a hydrogen leakage fault tree model for mobile emergency power supply vehicles, the key factors leading to hydrogen leakage events can be identified, and proactive prevention and control measures can be implemented to reduce the probability of hydrogen leakage, thereby ensuring the stable operation of mobile emergency power supply vehicles.

[0003] Traditional fault tree, event tree and static Bayesian network models often have significant limitations when analyzing multi-stage and component polymorphism of complex systems.

[0004] In actual engineering applications, the main sources of leakage in hydrogen-fueled mobile emergency power supply vehicles include hydrogen storage tanks, pipelines, and fuel cell stacks. During actual operation, the safety of pipelines is reduced due to the combined effects of various factors, including hydrogen embrittlement failure of components, mechanical damage, and misoperation. Once hydrogen leaks, due to its light weight and rapid diffusion rate, it easily forms a flammable hazardous zone, potentially causing significant harm to people, the environment, and assets. Therefore, implementing a safety risk assessment process, including hazard source identification and management, calculating accident probability, and proposing safety countermeasures, is of great significance for identifying safety risks in mobile emergency power supply vehicles.

[0005] Current traditional fault diagnosis and risk response methods lack bidirectional integration across the time dimension, resulting in incomplete leakage scenario analysis and poor real-time dynamic response to leakage situations. Therefore, a comprehensive, rapid, real-time, and accurate method for system fault tree diagnosis and risk response in various scenarios is needed. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree. By constructing sub-fault trees under four typical scenarios to analyze the causes and leakage status of each leakage, a dynamic Bayesian network is used to realize the dynamic analysis of the hydrogen leakage risk of the mobile emergency power supply, the safety barrier level of the risk response is adjusted according to the probability curve, and the posterior probability of hydrogen leakage is calculated by reverse reasoning to obtain the key factors of the leakage, which is conducive to ensuring the safe and stable operation of mobile hydrogen energy devices under various situations. In order to achieve the above-mentioned purpose and other advantages of the present invention, a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree is provided, comprising:

[0007] S1. Analyze the occurrence pattern of hydrogen leakage accidents in mobile hydrogen energy devices;

[0008] S2. Obtain the accident leakage location and leakage cause based on the existing hydrogen accident event database and the current main leakage components;

[0009] S3. Establish a hydrogen safety fault tree for mobile hydrogen energy devices under typical scenarios, and then obtain the occurrence rate of each fault;

[0010] S4. Establish a dynamic Bayesian model that integrates four types of sub-fault trees, and introduce the Leaky Noisy-or gate model to adjust the probability parameters;

[0011] S5. Use diagnostic reasoning to obtain a time series curve of hydrogen leakage and the posterior probability of basic events under various faults;

[0012] S6. Conduct risk responses at all levels for the Level 3 safety barrier through probabilistic results.

[0013] Preferably, the typical scenarios in step S3 include power generation, refueling, driving and fault detection, and the establishment of the hydrogen safety fault tree specifically includes the following steps:

[0014] S31. Determine the top event of the fault tree and the intermediate events in the general scenario through the hazard analysis method;

[0015] S32, determining the next level of fault events based on the leaking component according to a top-down indexing method;

[0016] S33. Establish sub-fault trees for each leakage scenario based on four typical hydrogen safety scenarios;

[0017] S34. Analyze the dynamic weight of the upper layer fault event caused by the lower layer fault according to the logic gate calculation method.

[0018] Preferably, the establishment of sub-fault trees under each leakage scenario for the four typical hydrogen safety scenarios in step S33 specifically includes: combining accident losses and logical analysis methods to obtain basic events and related causes in the power generation scenario, and establishing a hydrogen leakage sub-fault tree in the power generation scenario;

[0019] Combined with accident losses and logical analysis methods, the basic events and related causes in the power generation scenario were obtained, and a sub-fault tree for hydrogen leakage in the power generation scenario was established;

[0020] Combined with accident losses and logical analysis methods, the basic events and related causes in the refueling scenario were obtained, and a sub-fault tree for hydrogen leakage in the refueling scenario was established;

[0021] Combine accident losses with logical analysis methods to obtain basic events and related causes in driving scenarios, and establish a hydrogen leakage sub-fault tree in driving scenarios;

[0022] Combined with accident losses and logical analysis methods, the basic events and related causes in the fault detection scenario are obtained, and the hydrogen leakage sub-fault in the fault detection scenario is established.

[0023] Preferably, the establishment of the dynamic Bayesian model in step S4 specifically includes the following steps:

[0024] S41, combining the sub-fault tree analysis in step S3 with the event ET analysis to establish a BowTie (BT) model;

[0025] S42, mapping the events at each level of the BT model to the nodes at each level of the dynamic Bayesian model one by one;

[0026] S43, relying on the conditional probability table to establish the parameter probability between nodes in the state Bayesian model, wherein the parameter probability includes the conditional probability, the prior probability and the transition probability;

[0027] S44. The conditional probability is modified through the Leaky Noisy-or gate model, and the dynamic characteristics of node failure over time are captured through the transition probability.

[0028] Preferably, the step S6 of performing risk responses at different levels on the level 3 safety barriers specifically includes:

[0029] The three-level safety barrier includes first-level maintenance, second-level shutdown and third-level evacuation;

[0030] When the risk probability Pi is less than 0.04, the safety barrier is in a dormant state, and the equipment is considered to be in a healthy working state;

[0031] When the risk probability is 0.04<Pi<0.2, the first-level safety barrier is activated, the equipment automatically alarms, and enters the maintenance state;

[0032] When the risk probability is 0.2<Pi<0.37, the second-level safety barrier is activated and the equipment is shut down urgently to prevent further hydrogen leakage and accumulation;

[0033] When the risk probability Pi>0.37, the first two levels of safety barriers fail, the third level of safety barrier is activated, the leakage accident continues to escalate, and people near the mobile device are evacuated in an orderly manner.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] (1) The hydrogen leakage situation in each typical scenario is subdivided by constructing sub-fault trees corresponding to power generation, refueling, driving and fault repair scenarios, achieving a comprehensive analysis of the cause and status of the leakage event;

[0036] (2) A dynamic Bayesian network fault tree is used for diagnosis to visualize the logical relationships between system faults. The leaky noisy-or gate model is used to optimize the conditional probability, so that the results can more accurately and comprehensively describe the occurrence mechanism of fault-induced accidents and the evolution mechanism of accidents, which is more in line with the actual situation.

[0037] (3) Based on diagnostic reasoning, the time series curve of hydrogen pipeline leakage and the posterior probability of basic events are obtained. The posterior probability can be used to determine the important factors affecting hydrogen leakage accidents; using the probability results, a three-level safety barrier response method is adopted to reduce the harmfulness of hydrogen leakage accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree according to the present invention;

[0039] Figure 2 A fault tree model for hydrogen leakage in a general scenario according to embodiment 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0040] Figure 3 A fault tree model of hydrogen leakage in a power generation scenario according to embodiment 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0041] Figure 4 A fault tree model of hydrogen leakage in a refueling scenario according to embodiment 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0042] Figure 5 A fault tree model of hydrogen leakage in a driving scenario according to embodiment 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0043] Figure 6A fault tree model of hydrogen leakage in a fault repair scenario according to embodiment 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0044] Figure 7 A schematic diagram of a dynamic Bayesian network structure of Example 1 of a hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree according to the present invention;

[0045] Figure 8 A general dynamic Bayesian network model of a hydrogen leakage accident according to Example 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0046] Figure 9 A graph showing a hydrogen leakage risk probability curve of a mobile emergency power supply according to Example 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree of the present invention;

[0047] Figure 10 A comparison diagram of the prior probability and posterior probability of a basic event of hydrogen leakage in a general scenario of Example 1 of the hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree according to the present invention;

[0048] Figure 11 This is a comparison diagram of the prior probability and posterior probability of hydrogen leakage intermediate events in a general scenario of Example 1 of the hydrogen leakage detection and risk response method based on the dynamic Bayesian fault tree of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] Reference Figure 1 ,A hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree, comprising: S1, analyzing the occurrence mode of hydrogen leakage accidents in mobile hydrogen energy devices;

[0051] S2. Obtain the accident leakage location and leakage cause based on the existing hydrogen accident event database and the current main leakage components;

[0052] S3. Establish a hydrogen safety fault tree for a mobile hydrogen energy device under typical scenarios to obtain the incidence rate of each fault. Typical scenarios include power generation, refueling, driving, and fault detection. Establishing the hydrogen safety fault tree specifically includes the following steps:

[0053] S31. Determine the top event of the fault tree and the intermediate events in the general scenario through the hazard analysis method;

[0054] S32, determining the next level of fault events based on the leaking component according to a top-down indexing method;

[0055] S33. Establish sub-fault trees for each leakage scenario based on four typical hydrogen safety scenarios;

[0056] S34. Analyze the dynamic weight of the upper layer fault event caused by the lower layer fault according to the logic gate calculation method.

[0057] The sub-fault trees for each leakage scenario in four typical hydrogen safety scenarios are established. Specifically, the basic events and related causes in the power generation scenario are obtained by combining accident losses and logical analysis methods, and the hydrogen leakage sub-fault tree in the power generation scenario is established;

[0058] Combined with accident losses and logical analysis methods, the basic events and related causes in the power generation scenario were obtained, and a sub-fault tree for hydrogen leakage in the power generation scenario was established;

[0059] Combined with accident losses and logical analysis methods, the basic events and related causes in the refueling scenario were obtained, and a sub-fault tree for hydrogen leakage in the refueling scenario was established;

[0060] Combine accident losses with logical analysis methods to obtain basic events and related causes in driving scenarios, and establish a hydrogen leakage sub-fault tree in driving scenarios;

[0061] Combined with accident losses and logical analysis methods, the basic events and related causes in the fault detection scenario are obtained, and the hydrogen leakage sub-fault in the fault detection scenario is established.

[0062] S4. Establish a dynamic Bayesian model that integrates the four sub-fault trees and introduces a leaky noisy-or gate model to adjust the probability parameters. The general structure of the dynamic Bayesian model is that the current time slice is represented by t and the next time slice is represented by t+1. The establishment of the dynamic Bayesian model specifically includes the following steps:

[0063] S41, combining the sub-fault tree analysis in step S3 with the event ET analysis to establish a BowTie (BT) model;

[0064] S42, map the events at each level of the BT model to the nodes at each level of the dynamic Bayesian model one by one; map the events at each level of the BT model to the nodes at each level of the DBN model one by one, the prior probability of the parent node in the DBN model is numerically the same as the probability of occurrence of the basic event in the BT model, and the failure probability of the basic event is given by quantitative expression;

[0065] Among them, P(C) is the probability of accident consequences occurring, and P(T) is the probability of top event failure.

[0066] S43, the parameter probability between nodes in the dynamic Bayesian model is established by relying on the conditional probability table, which includes conditional probability, prior probability and transition probability; the conditional probability table (CPT) is related to the logic gates such as the "AND" gate and "OR" gate in the fault tree. The dynamic Bayesian network established is based on the prior network B0 and the transition network B → Got it.

[0067] S44, the conditional probability is modified by the Leaky Noisy-or gate model, and the dynamic characteristics of node failure over time are captured by the transition probability. The Leaky Noisy-or gate model adjusts the conditional probability by omitting the node X L Got it, X L The connection probability is P L =0.1. The conditional probability of node Y in the Leaky Noisy-or gate model is Among them, Y=T means event Y occurs, X i represents the i-th parent node of node Y, X p Indicates that node Y is divided by X i All parent nodes except .

[0068] Furthermore, the general structure of the dynamic Bayesian model uses arcs within the same time slice to represent X ti →Y t The development relationship between different variables in the slice is represented by the arc between slices, which represents the same variable in the continuous time step X. t →X ti and Y t →Y ti The development relationship between them.

[0069] S5. Use diagnostic reasoning to obtain a time series curve of hydrogen leakage and the posterior probability of basic events under various faults;

[0070] S6. Conduct risk responses at all levels for the Level 3 safety barrier using probability results. Specific risk responses at all levels for the Level 3 safety barrier include:

[0071] The three-level safety barrier includes first-level maintenance, second-level shutdown and third-level evacuation;

[0072] When the risk probability Pi is less than 0.04, the safety barrier is in a dormant state, and the equipment is considered to be in a healthy working state;

[0073] When the risk probability is 0.04<Pi<0.2, the first-level safety barrier is activated, the equipment automatically alarms, and enters the maintenance state;

[0074] When the risk probability is 0.2<Pi<0.37, the second-level safety barrier is activated and the equipment is shut down urgently to prevent further hydrogen leakage and accumulation;

[0075] When the risk probability Pi>0.37, the first two levels of safety barriers fail, the third level of safety barrier is activated, the leakage accident continues to escalate, and people near the mobile device are evacuated in an orderly manner.

[0076] Furthermore, the constructed three-level risk barrier is obtained by estimating the variable nodes at each time segment in the top event time series curve.

[0077] The present invention is further explained or illustrated by the following examples:

[0078] As shown in Table 1, based on the logical step-by-step analysis and combined with the actual operation and workflow of the hydrogen fuel mobile emergency power supply vehicle, the intermediate events and basic events are determined; the basic event X, intermediate event M and top event T in the general scenario of the hydrogen fuel mobile emergency power supply and four different working scenarios are shown in the following table.

[0079] Table 1 List of events in hydrogen leakage fault tree analysis for general scenarios

[0080]

[0081]

[0082] like Figure 2 As shown, after investigating the hydrogen leakage incident involving a mobile emergency power supply vehicle and analyzing the causal relationships between the incidents, it is generally determined that the top event is the one with the highest probability of investigation. Here, the hydrogen leakage from the hydrogen fuel mobile emergency power supply is selected as the top event for this analysis. The causes of the accident include direct and indirect causes. Direct causes include human negligence, damaged equipment and materials, and management defects. Other causes can be roughly classified as indirect causes. A hydrogen leakage fault tree model for the general scenario of a mobile emergency power supply vehicle was constructed, as shown in the figure. The symbols contained in the fault tree are connected in series with the event logic at each level in the list.

[0083] like Figure 3-Figure 6 As shown, in this embodiment, each sub-fault tree in the power generation, refueling, driving and fault repair scenarios follows the same deductive reasoning process as the general scenario. According to the principle of gradual progress, the top event is first determined, and then the analysis continues step by step until all the basic matters that lead to the occurrence of the above events are listed one by one.

[0084] Table 2 List of events in hydrogen leakage fault tree analysis for power generation scenario

[0085]

[0086]

[0087] Table 3 List of events in hydrogen leakage fault tree analysis of refueling scenario

[0088]

[0089]

[0090]

[0091] Table 4 List of events in the hydrogen leakage fault tree analysis of driving scenarios

[0092]

[0093]

[0094] Table 5 List of events in hydrogen leakage fault tree analysis of fault repair scenario

[0095]

[0096]

[0097] like Figure 7 As shown, in Example 1, the prior network B0 represents the network structure and probability distribution P([X1]) of the Bayesian network at the initial moment, and the joint probability is given by obtain, represents the i-th node at time t (i=1,2,…,N), Indicates the network The parent node of .

[0098] like Figure 8 As shown in the figure, the parent node, intermediate node and child node in the DBN model are represented by the basic event, intermediate event and top event in each sub-fault tree. The prior probability of the parent node in the DBN model is represented by Quantitative expression.

[0099] like Figure 9 As shown, the prior probabilities, conditional probabilities, and transition probabilities of each factor are input into the established DBN inference model. The number of time slices is set to 10 for probability update, and variable nodes are estimated over these 10 time slices. Event T (hydrogen leak) is set as the evidence node, and the accident begins at the initial time slice. The probability of a hydrogen pipeline leak in the initial time slice is 0.04, and then gradually increases with each time slice. The probability reaches 0.2 in the fourth time slice, and the increase in probability slows down thereafter, reaching 0.37 by the end of the cycle. This period presents a high risk of leakage, necessitating periodic maintenance.

[0100] Table 6 Prior probability table of basic events in hydrogen leakage fault tree for general scenarios

[0101]

[0102]

[0103] like Figure 10 As shown, the prior probability after BT mapping and the posterior probability after optimization by the LeakyNoisy-or gate model in the general scenario of Example 1 are Compare.

[0104] like Figure 11 As shown, the posterior probabilities of nodes M1 and M2 in Example 1 are the highest, indicating that among all factors, pipeline damage and valve failure are important factors leading to hydrogen leakage. When formulating maintenance measures, special attention should be paid to maintenance work in this regard.

[0105] The number of devices and processing scales described herein are intended to simplify the description of the present invention, and the application, modification, and variation of the present invention will be apparent to those skilled in the art. Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiment. They can be applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily implemented. Therefore, the present invention is not limited to the specific details and illustrations shown and described herein without departing from the general concept defined by the claims and their equivalents.

Claims

1. A hydrogen leakage detection and risk response method based on dynamic Bayesian fault tree, characterized in that: The following steps are involved: S1. Analyze the occurrence pattern of hydrogen leakage accidents in mobile hydrogen energy devices; S2. Obtain the accident leakage location and leakage cause based on the existing hydrogen accident event database and the current main leakage components; S3. Establish a hydrogen safety fault tree for mobile hydrogen energy devices under typical scenarios, and then obtain the occurrence rate of each fault; S4. Establish a dynamic Bayesian model that integrates four types of sub-fault trees, and introduce the LeakyNoisy-or gate model to adjust the probability parameters; S5. Use diagnostic reasoning to obtain a time series curve of hydrogen leakage and the posterior probability of basic events under various faults; S6. Conduct risk responses at all levels for the Level 3 safety barrier through probabilistic results.

2. A hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree according to claim 1, characterized in that: Typical scenarios in step S3 include power generation, refueling, driving, and fault detection, and the establishment of the hydrogen safety fault tree specifically includes the following steps: S31. Determine the top event of the fault tree and the intermediate events in the general scenario through the hazard analysis method; S32, determining the next level of fault events based on the leaking component according to a top-down indexing method; S33. Establish sub-fault trees for each leakage scenario based on four typical hydrogen safety scenarios; S34. Analyze the dynamic weight of the upper layer fault event caused by the lower layer fault according to the logic gate calculation method.

3. A hydrogen leakage detection and risk response method based on a dynamic Bayesian fault tree according to claim 2, characterized in that: The establishment of sub-fault trees for each leakage scenario in the four typical hydrogen safety scenarios in step S33 specifically includes: combining accident losses and logical analysis methods to obtain basic events and related causes in the power generation scenario, and establishing a hydrogen leakage sub-fault tree in the power generation scenario; Combined with accident losses and logical analysis methods, the basic events and related causes in the power generation scenario were obtained, and a sub-fault tree for hydrogen leakage in the power generation scenario was established; Combined with accident losses and logical analysis methods, the basic events and related causes in the refueling scenario were obtained, and a sub-fault tree for hydrogen leakage in the refueling scenario was established; Combine accident losses with logical analysis methods to obtain basic events and related causes in driving scenarios, and establish a hydrogen leakage sub-fault tree in driving scenarios; Combined with accident losses and logical analysis methods, the basic events and related causes in the fault detection scenario are obtained, and the hydrogen leakage sub-fault in the fault detection scenario is established.

4. The hydrogen leakage detection and risk response method based on dynamic Bayesian fault tree according to claim 1, characterized in that: The dynamic Bayesian model establishment in step S4 specifically includes the following steps: S41, combining the sub-fault tree analysis in step S3 with the event ET analysis to establish a Bow Tie (BT) model; S42, mapping the events at each level of the BT model to the nodes at each level of the dynamic Bayesian model one by one; S43, relying on the conditional probability table to establish the parameter probability between nodes in the state Bayesian model, wherein the parameter probability includes the conditional probability, the prior probability and the transition probability; S44. The conditional probability is modified through the LeakyNoisy-or gate model, and the dynamic characteristics of node failure over time are captured through the transition probability.

5. The hydrogen leakage detection and risk response method based on dynamic Bayesian fault tree according to claim 1, characterized in that: The step S6 specifically includes: The three-level safety barrier includes first-level maintenance, second-level shutdown and third-level evacuation; When the risk probability Pi is less than 0.04, the safety barrier is in a dormant state, and the equipment is considered to be in a healthy working state; When the risk probability is 0.04<Pi<0.2, the first-level safety barrier is activated, the equipment automatically alarms, and enters the maintenance state; When the risk probability is 0.2<Pi<0.37, the second-level safety barrier is activated and the equipment is shut down urgently to prevent further hydrogen leakage and accumulation; When the risk probability Pi>0.37, the first two levels of safety barriers fail, the third level of safety barrier is activated, the leakage accident continues to escalate, and people near the mobile device are evacuated in an orderly manner.

6. A hydrogen leakage detection and risk response method based on dynamic Bayesian fault tree according to claim 5, characterized in that: The constructed three-level risk barrier is obtained by estimating the variable nodes at each time segment in the top event time series curve.

Images