A method for analyzing a leakage accident of a liquefied natural gas receiving station

Abstract

The application discloses a liquefied natural gas receiving station leakage accident analysis method, comprising the following steps: obtaining different types of basic event records in a leakage accident to obtain a preliminary event feature set; constructing a dynamic correlation network of an accident scene according to the preliminary event feature set; selecting a typical key event as a basic event constituting the scene; dividing the risk degree of each basic event in the possible occurrence of a disaster accident according to a grade to establish a basic event risk degree grading table; constructing a basic event scene database according to the basic event risk degree grading table; and abstracting leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making. The application realizes intelligent early warning and rapid response to potential accidents of liquefied natural gas storage and transportation facilities, improves the scientificity and effectiveness of emergency management, can significantly reduce the accident occurrence probability and loss degree, and provides technical support for guaranteeing the safety of liquefied natural gas storage and transportation.

Description

Technical Field

[0001] This invention belongs to the field of liquefied natural gas safety technology, specifically relating to a method for analyzing leakage accidents at liquefied natural gas receiving stations. Background Technology

[0002] In recent years, impacted by the oil crisis and increasingly severe environmental problems caused by coal and oil, liquefied natural gas (LNG), as a high-quality clean energy source, has become one of the fastest-growing fuels globally. However, the rapid development of the LNG industry also presents significant safety hazards; accidents such as ruptures and leaks can have extremely serious consequences. Therefore, timely, scientific, and effective emergency decision-making and handling in the event of tank leaks are crucial issues that urgently need to be addressed.

[0003] Research on the concept of emergency decision-making began relatively early, and scholars have developed their own distinct perspectives through discussions on its connotation and characteristics. Fan Weicheng believes that emergency decision-making refers to formulating a decision-making plan after a sudden event occurs, using decision theory and computer-aided tools, and adjusting it according to the development of the event until the rescue operation concludes. Xue Lan believes that emergency decision-making is a series of measures to control the situation and reduce losses when information and resources are limited during an accident. Zhang Sichen et al., using diketene leaks as a case study, constructed an expert decision-making model for the entire process of emergency management of hazardous chemical leak accidents based on dynamic Bayesian networks. Song Xu et al.'s emergency model for natural gas leaks and fires at gas storage stations can not only achieve early emergency response to accidents but also identify the risk propagation paths of accidents.

[0004] Regarding the concept and application of "scenario-response," Porter argues that a scenario is a series of self-consistent viewpoints about possible future outcomes. Zhang Weirong et al. proposed a multi-module adaptive ventilation (MAV) method and studied its performance in controlling pollutant diffusion and removing pollutants in buildings across various scenarios. Wang Jiajun et al. proposed a method for assessing the satisfaction of Natech risk response needs in urban areas; only when all types of risk response needs are met can emergency response levels be better improved. Li Jie et al., through constructing a catastrophic scenario for VOCs treatment systems, found that enterprises still have shortcomings in emergency response responsibilities and should focus on building emergency response capabilities. Chen et al. conducted dynamic analysis of specific accident cases based on a scenario-response evolution model. Kathleen analyzed the evolution mechanism of sudden accidents based on a "scenario-response" type emergency decision-making model. Dormer analyzed how the evolution of scenarios affects decision-makers' selection and formulation of emergency response plans. Li Wei applied the "scenario-response" emergency decision-making method to conduct an empirical study on material needs and reserves under different natural disaster scenarios. Yang Jingzhao et al., in their study within the "scenario-coping" framework, found that temporary teams can influence team effectiveness through team cohesion in extreme situations.

[0005] However, existing emergency decision-making research mainly relies on calling upon and referencing pre-established or historical accident decision-making plans, lacking research on real-time emergency response decision-making. Furthermore, insufficient understanding of the dynamic evolution of complex accident scenarios makes it difficult to fully capture the interactive influence of multiple factors during an accident, thus limiting the accuracy and timeliness of emergency decision-making.

[0006] A deeper analysis of the challenges in this field reveals that the core issue lies in how to effectively construct a dynamic correlation mechanism for accident scenarios. First, since accidents often involve a chain reaction of multiple basic events, such as leaks, fires, and explosions, it is difficult to predict the possible paths of accident development without accurately identifying the logical relationships between these events. Second, this lack of logical relationships further leads to a lack of systematic support for emergency decision-making; decision-makers often have to rely on experience when faced with complex scenarios, rather than quickly formulating targeted countermeasures. These two aspects are closely related; the former directly affects the latter, forming a pressing technical challenge that needs to be addressed.

[0007] Therefore, how to integrate multiple basic events to construct a scenario database that can reflect the dynamic evolution of accidents, and on this basis, achieve scientific and precise emergency decision-making, has become a key research issue. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method for analyzing leakage accidents at liquefied natural gas receiving stations, which addresses the shortcomings of the prior art. This method enables intelligent early warning, rapid response, and precise control of potential accidents at liquefied natural gas storage and transportation facilities, improves the scientific nature and effectiveness of emergency management, significantly reduces the probability of accidents and the extent of losses, and provides technical support for ensuring the safety of liquefied natural gas storage and transportation.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for analyzing leakage accidents at liquefied natural gas receiving stations, the method comprising the following steps: Step S1: Obtain basic event records of different types in the leak accident through the pre-established accident history archive of liquefied natural gas storage and transportation facilities. Extract the triggering conditions and subsequent impact characteristics of each type of event to obtain a preliminary set of event characteristics. Step S2: Based on the preliminary set of event characteristics, the event interaction analysis method is used to sort out the logical relationship between basic events. Combining conditional triggering logic and scenario interaction mode, the possibility of chain reaction of each event in different environments is determined, and a dynamic correlation network of accident scenarios is constructed. Step S3: Select typical key events in the accident cases that affect the evolution of the scenario and have a high frequency of occurrence and a coupling relationship of common force, mutual force and driving force in the scenario evolution as the basic events constituting the scenario. Step S4: Classify the degree of danger of each basic event in the possible occurrence of disasters and accidents according to the level, and establish a basic event danger level classification table; Step S5: Based on the basic event hazard level classification table, the basic events are based on the scenario evolution of historical accident cases in chronological order. The scenario evolution process and the relationship between basic events are displayed through the hierarchical relationship. Basic events have temporal characteristics. The value of the basic event in the current scenario is determined only by the previous scenario, thus constructing a basic event scenario database. Step S6: Abstract the leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making.

[0010] The above-mentioned method for analyzing liquefied natural gas receiving terminal leakage accidents includes, after step S6: Step S7: Visualize and analyze the scenario in the form of images or videos.

[0011] The above-mentioned method for analyzing liquefied natural gas (LNG) receiving terminal leakage accidents, in step S1, involves obtaining records of different types of basic events in the leakage accident from a pre-established historical accident archive of LNG storage and transportation facilities. For each type of event, the triggering conditions and subsequent impact characteristics are extracted to obtain a preliminary set of event features. The specific process is as follows: Records of leaks, fires, and explosions are extracted from accident history archives, and a basic event dataset containing event type, timestamp, and location information is generated using structured data parsing methods. For the basic event dataset, triggering conditions are extracted, and a conditional filtering algorithm is used to determine whether the temperature range and pressure range exceed the preset thresholds to obtain a set of triggering conditions. Based on the set of triggering conditions, the subsequent impact of the event is analyzed. Spatial analysis methods are used to calculate the diffusion range of the leakage event, the damage level of the fire event, and the impact range of the explosion event, and a set of impact characteristics is generated. From the set of impact features, the diffusion range, degree of damage, and impact range are extracted. Cluster analysis algorithms are used to determine the similarity between event features and generate a preliminary set of event features.

[0012] The above-mentioned method for analyzing liquefied natural gas receiving terminal leakage accidents is characterized in that: in step S2, based on a preliminary set of event characteristics, an event interaction analysis method is used to sort out the logical relationships between basic events, and combined with conditional triggering logic and scenario interaction mode, the possibility of chain reactions of each event under different environments is determined, thus constructing a dynamic correlation network of accident scenarios; the specific process is as follows: By classifying the event features in the initial event feature set, data mining techniques are used to extract key feature information, and an initial dataset is constructed for basic events to obtain a structured description of the event features. Based on the structured description, the logical relationships between basic events are sorted out using the interaction analysis method. If the correlation between a certain event and other events exceeds a preset threshold, it is marked as a key node, and the priority ranking of key nodes is determined. By prioritizing key nodes and combining them with conditional triggering logic, the scenario interaction mode is simulated. If a certain condition triggers a chain reaction of multiple events, the reaction path is recorded to obtain the preliminary sequence of the chain reaction. Based on the preliminary sequence of the chain reaction, the impact of environmental differences on the reaction path is analyzed, and the random forest algorithm is used to predict the probability of the reaction under different environments and determine the probability of each path occurring in a specific environment. Based on the reaction probability prediction results, a dynamic network model is constructed for the accident scenario. If the probability of occurrence of a certain path is higher than a preset threshold, it is included in the core scenario branch to determine the core structure of the dynamic network. Based on the core structure of dynamic networks, iterative simulations are performed on the chain reactions of various events under different environments to obtain the evolution trend of the reaction path and obtain a complete dynamic correlation network of the accident scenario.

[0013] The basic events selected in step S3 of the above-mentioned liquefied natural gas receiving terminal leakage accident analysis method include: E1, LNG leakage; E 2、 Fire; E 3、 Explosion; E 4、 Tank wall rupture; E 5、 Failure of liquid inlet system equipment and facilities; E 6、 Tank outlet equipment and facilities failed; E 7、 An ignition source appears; E 8、 Storage tank overpressure; E 9、 Emergency decision-making and measures.

[0014] In the above-mentioned method for analyzing liquefied natural gas receiving station leakage accidents, step S4 describes classifying the degree of danger of each basic event in the possible occurrence of disaster accidents according to the level, and establishing a basic event danger level classification table. The danger and impact of the same event are defined in four levels, and the same level is further divided into intervals of [1-10] according to its impact.

[0015] The above-mentioned method for analyzing liquefied natural gas receiving terminal leakage accidents, in step S6 above, abstracts the leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making, and includes the following steps: The four possible values ​​of the basic event are defined in matrix form: Level 1 = [1 0 0 0], Level 2 = [0 1 00], Level 3 = [0 0 1 0], Level 4 = [0 0 0 1]; The one-step prediction result is obtained by multiplying the real-time input state with the probability transition matrix; let the initial value be matrix A and the transition matrix be B, and the one-step prediction formula be expressed as: A(predicted value) = A(initial value) B; By selecting the states of each basic event in the scenario at time t as input values, the occurrence or impact of each basic event in the scenario at time t+1 is predicted.

[0016] In the above-mentioned LNG receiving terminal leakage accident analysis method, when performing visualization analysis of the scenario in the form of images or videos in step S7, Fluent fluid simulation software is used. Based on the analysis of the leakage diffusion law of LNG storage tanks and combined with the analysis of historical accident cases, multiple typical LNG storage tank leakage scenarios are used as examples for scenario simulation analysis. By changing the leakage hole diameter, leakage hole location, and leakage hole height parameters, all leakage scenario simulation results are stored in the scenario database as a visual scenario analysis method in accident emergency decision-making.

[0017] The above-mentioned method for analyzing liquefied natural gas receiving terminal leakage accidents includes typical LNG storage tank leakage scenarios such as leakage of inlet equipment, leakage of outlet equipment, leakage of pipelines in the tank, leakage of tank wall rupture, and leakage of two tank walls rupture.

[0018] Compared with the prior art, the present invention has the following advantages: 1. This invention constructs a dynamic accident correlation network by combining event interaction analysis with the random forest algorithm, clarifying the evolutionary logic of the chain coupling between events such as leakage, ignition, and explosion, and making up for the shortcomings of traditional technologies that cannot carry out multi-factor linkage analysis and are difficult to grasp the dynamic evolution law of accidents.

[0019] 2. This invention sets up a unified quantitative evaluation system with four basic hazard levels and subdivision scales of 1 to 10, classifying the degree of harm of nine types of core accident basic events, changing the previous problem of relying solely on qualitative methods to judge risks and failing to accurately locate the source of risks.

[0020] 3. This invention constructs a time-series scenario database and introduces a Markov chain model to describe the time-series transmission characteristics of accidents. It can predict the probability of various risks occurring in the next stage based on the real-time accident status and predict serious derivative accidents in advance, thus solving the shortcomings of existing technologies that lack real-time dynamic prediction and can only refer to static historical plans.

[0021] 4. This invention relies on Fluent fluid simulation to simulate various typical leakage conditions and store visual cloud images of gas cloud diffusion. It combines quantitative numerical analysis with intuitive image analysis to improve the shortcomings of pure numerical prediction, which is abstract and difficult to understand, and the inability of on-site command to quickly determine the danger range.

[0022] 5. This invention can quantify the inhibitory effect of emergency response on disaster development, output targeted response plans based on predicted scenarios, get rid of the limitations of experience-based emergency response being lagging and blind, can not only provide early warning of potential hazards to reduce the accident rate, but also assist on-site rescue to reduce losses after an accident occurs, comprehensively improve the scientific level of emergency management of liquefied natural gas receiving stations, and provide a complete set of quantitative technical support for LNG storage and transportation safety.

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the method in Embodiment 2 of the present invention; Figure 3A This is a 30-second leakage cloud map of scenario 1 in a specific application scenario of the present invention; Figure 3BThis is a 100s leakage cloud map of scenario 2 in the specific application scenario of this invention; Figure 3C This is a 200s leakage cloud map of scenario 3 in the specific application scenario of this invention; Figure 3D This is a 300s leakage cloud map of scenario 4 in the specific application scenario of this invention; Figure 4A Diagram showing the cut-out location of the pipeline in tank TK-02; Figure 4B This is a diagram showing the situation after the accident at storage tank TK-02. Detailed Implementation

[0025] Example 1

[0026] like Figure 1 As shown, the liquefied natural gas receiving station leakage accident analysis method of this embodiment includes the following steps: Step S1: Obtain basic event records of different types in the leak accident through the pre-established accident history archive of liquefied natural gas storage and transportation facilities. Extract the triggering conditions and subsequent impact characteristics of each type of event to obtain a preliminary set of event characteristics. In this embodiment, step S1 involves obtaining basic event records of different types of leak accidents from a pre-established historical archive of liquefied natural gas storage and transportation facilities, extracting the triggering conditions and subsequent impact characteristics for each type of event, and obtaining a preliminary set of event features. The specific process is as follows: Records of leaks, fires, and explosions are extracted from accident history archives, and a basic event dataset containing event type, timestamp, and location information is generated using structured data parsing methods. For the basic event dataset, triggering conditions are extracted, and a conditional filtering algorithm is used to determine whether the temperature range and pressure range exceed the preset thresholds to obtain a set of triggering conditions. Based on the set of triggering conditions, the subsequent impact of the event is analyzed. Spatial analysis methods are used to calculate the diffusion range of the leakage event, the damage level of the fire event, and the impact range of the explosion event, and a set of impact characteristics is generated. From the set of impact features, the diffusion range, degree of damage, and impact range are extracted. Cluster analysis algorithms are used to determine the similarity between event features and generate a preliminary set of event features.

[0027] In this embodiment, the basic events of historical LNG storage tank leakage accidents are summarized in Table 1: Table 1. Summary of Basic Events in Historical LNG Tank Leakage Accidents

[0028] Table 1 shows that accidents can be caused by many basic events, and leaks, fires, and explosions are not caused by just one basic event; they may be the result of many basic events occurring together or in a chain reaction. Therefore, it is necessary to sort out and analyze the basic relationships between these events.

[0029] Step S2: Based on the preliminary set of event characteristics, the event interaction analysis method is used to sort out the logical relationship between basic events. Combining conditional triggering logic and scenario interaction mode, the possibility of chain reaction of each event in different environments is determined, and a dynamic correlation network of accident scenarios is constructed. In this embodiment, step S2 involves using an event interaction analysis method based on a preliminary set of event features to clarify the logical relationships between basic events. This is combined with conditional triggering logic and scenario interaction patterns to determine the likelihood of chain reactions for each event under different environments, thus constructing a dynamic correlation network of accident scenarios. The specific process is as follows: By classifying the event features in the initial event feature set, data mining techniques are used to extract key feature information, and an initial dataset is constructed for basic events to obtain a structured description of the event features. Based on the structured description, the logical relationships between basic events are sorted out using the interaction analysis method. If the correlation between a certain event and other events exceeds a preset threshold, it is marked as a key node, and the priority ranking of key nodes is determined. By prioritizing key nodes and combining them with conditional triggering logic, the scenario interaction mode is simulated. If a certain condition triggers a chain reaction of multiple events, the reaction path is recorded to obtain the preliminary sequence of the chain reaction. Based on the preliminary sequence of the chain reaction, the impact of environmental differences on the reaction path is analyzed, and the random forest algorithm is used to predict the probability of the reaction under different environments and determine the probability of each path occurring in a specific environment. Based on the reaction probability prediction results, a dynamic network model is constructed for the accident scenario. If the probability of occurrence of a certain path is higher than a preset threshold, it is included in the core scenario branch to determine the core structure of the dynamic network. Based on the core structure of the dynamic network, the chain reaction of each event under different environments is iteratively simulated to obtain the evolution trend of the reaction path and obtain the complete dynamic correlation network of the accident scenario.

[0030] Step S3: Select typical key events in the accident cases that affect the evolution of the scenario and have a high frequency of occurrence and a coupling relationship of common force, mutual force and driving force in the scenario evolution as the basic events constituting the scenario. In this embodiment, the logical relationships between the basic events are summarized in Table 2: Table 2 Logical Relationships Between Basic Events

[0031] In this embodiment, the basic events selected in step S3 to constitute the scenario include nine typical key events: E1, LNG leakage; E 2、 Fire; E 3、 Explosion; E 4、 Tank wall rupture; E 5、 Failure of liquid inlet system equipment and facilities; E 6、 Tank outlet equipment and facilities failed; E 7、 An ignition source appears; E 8、 Storage tank overpressure; E 9、 Emergency decision-making and measures.

[0032] Step S4: Classify the degree of danger of each basic event in the possible occurrence of disasters and accidents according to the level, and establish a basic event danger level classification table; In this embodiment, when establishing a basic event danger level classification table by classifying the degree of danger of each basic event in the possible occurrence of disasters and accidents in step S4, a total of 4 levels are defined for the danger and impact of the same event, and the same level is further divided into intervals of [1-10] according to its impact.

[0033] In this embodiment, the degree of influence of basic events at the same level is divided as shown in Table 3: Table 3. Impact Level of Basic Events at the Same Level

[0034] In this embodiment, the basic event hazard level classification table is shown in Table 4: Table 4 Classification of Risk Levels

[0035] Step S5: Based on the basic event hazard level classification table, the basic events are based on the scenario evolution of historical accident cases in chronological order. The scenario evolution process and the relationship between basic events are displayed through the hierarchical relationship. Basic events have temporal characteristics. The value of the basic event in the current scenario is determined only by the previous scenario, thus constructing a basic event scenario database. As an important part of the emergency decision support system, a comprehensive basic event scenario database provides scientific guidance for emergency decision-making.

[0036] In this embodiment, the basic event database is organized as shown in Table 5: Table 5. Basic Event Database Organization Format

[0037] Table 5 shows nine basic events, based on the chronological evolution of historical accident cases. The evolution of scenarios and the relationships between basic events are illustrated through hierarchical relationships. At this point, all basic events exhibit temporal characteristics, and the values ​​of basic events in the current scenario are determined solely by the previous scenario. This process establishes a basic event scenario database, which serves as the basic input for Markov chain models in accident emergency decision-making for prediction and analysis.

[0038] Step S6: Abstract the leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making.

[0039] In this embodiment, step S6, which involves abstracting the leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making, includes the following steps: The four possible values ​​of the basic event are defined in matrix form: Level 1 = [1 0 0 0], Level 2 = [0 1 00], Level 3 = [0 0 1 0], Level 4 = [0 0 0 1]; The one-step prediction result is obtained by multiplying the real-time input state with the probability transition matrix; let the initial value be matrix A and the transition matrix be B, and the one-step prediction formula be expressed as: A(predicted value) = A(initial value) B; By selecting the states of each basic event in the scenario at time t as input values, the occurrence or impact of each basic event in the scenario at time t+1 is predicted.

[0040] In this embodiment, examples of scenario prediction results are shown in Table 6: Table 6 Examples of Scenario Prediction Results

[0041] Example 2

[0042] like Figure 2 As shown, this embodiment differs from Embodiment 1 in that: after step S6, it further includes: Step S7: Visualize and analyze the scenario in the form of images or videos.

[0043] The remaining methods are the same as in Example 1.

[0044] In emergency response to sudden accidents, relying solely on theoretical reasoning and numerical prediction for decision-making is incomplete. Therefore, simultaneously predicting the direction of scenario evolution and visually analyzing the scenario using images or videos, combining quantitative and qualitative analysis, can make emergency response decisions more scenario-specific and provide sufficient theoretical guidance.

[0045] In this embodiment, when performing visual analysis of the scenario in the form of images or videos in step S7, Fluent fluid simulation software is used. Based on the analysis of the leakage and diffusion law of LNG storage tanks and combined with the analysis of historical accident cases, a variety of typical LNG storage tank leakage scenarios are used as the scenario simulation analysis objects. By changing the leakage hole diameter, leakage hole position, and leakage hole height parameters, all leakage scenario simulation results are stored in the scenario database as a visual scenario analysis method in accident emergency decision-making.

[0046] In this embodiment, the typical LNG storage tank leakage scenarios include leakage of the inlet equipment, leakage of the outlet equipment, leakage of pipelines in the tank, leakage of tank wall rupture, and leakage of two tank walls rupture.

[0047] To verify the technical effectiveness of this invention, the "11.2 Major Fire Accident at the Beihai LNG Receiving Terminal" was selected as a case study. The method of this invention was applied, selecting nine basic events as the main analytical objects constituting the scenario. Based on a scenario database, a Markov chain mathematical model was used to achieve scenario prediction. Simultaneously, scenario visualization analysis was combined to conduct responsive emergency decision-making and response measures. The dynamic emergency decision-making process from the initial occurrence of the accident to its complete resolution was simulated and demonstrated. The specific process is as follows: Accident Description: On the morning of November 2, 2020, workers carried out branch pipeline construction on the storage tank and performed hot work on the LNG export pipeline on the second-level platform in front of the storage tank. After cutting off a short section on the low-pressure pump outlet main pipe, a tee pipe was added.

[0048] The worker operated the equipment alone without supervision from relevant management personnel. At 11:44:48, the worker operated the SIS (Safety Instrumented System) to forcibly close the valve, which then reopened, causing LNG to spray out. Approximately 10 seconds after the spray, a fire broke out on the platform in front of the storage tank. At the time of the LNG spray and fire, there were 8 workers on the platform in front of the storage tank and 1 worker on the tank top. After the incident, Beihai LNG Company activated its emergency plan, and the company's fire brigade dispatched 4 fire trucks and 25 firefighters to the storage tank area to fight the fire. The process team urgently shut down the ESD (Emergency Shutdown Device) in the storage tank area. Due to the forced SIS system, the ESD did not activate, and the sprinkler system was subsequently activated. After receiving the alarm, the Beihai Fire and Rescue Brigade dispatched 18 fire trucks and 58 firefighters from Jingang, Nankang, Xinggang, Special Service, and Nanzhu stations to the scene. Sinopec Beihai Refining & Chemical dispatched 15 firefighters, 1 high-pressure water jet fire truck, 2 foam fire trucks, and 1 gas protection vehicle to the accident site to assist in emergency rescue and disposal. At 11:51:59, a short circuit occurred in the valve control circuit, the fuse in the cabinet blew, the valve lost power and closed, and the open flame in front of the tank was extinguished. At 13:36, the Beihai Municipal Emergency Management Bureau, Fire and Rescue Brigade, Market Supervision Bureau, Tieshangang District Government, and Beihai LNG Company established an accident site rescue command center and quickly organized personnel search and rescue, tank temperature monitoring, pressure relief and discharge of connected pipelines, and transfer of filling vehicles. As of 19:35, rescuers had rescued 8 people from the accident site. The emergency rescue operation was thus concluded.

[0049] Accident analysis: The direct cause of the accident was: during hot work cutting operations on the low-pressure pump outlet main pipe of the second-level platform of the storage tank, the isolation valve was opened. The storage tank valve did not employ a physical isolation method to isolate the power source, but instead used an instrument logic isolation method. This isolation failed during the valve closing operation, causing LNG in the low-pressure export pipeline to spray out from the cut. After the LNG leaked and mixed with air, it ignited upon contact with ignition energy, triggering a fire. The ignition source was the ignition energy generated by cables whose insulation protective layer had become brittle and detached after being impacted by the low-temperature LNG jet.

[0050] In order to realize the application of the emergency decision-making model of the present invention, it is first necessary to unfold the target accident in the form of scenario evolution in chronological order. Table 7 is the accident scenario evolution table of the "11.2" major fire accident at Beihai LNG receiving station over time.

[0051] Table 7. Scenario Evolution of the Beihai "11.2" Accident

[0052] Application of the method of this invention: Before applying emergency decision-making models and conducting predictive analysis of accident development, it is first necessary to determine the occurrence and severity of basic events in the initial scenario. This invention selects nine basic events as examples to simplify the analysis of the basic scenarios constituting the accident situation.

[0053] When an accident occurs, Scenario 1 is used as the input scenario, and the basic event components and levels of the scenario are analyzed. Table 8 shows the scenario information and emergency measures in Scenario 1.

[0054] Table 8 Scenario 1 Scenario Information

[0055] As shown in Table 8, Scenario 1, the values ​​of two basic events have changed in the current scenario. First, the severity grading of the basic events is analyzed to classify them into levels, which are then input into the Markov chain model for scenario prediction.

[0056] E1 Leakage Incident: LNG was ejected from the external output pipeline when the valve was opened, classified as Level 3; E6 Tank Outlet System Equipment Failure Incident: Valve failure caused by operator error, classified as Level 2; Other basic events are still in a state of not occurring or their values ​​have changed, classified as Level 1; E9 Emergency Decision and Response Incident: Emergency response measures have not been implemented in a timely manner after the accident, classified as Level 3.

[0057] By analyzing the scenarios, the judgment levels of each basic event are input into the Markov chain model to predict the next scenario. Table 9 shows the prediction results for scenario 2.

[0058] Table 9 Prediction Results for Scenario 2

[0059] In Table 9, the "Scenario 1" column shows the basic event states under Scenario 1, and the "Prediction" column shows the basic event states under Scenario 2 predicted based on the basic event states under Scenario 1. The same applies below.

[0060] The prediction results of Scenario 2 show the possible future scenarios. At this point, the levels of the basic events that constitute the scenario should be analyzed.

[0061] E1 Leakage Event Level 3: The pipeline valve is still leaking continuously; E2 Fire Event Level 3: A fire of significant danger has occurred; E3 Explosion Event Level 2: A physical explosion of equipment or facilities is possible; E4 Tank Wall Rupture Event Level 1: The danger of a fire or explosion has not yet affected the tank to the point of rupture or damage; E5 Tank Inlet System Equipment Failure Event Level 1: The outlet system equipment or facilities are malfunctioning but have not yet significantly affected the tank; E6 Tank Outlet System Equipment Failure Event Level 3: The tank outlet system equipment is malfunctioning; E7 Ignition Source Event Level 2: An ignition source event of moderate danger has occurred; E8 Tank Overpressure Event Level 2: The malfunction of the tank outlet equipment or facilities may affect the pressure balance of the tank; E9 Emergency Decision-Making and Measures Event Level 2: Some emergency decisions and measures have been made, but with a certain degree of lag.

[0062] Before making emergency decisions, it is necessary to analyze the visualized information of the scenarios. Since Scenario 1 and predicted Scenario 2 involve the same leakage scenario—a leak in the tank outlet system equipment—the leakage cloud map of this scenario is analyzed by accessing the visualized data in the scenario database to determine the trajectory of the accident. The leakage cloud map is shown below. Figure 3A , Figure 3B , Figure 3C , Figure 3D As shown.

[0063] Cloud map analysis indicates a leak at the bottom of the tank, specifically a valve or LNG pipeline. The resulting gas cloud will rapidly expand over time, primarily spreading near the ground. If workers approach the low temperatures at this point, frostbite or other injuries may occur. Simultaneously, the explosive gas cloud will accumulate mainly in the center of the cloud and on both sides of the tank. After 50 seconds of leakage, the width of the gas cloud will gradually increase, and the explosive gas cloud will accumulate near the process area. If an ignition source is found within this hazard zone, it could trigger a fire or explosion. Therefore, it is crucial to address the leaking gas cloud promptly and monitor its extent and movement in real time. Dilute any areas where the explosive gas cloud may be present. Furthermore, the leak diffusion pattern cloud map should guide the evacuation of personnel and direct real-time rescue and response measures for firefighters.

[0064] At this point, based on real-time, predictive, and visualized scenario data, decision-makers should promptly arrange and implement a series of emergency decision-making measures, as shown in Table 10.

[0065] Table 10 Emergency Decision-Making Measures for Scenario 1

[0066] In the Beihai "11.2" accident, the accident occurred suddenly and developed rapidly. The workers did not take immediate action at the accident site, and the scenario evolved into the next scenario of escalation. At this point, according to the accident emergency decision-making process, the evolved real-time scenario 2 should be used as the input scenario to continue a new round of model application. Table 11 shows the scenario information in scenario 2 of the target accident case.

[0067] Table 11 Scenario 2 Scenario Information

[0068] At this point, the scenario information in Scenario 2 should be analyzed promptly, the basic events should be graded, and predictions should continue for the next scenario.

[0069] Scenario 2 involves four basic event values ​​that change. E1 is a leakage event where the valve remains open and LNG continues to leak from the pipeline, classified as Level 3. E2 is a fire event where the LNG leak causes a fire on the platform in front of the storage tank upon contact with an ignition source, posing a significant fire hazard, also classified as Level 3. E6 is a storage tank outlet system equipment failure event where the valve system remains faulty but has not yet experienced further failure or damage, classified as Level 2. E7 is an ignition source event where cables, whose insulation has become brittle and detached after being impacted by cryogenic LNG injection, generate ignition energy, posing a significant danger, classified as Level 3. E9 is an emergency decision-making and response event where no corresponding emergency decision-making or response measures have been implemented despite the accident evolving to a certain extent, classified as Level 3. Other basic events remain in a state of non-occurrence or their values ​​change, classified as Level 1.

[0070] At the same time, the actual and predicted scenarios should also be compared. Table 12 shows the comparison between the predicted and actual scenarios for scenario 2.

[0071] Table 12 Comparison of Scenario 2 Predictions and Actual Scenario

[0072] Analysis comparing the actual and predicted scenarios reveals that, except for the E7 ignition source event, the basic event levels of the predicted scenarios are generally consistent with, or even more dangerous than, the actual scenarios. Therefore, the method of this invention has high reliability during this prediction phase. However, in the predicted scenario, certain emergency decisions and measures would be taken for the E9 emergency decision-making and response event; in the actual scenario, due to the rapid occurrence of the accident, no corresponding measures were taken.

[0073] At this point, we should promptly and continuously conduct predictive analysis on the next scenario, using scenario 2 as the real-time input scenario to predict scenario 3, as shown in Table 13.

[0074] Table 13 Prediction Results for Scenario 3

[0075] Predicting scenario 3 is based on the information from scenario 2 and is derived solely from the state evolution of scenario 2. Therefore, it is necessary to analyze the various basic events in predicting scenario 3.

[0076] E1 Leakage Event Level 3: Pipeline or valve remains leaking; E2 Fire Event Level 3: Fire may continue to spread and remains at a high level of danger; E3 Explosion Event Level 2: Due to the impact of the fire, it is very likely to encounter a flammable explosive gas cloud, triggering a serious chain fire and explosion event; E4 Tank Wall Rupture Event Level 1: The danger and severity of the fire or explosion event have not yet affected the local rupture of the tank; E5 Tank Inlet System Equipment Failure Event Level 1: The relevant facilities and equipment at the tank inlet are still operating normally or have not yet malfunctioned; E6 Tank Outlet System Equipment Failure Event Level 4: If a fire or explosion occurs in the tank area, it will affect the equipment and facilities around the tank, causing damage; E7 Ignition Source Event Level 3: Due to the gradual development of the fire event and its determination as Level 3, it has a high level of danger and will form a large-scale ignition energy; E8 Tank Overpressure Event Level 2: Due to the impact of the fire event, the tank will be affected by the heat radiation of the flame, causing an increase in internal pressure; E9 Emergency Decision and Measures Event Level 2: Some emergency decision and measures have been taken, but there is a certain lag.

[0077] Based on real-time and predicted scenario analysis, the leak location remains at the LNG pipeline valve outside the storage tank. To prevent damage to other equipment and facilities, decision-makers should continue analysis based on the leak scenario cloud map at the bottom of the storage tank. However, seven minutes have already passed. If no corresponding emergency decisions and measures are taken, the leak may have already covered the entire plant area, and highly explosive LNG gas clouds may form in some locations. At this point, analysis should be conducted based on the real-time and predicted scenarios, and emergency decisions and measures should be taken promptly for the basic events in the scenarios, as shown in Table 14.

[0078] Table 14 Emergency Decision-Making Measures for Scenario 2

[0079] In the Beihai "11.2" accident, the company promptly activated its emergency response plan and organized fire and rescue forces to handle the on-site accident, bringing it under control and allowing the accident scenario to evolve in an ideal direction. The real-time scenario 3 information is shown in Table 15 below.

[0080] Table 15 Scenario 3 Scenario Information

[0081] Analysis of Table 15 shows that Scenario 3 primarily involves changes in the values ​​of four basic events: E1 Leakage Event, where the valve shuts off due to power failure, but a large amount of LNG has already leaked, thus classified as Level 2; E2 Fire Event, where the open flame at the tank is extinguished due to timely firefighting and handling of related incidents, thus classified as Level 1; E6 Tank Outlet System Equipment Failure Event, where the valve control circuit of the external pipeline system short-circuits, thus classified as Level 3; E7 Ignition Source Event, where the open flame is extinguished and no further ignition source appears due to proper emergency response measures, thus classified as Level 1; E8 Tank Overpressure Event, where the pressure in the tank increases due to heat radiation from the flames surrounding the tank, thus classified as Level 2; and E9 Emergency Decision and Response Event, where the company promptly activates the emergency plan and deploys rescue forces to carry out emergency response, taking corresponding measures for the basic events that have occurred, thus classified as Level 1. Other basic events remain in a state of not occurring or their values ​​do not change, thus classified as Level 1.

[0082] Table 16 compares the actual and predicted results for Scenario 3.

[0083] Table 16 Comparison of Scenario 3 Predictions and Actual Scenario

[0084] A comparison of the actual and predicted scenarios in Scenario 3 reveals that the timely emergency response measures implemented in Scenario 2 may have resulted in the various basic events in the predicted scenario being more severe than in the real-time scenario. However, in actual emergency decision-making, preparing emergency response measures in advance for potentially severe scenarios not only allows for more effective handling of the current incident but also helps to prevent derivative incidents and more quickly control the development of disasters within a certain scope.

[0085] At this point, the disaster is not yet completely over, and the risks of various basic events have not completely disappeared. Disasters may still recur due to human error or the existence of potential hazards, necessitating a new round of prediction and analysis. Using scenario 3 as the input scenario, prediction and analysis are conducted for scenario 4, as shown in Table 17.

[0086] Table 17 Prediction Results for Scenario 4

[0087] Based on the prediction results of Scenario 4, the levels of each basic event in the scenario are analyzed. E1 Leakage Event Level 3: The pipeline or valve is still leaking or there is still a localized gas cloud accumulation; E2 Fire Event Level 3: The fire may not be completely extinguished; if a flammable and explosive gas cloud still exists, the risk is high; E3 Explosion Event Level 2: There is still a possibility of a small-scale explosion; E4 Tank Wall Rupture Event Level 1: The danger and severity of events such as fire or explosion have not yet affected the localized rupture of the tank; E5 Tank Inlet System Equipment Failure Event Level 1: The relevant facilities and equipment at the tank inlet are not affected; E6 Tank Outlet System Equipment Failure Event Level 3: The LNG pipeline valves outside the tank may be affected by flames or have their own system problems and are still in a faulty state; E7 Ignition Source Event Level 3: If the fire has not completely extinguished, there is still an open flame, posing a significant risk; E8 Tank Overpressure Event Level 2: The tank is still in a state of uneven internal pressure; E9 Emergency Decision and Measures Event Level 2: Some emergency decisions and measures have been taken, but there is a certain lag.

[0088] Analysis of the predicted scenario 4 and the real-time scenario 3 shows that no other facility or equipment leakage scenarios have occurred. Therefore, appropriate emergency decisions still need to be made based on the leakage cloud map at the bottom of the storage tank, as shown in Table 18.

[0089] Table 18 Emergency Decision-Making Measures for Scenario 3

[0090] In the Beihai "11.2" accident, relevant departments in Beihai City immediately activated the emergency response, established an on-site rescue command center, and continued to direct and implement a series of post-accident handling measures after the accident had been brought under control, gradually bringing the accident to a standstill and the scenario further evolved in an ideal direction before ending. The following is a display of the real-time scenario 4 information, as shown in Table 19.

[0091] Table 19 Scenario 4 Scenario Information

[0092] The basic events of Scenario 4 are analyzed and compared with the predicted scenarios.

[0093] Analysis of Table 19 shows that the values ​​of the four basic events in Scenario 4 changed: E1 Leakage Event, the leaked gas cloud was basically dealt with or dissipated by atmospheric winds, classified as Level 1; E6 Tank Outlet System Equipment Failure Event, the equipment and facilities have been basically repaired and dealt with, classified as Level 1; E8 Tank Overpressure Event, the tank is subjected to heat radiation from the flame, and the pressure imbalance will continue for a short period of time, classified as Level 2; E9 Emergency Decision-Making and Measures Event, relevant departments promptly and effectively carried out emergency response and handling, classified as Level 1. The values ​​of the other basic events remained unchanged, classified as Level 1.

[0094] Table 20 compares the actual and predicted results for Scenario 4.

[0095] Table 20 Comparison of Scenario 4 Predictions and Actual Scenario

[0096] As shown in Table 20, the basic event levels in the predicted scenario are more dangerous than those in the actual scenario, but the level of each basic event in the predicted scenario still indicates that the event still has the possibility of occurring.

[0097] Ultimately, with the concerted efforts of relevant departments, all personnel were successfully rescued, and the rescue operation concluded. The investigation determined that the Beihai LNG "11.2" fire was caused by LNG spillage and subsequent fire during hot work on the low-pressure pump outlet manifold of the LNG storage tank, constituting a major production safety accident. The accident resulted in damage to all related equipment and facilities within the storage tank system, such as... Figure 4A , 4B As shown.

[0098] This invention establishes an emergency decision-making model that analyzes the entire process of an accident from its occurrence to its resolution from the decision-maker's perspective. Through comparative analysis of three predicted and actual scenarios, it is evident that the emergency decision-making model can predict and analyze potential scenarios and existing accident hazards, thus enabling proactive emergency decision-making. Furthermore, by using the basic events constituting the scenario as the primary analysis object, the model can accurately and quickly identify the specific causes of an accident's evolution and make corresponding emergency decisions, making it more targeted. Simultaneously, comparisons between the three sets of predicted and actual scenarios reveal that emergency decisions and measures are crucial to the evolution of the scenario. Timely and scientific emergency decisions can effectively control the development trend of the accident, guiding the scenario towards an ideal direction; while untimely emergency decisions may delay the handling of the accident, causing the scenario to develop into a more serious and dangerous disaster. When a sudden accident occurs, decision-makers must make correct decisions in a short time, which is a significant challenge. The emergency decision-making model constructed in this invention can provide decision-makers with scientific and accurate guidance, assisting them in making timely and accurate decisions.

[0099] Comparative analysis with target accident cases shows that the method of this invention can predict the evolution of a scenario when emergency decision-making measures are not timely or handled improperly. At the same time, it enables emergency decision-making measures to be proactive and more scenario-responsive, effectively controlling the development trend of accidents and preventing the further expansion of disasters.

[0100] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for analyzing leakage accidents at liquefied natural gas receiving terminals, characterized in that, The method includes the following steps: Step S1: Obtain basic event records of different types in the leak accident through the pre-established accident history archive of liquefied natural gas storage and transportation facilities. Extract the triggering conditions and subsequent impact characteristics of each type of event to obtain a preliminary set of event characteristics. Step S2: Based on the preliminary set of event characteristics, the event interaction analysis method is used to sort out the logical relationship between basic events. Combining conditional triggering logic and scenario interaction mode, the possibility of chain reaction of each event in different environments is determined, and a dynamic correlation network of accident scenarios is constructed. Step S3: Select typical key events in the accident cases that affect the evolution of the scenario and have a high frequency of occurrence and a coupling relationship of common force, mutual force and driving force in the scenario evolution as the basic events constituting the scenario. Step S4: Classify the degree of danger of each basic event in the possible occurrence of disasters and accidents according to the level, and establish a basic event danger level classification table; Step S5: Based on the basic event hazard level classification table, the basic events are based on the scenario evolution of historical accident cases in chronological order. The scenario evolution process and the relationship between basic events are displayed through the hierarchical relationship. Basic events have temporal characteristics. The value of the basic event in the current scenario is determined only by the previous scenario, thus constructing a basic event scenario database. Step S6: Abstract the leakage accident analysis and prediction into a Markov chain model for prediction and analysis in accident emergency decision-making.

2. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 1, characterized in that: Step S6 is followed by: Step S7: Visualize and analyze the scenario in the form of images or videos.

3. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 1 or 2, characterized in that: Step S1 describes obtaining basic event records of different types of leak accidents from a pre-established historical archive of liquefied natural gas storage and transportation facilities. For each type of event, the triggering conditions and subsequent impact characteristics are extracted to obtain a preliminary set of event features. The specific process is as follows: Records of leaks, fires, and explosions are extracted from accident history archives, and a basic event dataset containing event type, timestamp, and location information is generated using structured data parsing methods. For the basic event dataset, triggering conditions are extracted, and a conditional filtering algorithm is used to determine whether the temperature range and pressure range exceed the preset thresholds to obtain a set of triggering conditions. Based on the set of triggering conditions, the subsequent impact of the event is analyzed. Spatial analysis methods are used to calculate the diffusion range of the leakage event, the damage level of the fire event, and the impact range of the explosion event, and a set of impact characteristics is generated. From the set of impact features, the diffusion range, degree of damage, and impact range are extracted. Cluster analysis algorithms are used to determine the similarity between event features and generate a preliminary set of event features.

4. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 1 or 2, characterized in that: Step S2 describes the process of using an event interaction analysis method based on a preliminary set of event features to analyze the logical relationships between basic events. By combining conditional triggering logic and scenario interaction patterns, the likelihood of chain reactions for each event under different environments is determined, thus constructing a dynamic correlation network of accident scenarios. The specific process is as follows: By classifying the event features in the initial event feature set, data mining techniques are used to extract key feature information, and an initial dataset is constructed for basic events to obtain a structured description of the event features. Based on the structured description, the logical relationships between basic events are sorted out using the interaction analysis method. If the correlation between a certain event and other events exceeds a preset threshold, it is marked as a key node, and the priority ranking of key nodes is determined. By prioritizing key nodes and combining them with conditional triggering logic, the scenario interaction mode is simulated. If a certain condition triggers a chain reaction of multiple events, the reaction path is recorded to obtain the preliminary sequence of the chain reaction. Based on the preliminary sequence of the chain reaction, the impact of environmental differences on the reaction path is analyzed, and the random forest algorithm is used to predict the probability of the reaction under different environments and determine the probability of each path occurring in a specific environment. Based on the reaction probability prediction results, a dynamic network model is constructed for the accident scenario. If the probability of occurrence of a certain path is higher than a preset threshold, it is included in the core scenario branch to determine the core structure of the dynamic network. Based on the core structure of dynamic networks, iterative simulations are performed on the chain reactions of various events under different environments to obtain the evolution trend of the reaction path and obtain a complete dynamic correlation network of the accident scenario.

5. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 1 or 2, characterized in that: The basic events selected in step S3 to constitute the scenario include: E1, LNG leakage; E 2、 Fire; E 3、 Explosion; E 4、 Tank wall rupture; E 5、 Failure of liquid inlet system equipment and facilities; E 6、 Tank outlet equipment and facilities failed; E 7、 An ignition source appears; E 8、 Storage tank overpressure; E 9、 Emergency decision-making and measures.

6. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 1 or 2, characterized in that: In step S4, when classifying the degree of danger of each basic event in the possible occurrence of disasters and accidents according to the level, and establishing a basic event danger level classification table, the danger and impact of the same event are defined into 4 levels, and the same level is further divided into the interval [1-10] according to its impact.

7. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 6, characterized in that: Step S6, which involves abstracting the leak accident analysis and prediction into a Markov chain model for prediction and analysis in emergency response decision-making, includes the following steps: The four possible values ​​of the basic event are defined in matrix form: Level 1 = [1 0 0 0], Level 2 = [0 1 0 0], Level 3 = [0 0 1 0], Level 4 = [0 0 0 1]; The one-step prediction result is obtained by multiplying the real-time input state with the probability transition matrix; let the initial value be matrix A and the transition matrix be B, and the one-step prediction formula be expressed as: A(predicted value) = A(initial value) B; By selecting the states of each basic event in the scenario at time t as input values, the occurrence or impact of each basic event in the scenario at time t+1 is predicted.

8. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 2, characterized in that: When performing visualization analysis of the scenario in the form of images or videos as described in step S7, Fluent fluid simulation software is used. Based on the analysis of the leakage and diffusion law of LNG storage tanks and combined with the analysis of historical accident cases, a variety of typical LNG storage tank leakage scenarios are used as examples for scenario simulation analysis. By changing the leakage hole diameter, leakage hole location, and leakage hole height parameters, all leakage scenario simulation results are stored in the scenario database as a visual scenario analysis method in accident emergency decision-making.

9. The method for analyzing liquefied natural gas receiving station leakage accidents according to claim 8, characterized in that: The typical LNG storage tank leakage scenarios include leakage from the inlet equipment, leakage from the outlet equipment, leakage from pipelines within the tank, leakage from tank wall rupture, and leakage from both tank walls rupture.

Images