CN-122022634-A - Fusion strategy-based marine methanol-hydrogen energy risk assessment method and apparatus

CN122022634ACN 122022634 ACN122022634 ACN 122022634ACN-122022634-A

Abstract

The invention provides a fusion strategy-based marine methanol-to-hydrogen energy risk assessment method and a fusion strategy-based marine methanol-to-hydrogen energy risk assessment device, which relate to the technical field of system security engineering and risk management, and the risk assessment method comprises the steps of identifying unsafe control behaviors of a power system through system-level hazard analysis, and constructing a hazard causal chain from a bottom-layer physical failure mode to a top-layer system-level hazard by combining component-level failure analysis; the method comprises the steps of constructing a hierarchical causal hazard and operability knowledge graph based on a hazard causal chain, converting unstructured multimode data streams into a structured semantic event sequence in real time by multimode sensing on the running state of a power system, and matching and cognizing the structured semantic event, the causal hazard and the operability knowledge graph to output a diagnosis report on the current risk state of the power system so as to complete risk assessment.

Inventors

Lou Qianyun
CUI YAN
Qiao Shoucheng
LI LU
KONG ZEYU

Assignees

中国船级社

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (10)

1. The marine methanol-hydrogen energy risk assessment method based on the fusion strategy is characterized by comprising the following steps of: Identifying unsafe control behaviors of the power system through system-level hazard analysis, and constructing a hazard causal chain from a bottom-layer physical failure mode to a top-layer system-level hazard by combining component-level failure analysis; Constructing a hierarchical causal hazard and operability knowledge map based on a hazard causal chain; converting unstructured multimode data streams into a structured semantic event sequence in real time by multimode sensing of the running state of the power system; and matching and cognizing the structural semantic event, the causal hazard and the operability knowledge graph to output a diagnosis report of the current risk state of the power system, thereby completing risk assessment.
2. The fusion strategy-based marine methanol-to-hydrogen energy risk assessment method of claim 1, wherein constructing a hazard causal chain comprises: Defining a hazard cause and effect chain tuple To represent a risk conduction path from the underlying physical failure mode to the top-level hazard, which is structured as follows: Wherein, the A system level hazard that is ultimately associated with the path; is a violated system-level security constraint in the path; To cause Violated unsafe control behavior; To cause A specific system context that presents a hazard; As a set, all that can result in Context of the present invention The underlying physical failure modes from component level failure analysis that occur, each element in the collection Are all a binary group, defined as Wherein As a specific physical component of the device, Is a particular failure mode for the component.
3. The method for evaluating the risk of methanol to hydrogen energy for a ship based on a fusion strategy as set forth in claim 2, wherein the step of constructing a hazard causal chain further comprises controlling the behavior of unsafe conditions Aggregate quantification of risk of (2) by calculating the aggregate risk priority thereof To realize: Wherein, the Is that Severity of (2); Is that Is defined by the aggregate probability By a preset mapping function The calculation mode of the aggregation probability is as follows: Wherein, the Is a collection Middle (f) The probability of the occurrence of a physical failure mode, Is a collection The total number of independent failure modes; Is that The aggregate detection degree of (2) is calculated by the following steps: Wherein, the Is the first Detection of individual physical failure modes.
4. The method for evaluating the risk of marine methanol-to-hydrogen energy based on a fusion strategy according to claim 3, wherein the step of constructing a hierarchical causal hazard and operability knowledge graph comprises the steps of introducing a risk path RiskPath as a core node type, wherein each RiskPath node in the graph uniquely corresponds to one hazard causal link tuple ; A hierarchical graph structure centered on RiskPath nodes is constructed, wherein system-level jeopardy nodes are connected to corresponding RiskPath nodes through a has_path relationship, and RiskPath nodes are connected to all failure mode nodes capable of inducing the path through a has_cause relationship.
5. The method for evaluating marine methanol-to-hydrogen energy risk based on a fusion strategy of claim 4, wherein the real-time conversion of unstructured multimodal data streams into a structured semantic event sequence specifically comprises: the power system is arranged in a time window Fusing the internally acquired visual characteristic sequence and the acoustic characteristic sequence to obtain a fused characteristic sequence ; Pairs of time ordered attention classification networks Processing to obtain a set of semantic events in a predefined manner Multiple tag probability distribution vector on The calculation mode is as follows: Wherein, the Is that A time hidden state vector; Is that Attention weight of moment; Is a context vector obtained by attention weight aggregation; is a learnable parameter of the network; the function is activated for Sigmoid.
6. The method for evaluating marine methanol-to-hydrogen energy risk based on a fusion strategy of claim 5, wherein the transforming step further comprises performing space-time localization on the instantiated semantic event, wherein the step of spatially localization specifically comprises: acquiring the occurrence time of a localized event Is a visual space feature map of (1) ; Computing a classifier for instantiated events Output score of (2) Regarding the characteristic diagram of the first embodiment Individual channels Obtain the weight of each channel ; Class activation thermodynamic diagrams of generated events Wherein the method comprises the steps of Comparing thermodynamic diagrams with pre-calibrated two-dimensional bounding boxes of key components to determine physical positions of event occurrence : 。
7. The method for evaluating marine methanol-to-hydrogen energy risk based on fusion strategy of claim 6, wherein the matching and cognitive reasoning specifically comprises combining a real-time semantic event with a real-time semantic event Initial set of excitation points linked to knowledge graph Wherein the linking is achieved by semantic mapping of event types to failure modes, i.e. calculating event types Text embedding and failure modes of (a) Is described in text of (a) Cosine similarity between embeddings of (a) : And select the similarity Above a preset threshold Form part of an initial set of articulation points.
8. The method for risk assessment of methanol-hydrogen energy for a ship based on a fusion strategy as recited in claim 7, wherein the step of matching and cognitive reasoning further comprises the steps of determining a set of slave initial articulation points Each candidate causal path searched by starting Calculating a diagnostic confidence score thereof : Wherein, the The detection confidence of the real-time semantic event is obtained; Path cost for the candidate causal path; Is a real-time system state vector With context defined in the path Matching scores of (2); and selecting the path with the highest score as the diagnosis of the current risk state.
9. The method for risk assessment of marine methanol-to-hydrogen energy based on fusion strategy of claim 8 further comprising dynamic optimization of knowledge graph, and when a risk diagnosis is confirmed, a causal link tuple for hazard to optimal interpretation Corresponding causal path Edges of the has_cause type in And (5) carrying out weight updating: Wherein, the Is the updated edge weight; Is the original edge weight; is knowledge update learning rate; Is the diagnostic confidence of the optimal interpretation.
10. The marine methanol-to-hydrogen energy risk assessment device based on the fusion strategy is characterized by comprising: the hazard causal chain generation module is used for identifying unsafe control behaviors of the power system through system-level hazard analysis and constructing a hazard causal chain from a bottom-layer physical failure mode to a top-layer system-level hazard by combining component-level failure analysis; the knowledge map construction module is used for constructing a hierarchical causal hazard and operability knowledge map based on the hazard causal link; The semantic event sequence acquisition module is used for converting unstructured multimode data streams into a structured semantic event sequence in real time by carrying out multimode sensing on the running state of the power system; And the matching and cognitive reasoning module is used for matching and cognitive reasoning the structural semantic event, the causal hazard and the operability knowledge graph so as to output a diagnosis report of the current risk state of the power system, thereby completing risk assessment.

Description

Fusion strategy-based marine methanol-hydrogen energy risk assessment method and apparatus Technical Field The invention relates to the technical field of system security engineering and risk management, in particular to a marine methanol-hydrogen energy risk assessment method and device based on a fusion strategy. More specifically, the invention utilizes artificial intelligence and data processing technology to dynamically and closely loop risk identification, analysis and optimization of highly integrated and intelligent systems such as marine methanol reforming hydrogen production fuel cells. Background With the increasing global environmental protection and carbon emissions requirements, the shipping industry is experiencing a profound energy structure transformation. The novel clean energy power system represented by the methanol reforming hydrogen production fuel cell is becoming an important technical path for replacing the traditional fuel engine due to the advantages of high energy density, convenient fuel storage and transportation, low emission and the like. However, such a novel power system is not a simple superposition of conventional technologies, but a highly complex system integrating multiple technical fields of chemical reactions, thermal control, electrical engineering, and software algorithms. The system comprises a large number of physical components, sensors, actuators and complex control logic, and the subsystems have a tight and nonlinear coupling relation. This complexity makes operational security and risk management of the system a significant challenge. Traditional methods of security analysis for simple electromechanical systems have been difficult to fully adapt to such emerging systems. FMEA is a widely used basic method in existing risk assessment practices. The method adopts bottom-up induction logic to identify potential risks by analyzing the potential failure modes of each physical component in the system one by one and evaluating the final influence of the potential failure modes on the system functions. FMEA has the advantages of strong systematicness and wide coverage for identifying single-point faults caused by physical failure of hardware components. However, the limitations are also significant. The core assumption of FMEA is that the risk originates from physical damage to the component, which makes it difficult to effectively identify system level risks posed by unintended interactions between multiple normal components. Particularly, in the present day with increasing software definition functions, the danger caused by software logic defects, man-machine interaction errors or system design imperfections per se is often not fully revealed by the traditional FMEA method. To remedy the shortcomings of conventional methods such as FMEA, a systematic theory-based analysis method has been developed, wherein a Systematic Theoretical Process Analysis (STPA) is a representative one. In contrast to FMEA, STPA employs top-down deduction logic to treat the system as a dynamic, hierarchical control structure. It does not pay attention to whether the component fails or not, but to whether the control instruction issued by the controller to the controlled object is proper or not during the running process of the system. By identifying unsafe control actions (Unsafe Control Actions, UCA) that may cause the system to enter a dangerous state, STPA is able to efficiently analyze systematic risks caused by software bugs, complex interactions, organizational management, human factors, and the like. Although STPA exhibits a powerful analytical capability at the macro system level, it also suffers from its own drawbacks. STPA is relatively abstract and sometimes difficult to trace directly and carefully to the specific underlying physical failure mechanism that causes the unsafe control behavior to occur. In addition, the quality of the analysis result is highly dependent on the understanding depth of the analysis personnel on the control relation of the system and the modeling accuracy, and is not easy to directly fuse with quantitative fault data at the component level. More importantly, once the analysis is completed, the results are cured into a static report. It cannot be dynamically verified, revised and evolved based on the system's operational data in the real world. When the system is exposed to unexpected operation conditions or a new unknown fault mode occurs, the static risk model gradually fails, and continuous and effective support cannot be provided for the full life cycle safety guarantee of the system. Therefore, how to construct a risk management and control method which can integrate the advantages of different analysis methods and can perform self-adaptive learning and dynamic optimization by using actual operation data is a technical problem to be solved in the field. Disclosure of Invention The specification provides a marine methanol-hydrogen energy risk asses