Hydrogen energy accident human factor identification and risk trend prediction method fusing hfacs-hea and psogm

Abstract

This invention belongs to the field of hydrogen energy safety technology, specifically relating to a method for identifying human factors in hydrogen energy accidents and predicting risk trends by integrating HFACS-HEA and PSOGM. Addressing the problems of existing human factor analysis in hydrogen energy accidents—lack of a dedicated and systematic classification framework, lack of quantitative verification of causal relationships between factors, and low accuracy in risk trend prediction—this invention proposes the following solution: Constructing a hydrogen-specific HFACS-HEA model encompassing four levels: organizational influence, unsafe supervision, preconditions for unsafe behavior, and unsafe behavior; establishing a labeled dataset of human factors; using chi-square tests and concession ratio analysis to quantify the causal transmission relationships between factors at adjacent levels and identify critical paths in the causal chain; using grey relational analysis to screen key human factors requiring priority control; and employing particle swarm optimization to globally optimize the grey model parameters, building a PSOGM model to achieve accident trend prediction. This invention can be applied to the safety management of hydrogen energy production, storage, transportation, and refueling systems.

Description

Technical Field

[0001] This invention belongs to the field of hydrogen energy safety technology, specifically relating to a method for identifying human factors and predicting risk trends in hydrogen energy accidents that integrates HFACS-HEA and PSOGM. Background Technology

[0002] Hydrogen energy, as an important component of clean energy, is experiencing rapid industrial development. However, the flammable and explosive nature of hydrogen, along with the complexity of operating and maintaining hydrogen energy systems, leads to frequent hydrogen energy accidents. Human factors are one of the core causes of these accidents. Therefore, systematically identifying human factors and predicting risk trends in hydrogen energy accidents is of great significance for improving hydrogen energy safety management and preventing accidents.

[0003] The classic Human Factors Analysis and Classification System (HFACS) framework has been widely used in human factor analysis of accidents in fields such as aviation and chemical engineering. For example, Chinese patent CN115809801A discloses a human factor analysis method for power safety accidents based on HFACS, which uses the HFACS model to classify human factors in power accidents and analyzes the causal relationships between levels through chi-square tests and odds ratio analysis. However, when this method is directly applied to hydrogen energy systems, there are obvious shortcomings: First, the classic HFACS framework and its application in fields such as power are not designed for hydrogen energy systems and cannot match the unique risk attributes of hydrogen energy systems, such as high diffusivity, hydrogen embrittlement, and low-temperature, high-pressure storage. This results in mismatched factor classification, inapplicable hierarchical relationships, and difficulty in guaranteeing the accuracy of the analysis results. Second, existing research on human factors in hydrogen energy accidents is mostly qualitative description or simple frequency statistics, lacking quantitative verification of causal relationships between factors and failing to identify the key transmission paths in the accident causal chain. Furthermore, in terms of risk prediction, traditional grey models and regression models suffer from problems such as reliance on experience for parameter optimization and low prediction accuracy, making it difficult to achieve accurate early warning of hydrogen energy accident trends. This results in a lack of data support for risk management measures, which are therefore lagging behind.

[0004] Furthermore, existing research on hydrogen energy safety management has not yet established a complete methodology system encompassing systematic identification of human factors, quantitative analysis of causal relationships, screening of key factors, and accurate prediction of risk trends. This results in a lack of targeted control over human factors in hydrogen energy accidents and insufficient effectiveness in risk warning.

[0005] To address the aforementioned issues, there is an urgent need in this field for a systematic analysis framework of human factors that can be adapted to the characteristics of hydrogen energy systems, and a technical solution that integrates factor identification, causal quantification, and risk prediction. Summary of the Invention

[0006] This invention aims to address the technical problems of existing human factor analysis in hydrogen energy accidents, such as the lack of a dedicated and systematic classification framework, the lack of quantitative verification of causal relationships between factors, and the low accuracy of risk trend prediction. It provides a method for identifying human factors and predicting risk trends in hydrogen energy accidents that integrates HFACS-HEA and PSOGM, comprising the following steps: S1. Construct a dedicated HFACS-HEA model for hydrogen energy accidents and establish a dataset for identifying human factors in hydrogen energy accidents: Based on the classic HFACS framework, and combined with the unique risks of hydrogen energy's high diffusivity, hydrogen embrittlement effect, and low-temperature, high-pressure storage and transportation, a scenario-based modification is made to construct an HFACS-HEA model with four levels: organizational influence, unsafe supervision, preconditions for unsafe behavior, and unsafe behavior. These four levels contain 12 core factors. The organizational influence level includes poor management of hydrogen energy-specific resources, lack of a hydrogen energy safety culture, and loopholes in the organizational process for hydrogen energy projects. The unsafe supervision level includes lack of supervision over hydrogen energy operations, insufficient auditing of hydrogen energy equipment, and absence of emergency supervision for hydrogen energy. The preconditions for unsafe behavior level includes insufficient personnel awareness of hydrogen energy, defects in hydrogen energy equipment, hydrogen-related environmental risks, and defects in the hydrogen energy workspace. The unsafe behavior level includes operational errors and violations of hydrogen energy regulations. Historical hydrogen energy accident reports are integrated, and human-caused factors are labeled based on the HFACS-HEA model to establish a dataset. S2. Quantitatively identify the hierarchical causal transmission relationship of human factors: Based on the dataset of S1, the chi-square test is used to verify the correlation between adjacent hierarchical factors in the HFACS-HEA model, and the degree of causal association is quantified by the concession ratio analysis to identify the key transmission path of the accident causal chain. S3. Screening key human factors for priority control of hydrogen energy accidents: Using the total number of hydrogen energy accidents as a reference sequence and the occurrence frequency of each human factor in the HFACS-HEA model as a comparison sequence, the grey relational analysis method is used to calculate the correlation between each factor and the total number of accidents, and the key human factors for priority control are screened according to the magnitude of the correlation. S4. Constructing a PSOGM model to predict hydrogen energy accident trends: Preprocess the raw data of hydrogen energy accidents, construct a basic GM(1,1) model, and use the particle swarm optimization algorithm to globally optimize the dynamic weights of the background value, the development coefficient a, and the gray action b of the model to construct a PSOGM model, and use this model to predict the trend of hydrogen energy accidents.

[0007] Furthermore, in S1, the construction requirements for the hydrogen energy accident human factor identification dataset are as follows: collect a preset number of hydrogen energy accident reports, covering the entire process of hydrogen energy production, storage, transportation, and refueling, obtain valid samples after data cleaning, and label the human factors causing each accident one by one based on the HFACS-HEA model. If a certain core factor is the direct / indirect cause of the accident, it is assigned a value of 1, otherwise it is assigned a value of 0. At the same time, auxiliary information such as the time of occurrence, the stage, and the severity of the accident are also labeled.

[0008] Furthermore, in S2, the formula for calculating the chi-square test statistic is as follows: Where A is the organizational influence layer, B is the unsafe monitoring layer, C is the precondition layer for unsafe behavior, and D is the unsafe behavior layer; the formula for calculating the concession ratio is: Analysis revealed that insufficient equipment audits significantly increased equipment-related risks, while personnel factors significantly induced violations.

[0009] Furthermore, in S3, the grey relational analysis uses Z-score standardization to process the data, with a resolution coefficient of 0.5, and the grey relational degree is the mean of the relational coefficients; through analysis, organizational process loopholes, insufficient equipment audits, and operational factors are identified as priority control key factors with a very significant correlation to hydrogen energy accidents.

[0010] Furthermore, in S4, the construction of the PSOGM model includes three steps: preprocessing of the original data by the average weakening buffer operator, construction of the GM(1,1) model, and optimization of PSO algorithm parameters. The PSO algorithm uses the minimum sum of squared errors as the fitness function.

[0011] Furthermore, the method is used for the safety management of hydrogen production, hydrogen storage, hydrogen transportation, or hydrogen refueling systems.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention constructs an HFACS-HEA model adapted to the characteristics of hydrogen energy systems, breaks through the limitations of the classic HFACS framework in hydrogen energy accident analysis, establishes a systematic human factor classification system containing 12 core factors at 4 levels, solves the problem of unsystematic and unfit identification of human factors in hydrogen energy accidents, and provides a new theoretical perspective for in-depth analysis of human causes.

[0013] (2) This invention comprehensively utilizes the chi-square test and concession ratio analysis, and for the first time realizes the quantitative verification of the causal transmission relationship between human factors in hydrogen energy accidents. It clarifies the key transmission paths in the accident causal chain, such as insufficient equipment review → equipment factor risk, and personnel factors → violations. This upgrades the analysis of human factors from qualitative description to quantitative analysis, and improves the scientificity and reliability of the analysis results.

[0014] (3) This invention uses grey relational analysis to quantify the importance of each human factor, accurately identify the key risk factors that should be prioritized for control, provide clear control priorities for hydrogen energy safety management, and avoid the blindness of control work.

[0015] (4) This invention introduces the PSOGM model to achieve accurate prediction of hydrogen energy accident trends. By using the particle swarm optimization algorithm to globally optimize the background value dynamic weight, development coefficient and gray action of the traditional gray model, it solves the problem of insufficient parameter optimization of the traditional gray model and significantly improves the prediction accuracy.

[0016] (5) The present invention constructs a complete method from systematic identification of human factors, quantitative analysis of causal relationships, screening of key factors to accurate prediction of risk trends, providing theoretical basis and practical support for the identification and prevention of hidden human factors in hydrogen energy accidents, and has important application value for reducing the probability of hydrogen energy accidents and improving the safety management level of hydrogen energy systems.

[0017] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a diagram illustrating the overall research framework of the present invention: a method for identifying human factors and predicting risk trends in hydrogen energy accidents that integrates HFACS-HEA and PSOGM. Figure 2 This is a schematic diagram of the four-layer architecture of the classic HFACS 8.0 model; Figure 3 This is a schematic diagram of the 4-level, 12-core components of the HFACS-HEA model constructed in this invention. Figure 4 A statistical distribution map of hydrogen energy accidents from 2000 to 2022; Figure 5 The grey relational ranking diagram of each level of the HFACS-HEA model and hydrogen energy accidents; Figure 6 The grey relational ranking diagram of the 12 core factors of the HFACS-HEA model and hydrogen energy accidents; Figure 7 A comparison of the simulated hydrogen accident number curves for the GM(1,1), GAGM, and PSOGM models; Figure 8 A comparison chart of the prediction performance metrics of the GM(1,1), GAGM, and PSOGM models. Detailed Implementation

[0019] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0020] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0021] Example 1 Please see Figures 1 to 8 This embodiment provides a method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM. Figure 1 This is a diagram illustrating the overall research framework of the method of this invention. This method solves the problems of unsystematic identification of human factors in existing hydrogen energy accidents, unclear causal transmission relationships, and low accuracy of risk prediction. It constructs a full-process analysis framework from factor identification and correlation analysis to risk prediction.

[0022] S1: Construct a HFACS-HEA model specifically for hydrogen energy accidents and establish a dataset for identifying human factors in hydrogen energy accidents.

[0023] Specifically, such as Figure 2 As shown, this implementation is based on the classic HFACS 8.0 framework. Considering the unique risks of hydrogen energy systems, such as high diffusivity, hydrogen embrittlement, cryogenic and high-pressure storage and transportation, and explosiveness, this embodiment modifies and refines the original four-layer architecture for hydrogen energy scenarios. The final implementation... Figure 3 The HFACS-HEA model shown contains four levels: organizational influence, unsafe monitoring, preconditions for unsafe behavior, and unsafe behavior, totaling 12 core factors.

[0024] In a preferred implementation, the HFACS-HEA model comprises 12 core factors across 4 levels: (1) Organizational impact layer (A): including poor management of hydrogen energy special resources (A1), lack of hydrogen energy safety culture (A2), and loopholes in the organizational process of hydrogen energy special projects (A3); (2) Unsafe supervision layer (B): including lack of supervision of hydrogen energy operation (B1), insufficient review of hydrogen energy equipment (B2), and lack of supervision of hydrogen energy emergency (B3); (3) Preconditions for unsafe acts (C): including insufficient knowledge of hydrogen energy among personnel (C1), defects in hydrogen energy equipment (C2), environmental risks associated with hydrogen energy (C3), and defects in the working space for hydrogen energy (C4); (4) Unsafe behavior layer (D): includes hydrogen energy operation errors (D1) and hydrogen energy violation operation (D2).

[0025] This embodiment integrates authoritative data sources such as the French ARIA event database, the EU eMARS database, and the HIAD2.1 dataset, filtering hydrogen energy accident reports from 2000 to 2022. To improve the statistical power of the model, such as... Figure 4 As shown, this embodiment collected and screened a total of 429 valid hydrogen energy accident reports, covering the entire process of hydrogen energy production (18%), storage (35%), transportation (22%), and refueling (25%).

[0026] Based on the HFACS-HEA model, the human-caused factors of each accident are structurally labeled and encoded using a binary representation method (0 / 1): if a core factor is a direct or indirect cause of the accident, it is assigned a value of 1, otherwise it is assigned a value of 0. Simultaneously, auxiliary information such as the time of occurrence, stage, and severity of the accident is labeled, thereby establishing a dataset for identifying human factors in hydrogen energy accidents.

[0027] S2: Based on the chi-square test and concession ratio analysis, quantitatively identify the hierarchical causal transmission relationship of human factors.

[0028] Based on the dataset constructed in S1, a 2×2 contingency table was built. The chi-square test was used to verify the correlation between adjacent level factors in the HFACS-HEA model, with a significance criterion of P < 0.05. The formula for calculating the chi-square test statistic is as follows:

[0029] For factor pairs with significant correlation, the degree of causal association is further quantified using the concession ratio (OR) analysis. The formula for calculating the concession ratio is: The criterion is: when OR>1 and P<0.001, the factor is considered a key transmission factor in the accident causation chain.

[0030] This embodiment found through the above analysis that: insufficient equipment audit (B2) significantly increases the risk of equipment factors (C2) (OR=8.125, P<0.001), and personnel factors (C1) significantly induce violations (D2) (OR=5.992, P<0.001).

[0031] In another variation, the significance level α is not limited to 0.05. Those skilled in the art can set the P-value threshold to 0.01 or 0.1 according to the sample size of the dataset and the actual situation to adapt to different accuracy requirements.

[0032] S3: Grey relational analysis was used to screen key human factors for priority management of hydrogen energy accidents.

[0033] Using the total number of hydrogen energy accidents each year from 2000 to 2022 as a reference series, and the annual frequency of occurrence of 12 individuals as factors in the HFACS-HEA model as a comparison series, a grey relational analysis was performed. The specific calculation steps are as follows: (1) Data standardization processing.

[0034] Z-score standardization is used to eliminate the influence of dimensions. The formula is:

[0035] in, This is the original data; This represents the average of the original data. This represents the standard deviation of the original data.

[0036] (2) Calculation of correlation coefficient.

[0037]

[0038] in, The resolution coefficient is set to 0.5. This is the standardized reference sequence; This is the standardized comparison sequence.

[0039] (3) Calculation of grey relational degree.

[0040] The formula for calculating the mean of the correlation coefficient is: .

[0041] Analysis results as follows Figure 5 and Figure 6 As shown in the figure, this embodiment identifies the following as key control factors with a very significant correlation to hydrogen energy accidents: organizational process vulnerabilities (A3, correlation 0.704979), insufficient equipment audits (B2, correlation 0.678142), and operational factors (C4, correlation 0.667055).

[0042] In yet another variant embodiment, the resolution coefficient Not limited to 0.5, other values ​​can be selected in the (0,1] range, such as 0.4 or 0.6, to adjust the discriminative power of the correlation coefficient, depending on the degree of fluctuation of the data sequence.

[0043] S4: Build a PSOGM model to achieve accurate prediction of hydrogen energy accident trends.

[0044] This step first preprocesses the raw data of hydrogen energy accidents using the average weakening buffer operator, then constructs a basic GM(1,1) model, and finally uses the particle swarm optimization (PSO) algorithm to globally optimize the background value dynamic weights, development coefficients a, and gray action b of the model to build the PSOGM model.

[0045] (1) Preprocessing of raw data.

[0046] The raw time-series data of hydrogen energy accidents are processed using the average weakening buffer operator:

[0047] in, The original data, For the processed data, n is the length of the original data sequence, and k is the index of the data point in the sequence.

[0048] (2) Construction of GM(1,1) model.

[0049] After performing a single accumulation of the preprocessed data (1-AGO), a first-order single-variable ordinary differential equation is established:

[0050] Among them, X (1) For a cumulative sequence, a is the development coefficient and b is the gray action quantity.

[0051] (3) Parameter optimization of PSO algorithm.

[0052] The fitness function is the one that minimizes the sum of squared errors.

[0053] in, The original data, For model prediction data.

[0054] The PSO algorithm iteratively updates particle velocity and position, optimizing background weight, development coefficient 'a', and gray action 'b'. In this embodiment, the PSO algorithm parameters are set as follows: particle swarm size of 50, maximum number of iterations of 100, and inertia weight factor linearly decreasing from 0.9 to 0.4.

[0055] (4) Model validation.

[0056] The accuracy of the model is verified using the posterior difference test. ;in, The standard deviation of the original sequence. The standard deviation of the residual sequence. Small probability error. .

[0057] In this embodiment, the predictive performance of the PSOGM model is as follows: mean relative error 14.00%, model accuracy 86.00%, posterior error ratio C < 0.35, and small probability error P = 1, achieving first-level prediction accuracy. This is significantly better than the traditional GM(1,1) model (mean relative error 15.16%, model accuracy 84.84%) and the GAGM model (mean relative error 14.16%, model accuracy 85.84%). Figure 7 and Figure 8 As shown.

[0058] In another variant, the parameters of the PSO algorithm are not fixed values. For systems with larger datasets or higher real-time prediction requirements, the population size and number of iterations can be appropriately increased to seek a better global solution; or on resource-constrained edge computing devices, these parameters can be appropriately reduced to balance computational efficiency and accuracy.

[0059] In yet another variant embodiment, the PSOGM prediction model of the present invention is not limited to comparison with GAGM and GM(1,1). In more complex engineering applications, it can also be compared and validated with deep learning methods such as LSTM and GRU to adapt to prediction needs of different time scales and data characteristics.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM, characterized in that, Includes the following steps: S1. Construct a dedicated HFACS-HEA model for hydrogen energy accidents and establish a dataset for identifying human factors in hydrogen energy accidents: Based on the classic HFACS framework, and combined with the unique risks of hydrogen energy's high diffusivity, hydrogen embrittlement effect, and low-temperature, high-pressure storage and transportation, a scenario-based modification is made to construct an HFACS-HEA model with four levels: organizational influence, unsafe supervision, preconditions for unsafe behavior, and unsafe behavior. These four levels contain 12 core factors. The organizational influence level includes poor management of hydrogen energy-specific resources, lack of a hydrogen energy safety culture, and loopholes in the organizational process for hydrogen energy projects. The unsafe supervision level includes lack of supervision over hydrogen energy operations, insufficient auditing of hydrogen energy equipment, and absence of emergency supervision for hydrogen energy. The preconditions for unsafe behavior level includes insufficient personnel awareness of hydrogen energy, defects in hydrogen energy equipment, hydrogen-related environmental risks, and defects in the hydrogen energy workspace. The unsafe behavior level includes operational errors and violations of hydrogen energy regulations. Historical hydrogen energy accident reports are integrated, and human-caused factors are labeled based on the HFACS-HEA model to establish a dataset. S2. Quantitatively identify the hierarchical causal transmission relationship of human factors: Based on the dataset of S1, the chi-square test is used to verify the correlation between adjacent hierarchical factors in the HFACS-HEA model, and the degree of causal association is quantified by the concession ratio analysis to identify the key transmission path of the accident causal chain. S3. Screening key human factors for priority control of hydrogen energy accidents: Using the total number of hydrogen energy accidents as a reference sequence and the occurrence frequency of each human factor in the HFACS-HEA model as a comparison sequence, the grey relational analysis method is used to calculate the correlation between each factor and the total number of accidents, and the key human factors for priority control are screened according to the magnitude of the correlation. S4. Constructing a PSOGM model to predict hydrogen energy accident trends: Preprocess the raw data of hydrogen energy accidents, construct a basic GM(1,1) model, and use the particle swarm optimization algorithm to globally optimize the dynamic weights of the background value, the development coefficient a, and the gray action b of the model to construct a PSOGM model, and use this model to predict the trend of hydrogen energy accidents.

2. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 1, is characterized in that... In S1, the construction requirements for the hydrogen energy accident human factor identification dataset are as follows: collect a preset number of hydrogen energy accident reports, covering the entire process of hydrogen energy production, storage, transportation, and refueling, obtain valid samples after data cleaning, and label the human factors causing each accident one by one based on the HFACS-HEA model. If a certain core factor is the direct / indirect cause of the accident, it is assigned a value of 1, otherwise it is assigned a value of 0. At the same time, auxiliary information such as the time of occurrence, the stage, and the severity of the accident are also labeled.

3. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 1, characterized in that, In S2, the formula for calculating the chi-square test statistic is as follows: Among them, A is the organizational influence layer, B is the unsafe supervision layer, C is the precondition layer for unsafe behavior, and D is the unsafe behavior layer; The formula for calculating the concession ratio is as follows: ; Analysis revealed that insufficient equipment audits significantly increased equipment-related risks, while personnel factors significantly induced violations.

4. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 1, characterized in that, In S3, the grey relational analysis uses Z-score standardization to process the data, with a resolution coefficient of 0.5, and the grey relational degree is the mean of the relational coefficients. Through analysis, organizational process loopholes, insufficient equipment review, and operational factors are identified as priority control key factors with a very significant correlation to hydrogen energy accidents.

5. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 1, characterized in that, In S4, the construction of the PSOGM model includes three steps: preprocessing of the original data by the average weakening buffer operator, construction of the GM(1,1) model, and optimization of PSO algorithm parameters. The PSO algorithm uses the minimum sum of squared errors as the fitness function.

6. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 5, is characterized in that... In S4, the preprocessing formula for the average weakening buffer operator is as follows: ;in, The original data, For the processed data, n is the length of the original data sequence, and k is the index of the data point in the sequence.

7. The method for identifying human factors and predicting risk trends in hydrogen energy accidents by integrating HFACS-HEA and PSOGM as described in claim 1, characterized in that: The method is used for the safety management of hydrogen production, hydrogen storage, hydrogen transportation, or hydrogen refueling systems.

Images