Oil and gas pipeline quantitative risk evaluation method, system, equipment and medium

Abstract

The invention belongs to the technical field of oil and gas chemical engineering risk assessment, and discloses an oil and gas pipeline quantitative risk evaluation method, system, equipment and medium, and the method comprises the steps: analyzing an accident case of an oil and gas pipeline, and constructing a system comprising multiple levels of risk evaluation indexes based on an analysis result; calculating a risk value and a weight of each risk evaluation index in each multi-level risk evaluation index system; inputting the risk value of each risk evaluation index into a risk level evaluation model for classification prediction to obtain a corresponding risk level; calculating a corresponding SHAP value based on the weight and the risk level of each risk evaluation index; sorting each risk evaluation index according to the size of the SHAP value to obtain a risk priority; and carrying out safety supervision work on the oil and gas pipeline according to a risk grade evaluation result. According to the technical scheme, the safety level of the risk index can be scientifically, accurately and efficiently assessed, and the method is more suitable for the field process.

Description

Technical Field

[0001] The present invention belongs to the technical field of oil and gas chemical industry risk assessment, and in particular relates to a method, system, equipment and medium for quantitative risk assessment of oil and gas pipelines. Background Art

[0002] Pipeline transportation is the primary mode of transport for oil and natural gas. Regular risk assessments and subsequent risk management of oil and gas pipeline systems are crucial for their safe operation. Previous assessment methods have relied on expert judgment, employing semi-quantitative risk assessment methods to quantify the causes and consequences of oil pipeline failures and conduct current risk assessments. These methods fail to account for the subjectivity of expert judgment, making risk intervention measures ineffective in impacting risk assessment results. Furthermore, they fail to consider the need for strong generalization capabilities, resulting in risk assessments that are not tailored to field processes and incapable of accurately assessing the risk level of risk indicators. Therefore, a scientific and efficient risk assessment model is needed for accurate assessments. Summary of the Invention

[0003] The purpose of the present invention is to provide a method, system, equipment and medium for quantitative risk assessment of oil and gas pipelines to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, the present invention provides a quantitative risk assessment method for oil and gas pipelines, comprising:

[0005] Analyze oil and gas pipeline accident cases and build a multi-level risk assessment indicator system based on the analysis results;

[0006] Calculating the risk value and weight of each risk assessment indicator in each of the multi-level risk assessment indicator systems;

[0007] The risk value of each risk assessment indicator is input into the risk level assessment model for classification prediction to obtain the corresponding risk level; wherein the risk level assessment model is constructed based on a double hidden layer adaptive regularized neural network;

[0008] Calculate the corresponding SHAP value based on the weight and risk level of each risk assessment indicator;

[0009] Sort each risk assessment indicator according to the size of the SHAP value to obtain the risk priority;

[0010] Carry out safety monitoring of oil and gas pipelines based on risk level assessment results.

[0011] Optionally, the multi-level risk assessment indicator system includes first-level risk assessment indicators, second-level risk assessment indicators and third-level risk assessment indicators.

[0012] Optionally, the risk value calculation process specifically includes:

[0013] Constructing a frequency-consequence matrix corresponding to the multi-level risk assessment indicator system according to the LEC evaluation method;

[0014] The risk value of each risk assessment indicator is calculated based on the constructed matrix.

[0015] Optionally, the weight calculation process is:

[0016] Based on the improved hierarchical analysis method, the indicators of the secondary risk assessment indicators and the initial single weights of the tertiary risk assessment indicators are calculated. The weights of the secondary risk assessment indicators are used as the intermediate event weights in the comprehensive weight method, and the initial single weights are used as the single event weights in the comprehensive weight method. The intermediate event weights and the single event weights are multiplied together to obtain the comprehensive weights of the tertiary risk assessment indicators.

[0017] Optionally, the risk level assessment model training process:

[0018] Acquiring training data, wherein the training data includes risk training values ​​and corresponding risk levels;

[0019] An initial risk level evaluation model is constructed, the training data is input into the initial risk level evaluation model for classification prediction, and training is performed with the goal of minimizing the loss between the initial training result after classification prediction and the risk level corresponding to the risk training value to obtain a trained risk level evaluation model.

[0020] Optionally, the processing of the risk level evaluation model specifically includes:

[0021] The risk value of each risk assessment indicator is input into a double hidden layer adaptive regularized neural network, the risk value is discretized into multiple risk levels and used as output to obtain the corresponding risk level.

[0022] Optionally, the corresponding SHAP value is calculated based on the weight of each risk assessment indicator and the risk level. The specific calculation process is:

[0023] baseline=average(RV)

[0024] SHAP_values=(RV-baseline)·W3

[0025] Where baseline is the average of all risk values, which is used to measure the difference between individual risk values ​​and the overall average level; average is the average of the risk values ​​of all risk assessment indicators; RV is the risk value of all risk assessment indicators; SHAP_values ​​is the SHAP value of each risk assessment indicator; and W3 is the comprehensive weight of the three-level risk assessment indicators.

[0026] A quantitative risk assessment system for oil and gas pipelines, comprising:

[0027] The indicator system construction module is used to analyze accident cases of oil and gas pipelines and build a multi-level risk assessment indicator system based on the analysis results;

[0028] A risk level prediction module is used to calculate the risk value and weight of each risk assessment indicator in the multi-level risk assessment indicator system; input the risk value of each risk assessment indicator into the risk level assessment model for classification prediction to obtain the corresponding risk level; wherein the risk level assessment model is constructed based on a dual hidden layer adaptive regularized neural network;

[0029] The risk priority calculation module is used to calculate the corresponding SHAP value based on the weight and risk level of each risk assessment indicator; sort the risk assessment indicators according to the size of the SHAP value to obtain the risk priority; and conduct safety monitoring of oil and gas pipelines based on the risk level assessment results.

[0030] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for quantitative risk assessment of oil and gas pipelines.

[0031] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for quantitative risk assessment of oil and gas pipelines.

[0032] The technical effects of the present invention are:

[0033] 1. Based on the case analysis of oil and gas pipeline accidents, the present invention establishes an oil and gas pipeline risk indicator assessment system to make risk assessment more in line with the on-site process.

[0034] 2. The present invention applies the risk assessment index occurrence frequency-consequence severity matrix to the analytic hierarchy process to improve the analytic hierarchy process so that the analytic hierarchy process no longer needs to perform consistency checks when calculating weights, thus saving calculation costs.

[0035] 3. According to the risk level classification standard, the present invention integrates a double hidden layer adaptive regularized neural network. The network adopts a double hidden layer decreasing structure to gradually refine the characteristics of the risk level to avoid information redundancy; the hyperbolic tangent function (tansig) that is more suitable for small sample training is used as the activation function to alleviate the risk of gradient disappearance; a dynamically adjusted early stopping mechanism is set to adaptively control the number of training rounds to balance the model complexity and generalization ability; an L2 regularization weight decay term is introduced in the cross entropy loss function, and used in conjunction with the early stopping mechanism to achieve the effect of double overfitting; in order to adapt to the regularization and early stopping mechanisms, the conjugate gradient optimization algorithm (SCG) is optimized so that the optimization algorithm only requires the information of the first-order derivative and can automatically update the learning rate, converge quickly within the number of iterations, and meet the requirements of high-frequency data updates. Through the coordinated design of structural optimization and training strategies, the risk level of each risk assessment indicator is accurately given. The neural network structure supports the expansion of input dimensions and enhances the generalization ability of the model.

[0036] 4. According to the risk value and weight of each risk assessment indicator, the present invention introduces the SHAP (Shapley Additive Explanations) framework to analyze the contribution of each risk assessment indicator to the final assessment result, obtain the risk priority of each risk assessment indicator, and improve the transparency of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0039] Figure 1 : is a structural diagram of a dual hidden layer adaptive regularized neural network in an embodiment of the present invention;

[0040] Figure 2 SHAP contribution graph of the risk assessment indicator in an embodiment of the present invention;

[0041] Figure 3 Schematic diagram of the indicator system in an embodiment of the present invention;

[0042] Figure 4 This is an implementation flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0043] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as limiting the present invention, but rather as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0044] It should be understood that the terms described herein are intended only to describe particular embodiments and are not intended to limit the present invention. In addition, for numerical ranges herein, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Each smaller range between any intermediate value within a stated value or stated range and any other stated value or intermediate value within the stated range is also encompassed by the present invention. The upper and lower limits of these smaller ranges may be independently included or excluded within the scope.

[0045] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present invention without departing from the scope or spirit of the invention. Other embodiments will be apparent to those skilled in the art from the present invention. The present description and examples are intended to be illustrative only.

[0046] The words “include,” “including,” “have,” “contain,” etc. used in this article are open-ended terms, meaning including but not limited to.

[0047] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] like Figure 1 - Figure 4 As shown, this embodiment provides a quantitative risk assessment method for oil and gas pipelines, including: analyzing accident cases of oil and gas pipelines, and constructing a multi-level risk assessment index system based on the analysis results; calculating the risk value and weight of each risk assessment indicator in the multi-level risk assessment index system; inputting the risk value of each risk assessment indicator into a risk level assessment model for classification prediction to obtain a corresponding risk level; wherein the risk level assessment model is constructed based on a double hidden layer adaptive regularized neural network; calculating the corresponding SHAP value based on the weight and risk level of each risk assessment indicator; sorting each risk assessment indicator according to the size of the SHAP value to obtain the risk priority; and conducting safety monitoring of the oil and gas pipeline according to the risk level assessment results.

[0049] This embodiment discloses a method for constructing a quantitative risk assessment model for oil and gas pipelines based on an improved analytic hierarchy process (AHP) and a dual-hidden-layer adaptive regularized neural network. This method, which belongs to the field of oil and gas chemical industry risk assessment, includes the following steps: Step 1: Establishing an oil and gas pipeline risk assessment index system based on five key risk factors: corrosion, structural defects, third-party sabotage, natural disasters, and physical damage; Step 2: Applying an improved AHP based on the five-scaling method to calculate the weights of the oil and gas pipeline risk assessment indicators; Step 3: Integrating a novel dual-hidden-layer adaptive regularized neural network to derive the risk level of each risk assessment indicator based on the risk level classification criteria; Step 4: Introducing the SHAP (Shapley Additive Explanations) analysis module to quantify the importance of each risk assessment indicator and derive the risk priority of each risk assessment indicator; Step 5: Verifying the high accuracy of the method for constructing a quantitative risk assessment model for oil and gas pipelines based on the improved AHP and dual-hidden-layer adaptive regularized neural network. Experimental results demonstrate that this embodiment can scientifically, accurately, and efficiently assess the safety level of risk indicators, more closely matching on-site processes.

[0050] This embodiment proposes a method for constructing a quantitative risk assessment model for oil and gas pipelines based on an improved hierarchical analysis method and a dual-hidden layer adaptive regularized neural network. This method addresses existing issues such as risk assessment not being tailored to on-site processes, risk intervention measures not being able to take effect based on risk assessment results, and low risk indicator classification accuracy, which results in inefficient and inaccurate risk assessment models. The specific technical solutions of this embodiment include:

[0051] Step 1: Establish a risk indicator system. Based on the data collected from the PHMSA website on the frequency of occurrence and number of casualties of risk factors that cause oil and gas pipelines to be unsafe to operate, as shown in Table 1, the risk indicator system is divided into three levels;

[0052] Step 2: Based on the LEC evaluation method, a frequency-consequence matrix compatible with the 5-scale method is designed, as shown in Table 2. The principle for calculating the risk value of risk factors in the frequency-consequence matrix designed in this paper is equivalent to the risk factor calculation principle of the risk matrix and complies with the ISO31000 standard. The frequency-consequence matrix is ​​then integrated into the analytic hierarchy process for improvement, and the weights of the second-level risk assessment indicators and the single weights of the third-level risk assessment indicators are calculated. Then, based on the concept of the comprehensive weighting method, the comprehensive weights of the third-level risk assessment indicators are calculated.

[0053] Step 3: Integrate a dual hidden layer adaptive regularized neural network to calculate the risk level of each risk assessment indicator;

[0054] Step 4: Introduce the SHAP (Shapley Additive Explanations) analysis module to calculate the risk priority of each risk assessment indicator;

[0055] Step 5: Conduct case analysis to verify the high accuracy of the method for constructing a quantitative risk assessment model for oil and gas pipelines based on the improved hierarchical analysis method and double hidden layer adaptive regularized neural network. Based on the evaluation data of the oil and gas pipelines in a certain oil field in 2023, the risk indicator system established based on the analysis of historical accident cases of oil and gas pipelines is used for analysis. Based on the evaluation results, effective suggestions are put forward for the safety supervision of oil and gas pipelines.

[0056] Optionally, step 2 includes the following steps:

[0057] Step 2.1: Based on the established three-level risk indicator evaluation system, design the frequency-consequence matrix shown in Table 3. Then, based on the risk evaluation indicator risk values ​​calculated from the frequency-consequence matrix, construct the risk evaluation indicator judgment matrix, which is represented as follows:

[0058]

[0059] Step 2.2: Calculate the weight of the secondary risk assessment indicator and the initial single weight of the tertiary risk assessment indicator, which are represented as follows: R·W s =λ max W s

[0060] Where: max is the maximum eigenvalue of R, W s is λ max The eigenvector of

[0061] Step 2.3: Based on the idea of ​​the comprehensive weighting method, the weight of the second-level risk assessment indicator calculated using the improved analytic hierarchy process is used as the intermediate event weight in the comprehensive weighting method. The single weight of the third-level risk assessment indicator calculated using the improved analytic hierarchy process is used as the single event weight in the comprehensive weighting method. Finally, the two weights are multiplied together to obtain the comprehensive weight of the third-level risk assessment indicator. The representation is as follows:

[0062]

[0063] It is feasible that the step 3 specifically includes the following steps:

[0064] Step 3.1: Calculate the risk value of each risk assessment indicator based on the collected data on the frequency of occurrence of risk assessment indicators and the number of casualties caused. This is used as the core feature input of the dual hidden layer adaptive regularized neural network. The risk value is discretized into multiple risk levels and used as the output. The network structure of this neural network is: 1 input - 8 hidden layers - 6 hidden layers - multiple outputs. The specific structure is as follows: Figure 1 As shown;

[0065] Step 3.2: Use the hyperbolic tangent function (tansig) as the hidden layer activation function to map the risk value and provide a nonlinear transformation; use the multi-classification probability function (softmax) as the output layer activation function to output the probability of each risk indicator belonging to each risk level. The representation is as follows:

[0066]

[0067] Where: z input is the input of each output node, K is the number of risk levels;

[0068] Step 3.3: Use the regularized cross entropy function as the loss function and add a weight decay term to the loss function to perform double overfitting. The representation is as follows:

[0069]

[0070] Where: t ik is the actual risk level of the i-th risk assessment indicator, y ik is the probability of predicting the i-th risk assessment indicator to be the K-th risk level, λ is the regularization coefficient;

[0071] Step 3.4: At the Kth training round, calculate the cross entropy loss of the validation set; if the current validation set cross entropy loss is lower than the historical best value, update the best cross entropy loss and reset the failure counter; otherwise, increase the failure counter; when the number of consecutive failures reaches the threshold, trigger the early stopping mechanism and terminate the training;

[0072] Step 3.5: The optimized SCG algorithm accelerates convergence through first-order derivative information, avoiding the problem of falling into local optimality caused by the fixed learning rate of traditional gradient descent. The core formula of the optimization algorithm is as follows:

[0073] The curvature information of the K-th iteration is characterized as follows:

[0074]

[0075]

[0076] Where: w k is the weight vector of the Kth iteration, σ k is the scaling factor for the Kth iteration, d k is the conjugate gradient direction of the Kth iteration, δ k is the adjusted curvature estimate.

[0077] When δ k When ≤0, the curvature estimation needs to be corrected as follows:

[0078]

[0079] δ k =-δ k +λ k ||d k || 2

[0080] The learning rate is represented as follows:

[0081]

[0082] The weight update is represented as follows:

[0083] w k+1 =w k +α k d k

[0084] The gradient and direction update are characterized as follows:

[0085]

[0086]

[0087] d k+1 =-g k+1 +β k d k

[0088] The scaling factor is adaptively adjusted according to the curvature estimation error, which can be characterized as follows:

[0089]

[0090] It is feasible that step 4 specifically includes the following steps:

[0091] Step 4.1: The Shapley Additive Explanations framework analyzes the SHAP contribution values ​​of multiple risk assessment indicators to reveal the core factors that influence the risk assessment results, increase the interpretability of the model, and enhance the credibility of the model. The calculation of the SHAP contribution value is as follows:

[0092] baseline=average(RV)

[0093] SHAP_values=(RV-baseline)·W3

[0094] Where baseline is the average of all risk values, which is used to measure the difference between individual risk values ​​and the overall average level; average is the average of the risk values ​​of all risk assessment indicators; RV is the risk value of all risk assessment indicators; SHAP_values ​​is the SHAP value of each risk assessment indicator; and W3 is the comprehensive weight of the three-level risk assessment indicators.

[0095] Arrange the SHAP contribution values ​​of each risk assessment indicator in descending order to obtain the priority of each risk assessment indicator;

[0096] It is feasible that the step 5 specifically includes the following steps:

[0097] Step 5.1. Based on the established three-level risk assessment indicator frequency-consequence matrix, the risk level is divided into four levels: Level 1 (U1) represents low risk; Level 2 (U2) represents medium risk; Level 3 (U3) represents high risk; Level 4 (U4) represents extreme risk. The corresponding risk value intervals are: D1[0,4), D2[4,9), D3[9,16), D4[16,25]. The risk values ​​of each risk assessment indicator are brought into the double hidden layer adaptive regularized neural network (DHAR-Net) designed in this paper, and the results are shown in Table 3. As can be seen from Table 3, the accuracy of the double hidden layer adaptive regularized neural network (DHAR-Net) for the samples in this paper can reach 100%;

[0098] Step 5.2: Introduce the SHAP (Shapley Additive Explanations) framework to analyze the contribution of each risk indicator to the final assessment result. The SHAP contribution of each risk assessment indicator is as follows: Figure 2 As shown. Figure 2 It can be seen that the contribution of each risk assessment indicator to the final assessment result is calculated as follows: 0.0326, 0.0592, 0.1075, 0.0197, 0.0111, 0.0372, 0.0197, 0.0372, 0.0157, 0.0038, 0.0260, 0.0093, 0.0057, 0.0157, 0.0074, 0.0147, 0.0074, 0.0065, 0.0122, 0.0065, 0.0074, and the risk priority of each risk assessment indicator is shown in Table 3. Among them, the contribution of the risk assessment indicator internal corrosion is significantly higher than that of other indicators, indicating that the risk assessment indicator internal corrosion plays a decisive role in the risk assessment, and the future risk management of the pipeline should focus on this indicator.

[0099] In summary, compared with the prior art, the beneficial effects of this embodiment are:

[0100] 1. This embodiment establishes an oil and gas pipeline risk indicator assessment system based on the analysis of oil and gas pipeline accident cases, so that the risk assessment is more in line with the on-site process.

[0101] 2. This embodiment applies the risk assessment indicator occurrence frequency-consequence severity matrix to the AHP to improve the AHP so that the AHP no longer needs to perform consistency checks when calculating weights, thus saving calculation costs.

[0102] 3. Based on the risk level classification criteria, a dual-hidden layer adaptive regularized neural network is integrated. This network uses a dual-hidden layer decreasing structure to gradually refine the characteristics of the risk level, avoiding information redundancy. The hyperbolic tangent function (tansig), which is more suitable for small sample training, is used as the activation function to mitigate the risk of gradient vanishing. A dynamically adjusted early stopping mechanism is set to adaptively control the number of training rounds to balance model complexity and generalization ability. An L2 regularization weight decay term is introduced in the cross-entropy loss function, and used in conjunction with the early stopping mechanism to achieve a double overfitting effect. To adapt to the regularization and early stopping mechanisms, the conjugate gradient optimization algorithm (SCG) is optimized so that it only requires first-order derivative information and can automatically update the learning rate, converge quickly within the number of iterations, and meet the requirements of high-frequency data updates. Through the coordinated design of structural optimization and training strategies, the risk level assigned to each risk assessment indicator is accurately given. The neural network structure supports the expansion of input dimensions and enhances the generalization ability of the model.

[0103] 4. Based on the risk value and weight of each risk assessment indicator, the SHAP (Shapley Additive Explanations) framework is introduced to analyze the contribution of each risk assessment indicator to the final assessment result, obtain the risk priority of each risk assessment indicator, and improve the transparency of the model.

[0104] Specific implementation examples of this embodiment include:

[0105] By statistically analyzing the oil and gas pipeline accident data on the PHMSA website, a risk assessment was conducted on an oil and gas pipeline built in 2023. The oil and gas pipeline safety is used as the first-level risk assessment indicator F; the second-level risk assessment indicators are corrosion factors S1, structural defects S2, third-party damage S3, natural disasters S4, and physical damage S5; the third-level risk indicators are internal corrosion T1, external corrosion T2, corrosion prevention T3, accessory defects T4, equipment body failure T5, accessory joint failure T6, manufacturing defects T7, equipment operation failure T8, operator damage T9, and previous damage T10. 10 、Third party damages T 11 、Construction defects T 12 , Environmental protection related material defects T 13 、Contractor destroys T 14 、Earthquake T 15 、Extreme Weather T 16 、Other natural disasters T 17 、Arc fire T 18 、External force damage T 19 , Mechanical damage T 20 、Vehicle impact T 21 .

[0106] The data collected from the PHMSA website are organized as shown in Table 1. Based on the data in Table 1, the dynamic frequency-consequence matrix of the secondary risk assessment indicators and the tertiary risk assessment indicators is established as shown in Table 2. The weights of the secondary risk assessment indicators corrosion factor S1, structural defect S2, third-party damage S3, natural disaster S4, and physical damage S5 are 0.4185, 0.2625, 0.1599, 0.0618, and 0.0973 respectively. The tertiary risk assessment indicators are internal corrosion T1, external corrosion T2, corrosion prevention T3, accessory defect T4, equipment body failure T5, accessory joint failure T6, manufacturing defect T7, equipment operation failure T8, operator damage T9, and previous damage T10. 10 、Third party damages T 11 、Construction defects T 12 , Environmental protection related material defects T 13 、Contractor destroys T 14 、Earthquake T 15 、Extreme Weather T 16 、Other natural disasters T 17 、Arc fire T 18 、External force damage T 19 , Mechanical damage T 20 、Vehicle impact T 21 The weights are: 0.1634, 0.2970, 0.5396, 0.1578, 0.0888, 0.2978, 0.1578, 0.2978, 0.2057, 0.0492, 0.3422, 0.1219, 0.0753, 0.2057, 0.25, 0.5, 0.25, 0.1409, 0.2628, 0.1409, 0.4554.

[0107] Table 1 Risk indicator evaluation data

[0108]

[0109] Table 2 Risk assessment indicator frequency-consequence matrix

[0110]

[0111]

[0112] To address the overfitting problem caused by high data dimensionality, sparse samples, and high environmental noise in oil and gas pipeline risk assessment, this study proposes a dual-hidden-layer adaptive regularized neural network (DHAR-Net). Through the coordinated design of structural optimization and training strategies, the model's generalization ability is significantly improved, accelerating the risk assessment of new pipelines. Oil and gas pipeline risk data typically contains multi-source heterogeneous features, and traditional single-hidden-layer networks have difficulty capturing nonlinear coupling relationships. To this end, this model adopts a descending dual-hidden-layer structure with 8-6 neurons:

[0113] First-layer feature fusion: 8 neurons are used to fuse input features across dimensions, using the hyperbolic tangent function (tansig) as the activation function to generate a highly discriminative intermediate representation. The traditional ReLU function's characteristic of returning to zero gradients in the negative range can cause some weak risk signals to be lost, while the Sigmoid function's vanishing gradient problem in the saturation region affects convergence speed. Therefore, this model uses the hyperbolic tangent function (tansig) as the hidden layer activation function:

[0114] Second-layer risk compression: 6 neurons selectively enhance the fusion features, suppress redundant noise, and output low-dimensional risk-sensitive features.

[0115] This process simulates the hierarchical cognitive process of "feature screening-decision concentration" in industrial risk assessment, which is consistent with the risk level classification logic in the ASMEB31.8 standard.

[0116] According to the established three-level risk assessment indicator frequency-consequence matrix, the risk level is divided into four levels: Level 1 (U1) represents low risk; Level 2 (U2) represents medium risk; Level 3 (U3) represents high risk; Level 4 (U4) represents extreme risk. The corresponding risk value intervals are: D1[0,4), D2[4,9), D3[9,16), D4[16,25]. The risk values ​​of each risk assessment indicator are brought into the double hidden layer adaptive regularized neural network (DHAR-Net) designed in this paper, and the results are shown in Table 3. As can be seen from Table 3, the accuracy of the double hidden layer adaptive regularized neural network (DHAR-Net) for the samples in this paper can reach 100%;

[0117] Table 3 Risk assessment index evaluation results

[0118]

[0119]

[0120] The SHAP (Shapley Additive Explanations) framework

[21] is introduced to analyze the contribution of each risk indicator to the final evaluation result. The SHAP contribution of each risk evaluation indicator is as follows: Figure 2 As shown. Figure 2It can be seen that the contribution of each risk assessment indicator to the final assessment result is calculated as follows: 0.0326, 0.0592, 0.1075, 0.0197, 0.0111, 0.0372, 0.0197, 0.0372, 0.0157, 0.0038, 0.0260, 0.0093, 0.0057, 0.0157, 0.0074, 0.0147, 0.0074, 0.0065, 0.0122, 0.0065, 0.0074, and the risk priority of each risk assessment indicator is shown in Table 11. Among them, the contribution of the risk assessment indicator internal corrosion T2 is significantly higher than that of other indicators, indicating that the risk assessment indicator internal corrosion T2 plays a decisive role in risk assessment, and future risk management of the pipeline should focus on this indicator.

[0121] Oil and gas pipeline risk assessment technology is a crucial tool for ensuring safe operation of oil and gas pipelines. This paper addresses the issues of previous oil and gas pipeline risk assessment models, which suffer from weak generalization capabilities and over-reliance on expert evaluation, leading to uncertainty in the assessment of risk indicators and consequently inaccurate assessment results. This paper proposes a method for constructing a quantitative risk assessment model for oil and gas pipelines based on an improved analytic hierarchy process (AHP) and a dual-hidden-layer adaptive regularized neural network. This method improves the accuracy of oil and gas pipeline risk assessment by combining the improved AHP and dual-hidden-layer adaptive regularized neural network. This method constructs a dynamic frequency-consequence matrix for risk assessment indicators and incorporates it into the AHP. This eliminates the reliance on expert scoring and consistency checks, resulting in more accurate weights. A comprehensive weighting system is constructed to obtain more accurate weights for the three-level risk indicators. The constructed dual-hidden-layer adaptive regularized neural network extracts the characteristics of each risk assessment indicator and calculates its risk level. The SHAP contribution of each risk assessment indicator, calculated using the Shapley Additive Explanations framework, is ranked in descending order to determine its priority. Finally, the improved AHP and dual-hidden-layer adaptive regularized neural network quantitative risk assessment model construction method for oil and gas pipelines was applied to an oil and gas pipeline risk assessment case study. The results showed that the improved AHP and dual-hidden-layer adaptive regularized neural network quantitative risk assessment model construction method for oil and gas pipelines produced more accurate assessment results. In oil and gas pipeline risk assessment, the improved AHP and dual-hidden-layer adaptive regularized neural network quantitative risk assessment model for oil and gas pipelines effectively addresses the poor generalization capabilities of previous oil and gas pipeline risk assessment models and their over-reliance on expert evaluation, which leads to uncertainty in the judgment of risk assessment indicators and thus inaccurate assessment results. However, the established risk indicator system relies on the analysis of previous oil and gas pipeline accident cases, which may increase the subjectivity of the model. The improved AHP and dual-hidden-layer adaptive regularized neural network quantitative risk assessment model construction method has strong generalization capabilities and can be combined with other risk assessment cases with complex processes and multiple risk factors. For different fields, the assessment purpose can be achieved by changing the risk indicator system.

[0122] This embodiment may also provide a quantitative risk assessment system for oil and gas pipelines, including:

[0123] The indicator system construction module is used to analyze accident cases of oil and gas pipelines and build a multi-level risk assessment indicator system based on the analysis results;

[0124] A risk level prediction module is used to calculate the risk value and weight of each risk assessment indicator in the multi-level risk assessment indicator system; input the risk value of each risk assessment indicator into the risk level assessment model for classification prediction to obtain the corresponding risk level; wherein the risk level assessment model is constructed based on a dual hidden layer adaptive regularized neural network;

[0125] The risk priority calculation module is used to calculate the corresponding SHAP value based on the weight and risk level of each risk assessment indicator; sort the risk assessment indicators according to the size of the SHAP value to obtain the risk priority; and conduct safety monitoring of oil and gas pipelines based on the risk level assessment results.

[0126] This embodiment may further provide an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for quantitative risk assessment of oil and gas pipelines.

[0127] This embodiment may further provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method for quantitative risk assessment of oil and gas pipelines.

[0128] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A quantitative risk assessment method for oil and gas pipelines, characterized in that: include: Analyze oil and gas pipeline accident cases and build a multi-level risk assessment indicator system based on the analysis results; Calculating the risk value and weight of each risk assessment indicator in each of the multi-level risk assessment indicator systems; The risk value of each risk assessment indicator is input into the risk level assessment model for classification prediction to obtain the corresponding risk level; wherein the risk level assessment model is constructed based on a double hidden layer adaptive regularized neural network; Calculate the corresponding SHAP value based on the weight and risk level of each risk assessment indicator; Sort each risk assessment indicator according to the size of the SHAP value to obtain the risk priority; Carry out safety monitoring of oil and gas pipelines based on risk level assessment results.

2. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The multi-level risk assessment indicator system includes first-level risk assessment indicators, second-level risk assessment indicators and third-level risk assessment indicators.

3. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The risk value calculation process specifically includes: Constructing a frequency-consequence matrix corresponding to the multi-level risk assessment indicator system according to the LEC evaluation method; The risk value of each risk assessment indicator is calculated based on the constructed matrix.

4. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The weight calculation process, Based on the improved hierarchical analysis method, the indicators of the secondary risk assessment indicators and the initial single weights of the tertiary risk assessment indicators are calculated. The weights of the secondary risk assessment indicators are used as the intermediate event weights in the comprehensive weight method, and the initial single weights are used as the single event weights in the comprehensive weight method. The intermediate event weights and the single event weights are multiplied together to obtain the comprehensive weights of the tertiary risk assessment indicators.

5. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The training process of the risk level assessment model: Acquiring training data, wherein the training data includes risk training values ​​and corresponding risk levels; An initial risk level evaluation model is constructed, the training data is input into the initial risk level evaluation model for classification prediction, and training is performed with the goal of minimizing the loss between the initial training result after classification prediction and the risk level corresponding to the risk training value to obtain a trained risk level evaluation model.

6. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The processing process of the risk level evaluation model specifically includes: The risk value of each risk assessment indicator is input into a double hidden layer adaptive regularized neural network, the risk value is discretized into multiple risk levels and used as output to obtain the corresponding risk level.

7. The method for quantitative risk assessment of oil and gas pipelines according to claim 1, characterized in that: The corresponding SHAP value is calculated based on the weight of each risk assessment indicator and the risk level. The specific calculation process is: baseline=average(RV) SHAP_values=(RV-baseline)·W3 Where baseline is the average of all risk values, which is used to measure the difference between individual risk values ​​and the overall average level; average is the average of the risk values ​​of all risk assessment indicators; RV is the risk value of all risk assessment indicators; SHAP_values ​​is the SHAP value of each risk assessment indicator; and W3 is the comprehensive weight of the three-level risk assessment indicators.

8. A quantitative risk assessment system for oil and gas pipelines, characterized by: include: The indicator system construction module is used to analyze accident cases of oil and gas pipelines and build a multi-level risk assessment indicator system based on the analysis results; A risk level prediction module is used to calculate the risk value and weight of each risk assessment indicator in the multi-level risk assessment indicator system; The risk value of each risk assessment indicator is input into the risk level assessment model for classification prediction to obtain the corresponding risk level; wherein the risk level assessment model is constructed based on a double hidden layer adaptive regularized neural network; The risk priority calculation module is used to calculate the corresponding SHAP value based on the weight and risk level of each risk assessment indicator; sort the risk assessment indicators according to the size of the SHAP value to obtain the risk priority; and conduct safety monitoring of oil and gas pipelines based on the risk level assessment results.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a quantitative risk assessment method for an oil and gas pipeline according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that It stores a computer program, which, when executed by a processor, implements a quantitative risk assessment method for oil and gas pipelines according to any one of claims 1 to 7.

Images