Intelligent monitoring method and system for full-life-cycle safety state of oil and gas pipeline

Abstract

The invention provides an intelligent monitoring method and system for the full life cycle safety state of an oil and gas pipeline, and relates to the technical field of oil and gas pipelines, and the method comprises the steps: collecting the operation state data of the oil and gas pipeline, building a digital twin model, constructing a multi-dimensional feature vector, and carrying out the feature fusion; and constructing a fault evolution prediction model based on the fusion feature vector in combination with a graph neural network, calculating a residual life prediction value, continuously optimizing the safety state evaluation accuracy through online iterative training, and generating a safety state evaluation report and a maintenance decision suggestion. Intelligent monitoring and prediction of the safety state of the oil and gas pipeline are achieved, and the reliability of safe operation of the pipeline is improved.

Description

Technical Field

[0001] The present invention relates to oil and gas pipeline technology, and in particular to a method and system for intelligently monitoring the safety status of oil and gas pipelines throughout their entire life cycle. Background Art

[0002] With the rapid development of the oil and gas industry, oil and gas pipelines, as crucial infrastructure for energy transportation, have become crucial for national energy security and socioeconomic stability. Over their long service lives, oil and gas pipelines face numerous threats, including corrosion, fatigue, and third-party sabotage, which can lead to safety incidents such as leaks and explosions. Traditional oil and gas pipeline safety management relies primarily on regular inspections and manual patrols, which are inefficient and make it difficult to monitor the safety status of pipelines throughout their entire lifecycle in real time.

[0003] In recent years, with the development of emerging technologies such as the Internet of Things, big data, and artificial intelligence, oil and gas pipeline safety monitoring technology has gradually moved towards digitalization and intelligence. Currently, the industry has begun using various sensors to collect real-time pipeline operation data and conduct safety status assessments based on this data. However, existing oil and gas pipeline safety monitoring technology still has the following shortcomings:

[0004] First, most existing technologies only focus on single-dimensional pipeline safety parameter monitoring, such as monitoring only pressure, temperature or corrosion conditions, and lack comprehensive analysis of multi-dimensional data such as pipeline structural characteristics, environmental characteristics and historical fault characteristics, making it difficult to fully reflect the actual safety status of the pipeline.

[0005] Secondly, traditional pipeline safety assessment methods usually use static assessment models, which cannot effectively describe the dynamic evolution process of pipeline failures. The prediction accuracy of the remaining life of the pipeline is limited and it cannot provide timely warnings of potential safety risks.

[0006] Finally, existing pipeline safety status assessment systems generally lack self-learning and self-optimization capabilities and are unable to continuously optimize assessment models based on newly added pipeline operation data and failure cases. This results in insufficient timeliness and accuracy in safety assessment results, making it difficult to provide a reliable scientific basis for pipeline maintenance decisions. Summary of the Invention

[0007] The embodiments of the present invention provide a method and system for intelligently monitoring the safety status of oil and gas pipelines throughout their entire life cycle, which can solve the problems in the prior art.

[0008] A first aspect of an embodiment of the present invention provides an intelligent monitoring method for the safety status of an oil and gas pipeline throughout its entire life cycle, comprising:

[0009] Collecting oil and gas pipeline operation status data, the oil and gas pipeline operation status data including pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data;

[0010] Establishing a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; constructing a multi-dimensional feature vector based on the oil and gas pipeline digital twin model, the multi-dimensional feature vector including a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector;

[0011] Performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network;

[0012] Calculating a predicted value of the remaining service life of the oil and gas pipeline based on the output result of the oil and gas pipeline fault evolution prediction model;

[0013] Based on the predicted value of the remaining life of the oil and gas pipeline and combined with the pipeline safety status assessment rule base, the accuracy of the safety status assessment is continuously optimized through online iterative training; based on the optimized safety status assessment accuracy, an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations are generated.

[0014] Establishing a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; and constructing a multi-dimensional feature vector based on the oil and gas pipeline digital twin model, including:

[0015] constructing a physical-virtual mapping matrix based on the oil and gas pipeline operation status data, wherein the physical-virtual mapping matrix includes a plurality of mapping relationship coefficients;

[0016] Establishing a digital twin model of the oil and gas pipeline according to the physical-virtual mapping matrix to realize virtual simulation of the operating status data of the oil and gas pipeline;

[0017] Obtaining input variables and model parameters, constructing a state update equation based on the input variables and the model parameters, and dynamically updating the oil and gas pipeline digital twin model through the state update equation;

[0018] Extracting multi-dimensional features based on the oil and gas pipeline digital twin model includes: calculating pressure fluctuation features based on the variance of pressure sampling values ​​and average pressure, calculating temperature gradient features based on the partial derivative of the temperature field in three-dimensional space, and calculating environmental impact factors based on weight coefficients of various environmental parameters, wherein the weight coefficients are determined through historical data analysis;

[0019] The pressure fluctuation characteristics, the temperature gradient characteristics and the environmental influencing factors are fused to construct a multi-dimensional feature vector of the oil and gas pipeline.

[0020] Performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector and combining it with a graph neural network includes:

[0021] The multidimensional feature vector is mapped into a state feature using a state mapping weight matrix; a time feature is constructed based on a timestamp, a period code, and a feature change rate; a spatial feature is constructed based on a position coordinate, an azimuth angle, and a feature space gradient; and the state feature, the time feature, and the space feature are combined to form a spatiotemporal knowledge node;

[0022] Extracting graph features based on the spatiotemporal knowledge nodes through node feature aggregation operations, wherein the node feature aggregation operations use attention weights to perform weighted combination of adjacent node features, wherein the attention weights are calculated through spatiotemporal attention;

[0023] Using a graph neural network to process the graph features to obtain enhanced features, and combining the enhanced features with the multidimensional feature vector to obtain a fused feature vector;

[0024] Performing causal chain discovery on the fused feature vector, including: calculating the causal relationship between features through a conditional independence test, obtaining causal strength based on a causal gradient; and constructing a causal attention mechanism based on the causal relationship and the causal strength;

[0025] The causal attention mechanism is combined with the fusion feature vector, and an oil and gas pipeline fault evolution prediction model is constructed through the graph neural network.

[0026] The causal relationship between features is calculated through conditional independence testing, and the causal strength is obtained based on the causal gradient. The causal attention mechanism is constructed based on the causal relationship and the causal strength, including:

[0027] Acquire time series feature data and state feature data, calculate conditional entropy of the time series feature data and the state feature data, perform a conditional independence test based on the conditional entropy, and obtain a conditional independence test result between features;

[0028] Performing conditional probability modeling on the time series feature data, calculating the probabilistic dependency between features using a Bayesian network, and determining the causal relationship between features based on the result of the conditional independence test and the probabilistic dependency;

[0029] Calculating the causal gradient between features based on the causal relationship, and obtaining the direct causal strength according to the causal gradient; calculating the indirect causal strength between features along the causal transmission path; and combining the direct causal strength and the indirect causal strength to obtain the comprehensive causal strength;

[0030] The comprehensive causal strength is combined with the feature representation to generate a causal query vector and a causal key-value vector; a causal attention matrix is ​​constructed based on the causal query vector and the causal key-value vector; and the feature representation is reconstructed according to the causal attention matrix to obtain a causal attention mechanism.

[0031] Calculating the predicted value of the remaining service life of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model includes:

[0032] Decoupling the output results of the oil and gas pipeline fault evolution prediction model according to the failure mechanism to construct a crack evolution feature vector and a corrosion degradation feature vector;

[0033] Establishing a Wiener random process model based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, combining the drift coefficient, the diffusion coefficient and the measurement noise to construct a degradation state evolution equation; introducing a nonlinear compensation term into the degradation state evolution equation;

[0034] Calculating fatigue damage accumulation based on the Wiener random process model, and calculating plastic strain in combination with the corrosion rate in the corrosion degradation characteristic vector and the crack evolution characteristic vector; performing a weighted combination of the fatigue damage accumulation, the corrosion rate, and the plastic strain to obtain a physical life consumption rate;

[0035] The physical life consumption rate is input into the degradation state evolution equation to obtain a data-driven life prediction value; the weight coefficients of the physical life consumption rate and the data-driven life prediction value are adaptively adjusted according to the prediction error to output the remaining life prediction value of the oil and gas pipeline.

[0036] A Wiener random process model is established based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, and a degradation state evolution equation is constructed by combining the drift coefficient, the diffusion coefficient and the measurement noise. The nonlinear compensation term is introduced into the degradation state evolution equation, including:

[0037] Obtaining degradation state observation values ​​at multiple time points, wherein the degradation state observation values ​​include state measurement data and timestamp information;

[0038] Calculating the difference between state observation values ​​at adjacent time points and dividing the difference by the corresponding time interval to obtain a degradation rate; processing the degradation rate using a maximum likelihood estimation method to obtain a drift coefficient; performing a square sum operation on the deviation between the degradation rate and the drift coefficient to obtain a diffusion coefficient;

[0039] Performing a time domain analysis on the measurement noise in the degradation state observation value, calculating the autocorrelation function of the measurement noise under different time delays; performing a Fourier transform on the autocorrelation function to obtain a frequency domain characteristic;

[0040] Based on the frequency domain characteristics, the state is predicted, and the difference between the predicted value and the observed value is dynamically corrected in combination with the Kalman gain to eliminate the influence of measurement noise to obtain the state estimation value;

[0041] Performing local linearization processing on the state estimate, calculating a nonlinear approximation term through a second-order Taylor expansion; constructing a dynamic compensation term based on the nonlinear approximation term combined with a historical state data sequence and a historical prediction error;

[0042] The drift coefficient, the diffusion coefficient, the state estimation value and the dynamic compensation term are combined to construct a degradation state evolution equation, wherein the degradation state evolution equation includes a Brownian motion term, a filtered noise term and a nonlinear compensation term, and is used to predict the degradation state at a future moment.

[0043] Based on the optimized safety status assessment accuracy, a pipeline safety status assessment report and pipeline maintenance decision recommendations are generated, including:

[0044] The optimized safety status assessment accuracy includes the assessment indicator accuracy and the assessment model confidence; a correction coefficient is calculated based on the assessment model confidence, and the initial weight of the assessment indicator is linearly corrected based on the correction coefficient to obtain a dynamic correction weight; the assessment indicator and the dynamic correction weight are weighted and summed to obtain a comprehensive safety status score;

[0045] Performing K-means cluster analysis on historical assessment data to obtain initial risk level classification criteria; using the assessment model confidence as an adaptive factor, updating the initial classification criteria using an exponential sliding weighted average method; matching the comprehensive safety status score with the updated classification criteria to determine the risk level classification criteria;

[0046] A maintenance priority assessment model is constructed based on the risk level classification standard, and the risk level, maintenance difficulty, and time urgency are quantified and input into the maintenance priority assessment model; the assessment factor weights are set using the assessment indicator accuracy, and the maintenance priority score is calculated by weighted summation;

[0047] Writing the comprehensive safety status score, the risk level classification standard, and the maintenance priority score into a preset assessment report template in order of importance to generate a pipeline safety status assessment report;

[0048] A word extraction algorithm is used to identify assessment information from the pipeline safety status assessment report, and the assessment information is matched with a maintenance decision rule library to generate a pipeline maintenance decision suggestion including maintenance content, maintenance sequence and resource allocation.

[0049] A second aspect of an embodiment of the present invention provides an intelligent monitoring system for the safety status of an oil and gas pipeline throughout its entire life cycle, comprising:

[0050] The first unit is used to collect oil and gas pipeline operation status data, wherein the oil and gas pipeline operation status data includes pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data;

[0051] The second unit is configured to establish a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; and construct a multi-dimensional feature vector based on the digital twin model of the oil and gas pipeline, wherein the multi-dimensional feature vector includes a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector;

[0052] The third unit is configured to perform feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; and construct an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network;

[0053] The fourth unit is used to calculate the remaining service life prediction value of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model;

[0054] The fifth unit is used to continuously optimize the safety status assessment accuracy through online iterative training based on the remaining life prediction value of the oil and gas pipeline and the pipeline safety status assessment rule base; and generate an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations based on the optimized safety status assessment accuracy.

[0055] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0056] processor;

[0057] a memory for storing processor-executable instructions;

[0058] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0059] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0060] The beneficial effects of this application are as follows:

[0061] The present invention establishes a digital twin model by collecting oil and gas pipeline operating status data, constructs a multi-dimensional feature vector for feature fusion, and combines it with a graph neural network to build a fault evolution prediction model, thereby realizing intelligent monitoring of the entire life cycle of oil and gas pipelines and improving the accuracy and reliability of safety status assessment.

[0062] The present invention uses online iterative training to continuously optimize the accuracy of safety status assessment, can promptly adapt to changes in pipeline operating environment and status, dynamically adjust the assessment model, make the safety status assessment results more in line with actual conditions, and provide more accurate pipeline maintenance decision recommendations.

[0063] The present invention integrates and analyzes pipeline structural characteristics, environmental characteristics, and historical fault characteristics, comprehensively considering the impact of multiple factors on pipeline safety, breaking through the limitations of traditional single-dimensional assessment, making the remaining life prediction of oil and gas pipelines more scientific and reasonable, effectively reducing the risk of pipeline safety accidents, extending the service life of pipelines, and improving the economic benefits of pipeline operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Schematic diagram of the process of an intelligent monitoring method for the safety status of an oil and gas pipeline throughout its life cycle according to an embodiment of the present invention;

[0065] Figure 2 This is a bar chart comparing the performance of the oil and gas pipeline fault evolution prediction model according to an embodiment of the present invention;

[0066] Figure 3 Construct a flow chart for the causal attention mechanism based on conditional independence and Bayesian networks in an embodiment of the present invention;

[0067] Figure 4 This is a schematic diagram of the relationship between fatigue damage accumulation and prediction error according to an embodiment of the present invention;

[0068] Figure 5 This is a bar chart comparing the performance of the pipeline safety status assessment and maintenance decision-making method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0070] The technical solution of the present invention is described in detail below with reference to specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0071] Figure 1 FIG. 1 is a flow chart of an intelligent monitoring method for the safety status of an oil and gas pipeline throughout its life cycle according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0072] Collecting oil and gas pipeline operation status data, the oil and gas pipeline operation status data including pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data;

[0073] Establishing a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; constructing a multi-dimensional feature vector based on the oil and gas pipeline digital twin model, the multi-dimensional feature vector including a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector;

[0074] Performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network;

[0075] Calculating a predicted value of the remaining service life of the oil and gas pipeline based on the output result of the oil and gas pipeline fault evolution prediction model;

[0076] Based on the predicted value of the remaining life of the oil and gas pipeline and combined with the pipeline safety status assessment rule base, the accuracy of the safety status assessment is continuously optimized through online iterative training; based on the optimized safety status assessment accuracy, an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations are generated.

[0077] In an optional embodiment, a digital twin model of an oil and gas pipeline is established based on the oil and gas pipeline operation status data; and constructing a multi-dimensional feature vector based on the oil and gas pipeline digital twin model includes:

[0078] constructing a physical-virtual mapping matrix based on the oil and gas pipeline operation status data, wherein the physical-virtual mapping matrix includes a plurality of mapping relationship coefficients;

[0079] Establishing a digital twin model of the oil and gas pipeline according to the physical-virtual mapping matrix to realize virtual simulation of the operating status data of the oil and gas pipeline;

[0080] Obtaining input variables and model parameters, constructing a state update equation based on the input variables and the model parameters, and dynamically updating the oil and gas pipeline digital twin model through the state update equation;

[0081] Extracting multi-dimensional features based on the oil and gas pipeline digital twin model includes: calculating pressure fluctuation features based on the variance of pressure sampling values ​​and average pressure, calculating temperature gradient features based on the partial derivative of the temperature field in three-dimensional space, and calculating environmental impact factors based on weight coefficients of various environmental parameters, wherein the weight coefficients are determined through historical data analysis;

[0082] The pressure fluctuation characteristics, the temperature gradient characteristics and the environmental influencing factors are fused to construct a multi-dimensional feature vector of the oil and gas pipeline.

[0083] In the stage of obtaining the operating status data of the oil and gas pipeline, the system collects data through the sensor network distributed at each key node of the pipeline. These data include pressure data, temperature data, flow data, environmental data, etc. For example, on a 100-kilometer oil pipeline, a sensor node is set every 5 kilometers, for a total of 20 nodes. Each node collects pressure (0-10MPa range), temperature (-20℃ to 80℃ range), flow (0-2000m 3 / h range) and other parameters, and the sampling frequency is 1 time / minute.

[0084] Constructing a physical-virtual mapping matrix is ​​a key step in building a digital twin model. This mapping matrix contains multiple mapping coefficients, which are used to map parameters in the physical world to the virtual space. In implementation, the system analyzes historical operating data, identifies correlations between parameters, and establishes these mappings. For example, for the mapping relationship between pressure P and flow Q, analyzing 10,000 sets of historical data determined a mapping coefficient of 0.85; for temperature T and ambient temperature Te, a mapping coefficient of 0.72; and for pressure fluctuation and pipeline wear, a mapping coefficient of 0.68. These coefficients form a 15×15 mapping matrix that covers the mapping relationships between all key parameters.

[0085] When building a digital twin model of an oil and gas pipeline based on a physical-virtual mapping matrix, the system uses virtual simulation technology to construct a three-dimensional geometric model and an operational status model of the pipeline. The geometric model accurately restores the pipeline's spatial layout, material properties, and connection methods, including parameters such as pipe diameter, wall thickness, and bend angle. The operational status model uses a mapping matrix to convert real-time collected physical data into state variables in the virtual model. For example, when the pressure in a certain section of the actual pipeline rises to 8.5 MPa, the pressure value at the corresponding location in the virtual model is updated and visually displayed through a color change (e.g., from green to yellow).

[0086] To dynamically update the digital twin model of oil and gas pipelines, the system acquires input variables and model parameters and constructs a state update equation. Input variables include real-time monitored data such as pressure, temperature, and flow rate; model parameters include pipeline material parameters, medium physical parameters, and environmental impact coefficients. The state update equation uses time series analysis to predict the system state at the next moment.

[0087] When predicting pipeline pressure changes, the system takes into account the current pressure value of 7.2MPa and flow rate of 1500m 3 / h, temperature 45℃ and other input variables, combined with the pipeline roughness coefficient 0.025, medium density 850kg / m3 Using the same model parameters, the system predicts that the pressure will reach 7.5 MPa in 0.5 hours. The system updates its status every 10 minutes to ensure that the virtual model remains highly consistent with the actual pipeline status.

[0088] Based on the digital twin model of oil and gas pipelines, the system extracts multi-dimensional features. Pressure fluctuation features are obtained by calculating the variance between the sampled pressure values ​​and the average pressure. For example, if the average pressure at a monitoring point over 24 hours is 6.8 MPa with a standard deviation of 0.15 MPa, the system calculates a pressure fluctuation feature value of 0.0225, indicating good pressure stability at that point. Temperature gradient features are obtained by calculating the partial derivative of the temperature field in three-dimensional space.

[0089] The pipeline was divided into 500 temperature monitoring areas, and the temperature change rate of each area was calculated. For example, in a certain section, the temperature along the route was detected to drop from 55°C to 45°C, with a length of 2,000 meters, resulting in a temperature gradient of 0.005°C / meter. The environmental impact factor was calculated by weighting each environmental parameter. The system analyzed three years of historical data and determined the weight coefficients of the environmental parameters: air temperature 0.35, humidity 0.25, geological conditions 0.30, and corrosion environment 0.10. The environmental parameters of a certain section of pipeline were: air temperature 35°C, humidity 65%, geological conditions score 80 points (out of 100), and corrosion environment score 30 points (out of 100). The calculated environmental impact factor was 65.5.

[0090] The pressure fluctuation characteristics, temperature gradient characteristics, and environmental impact factors are fused to construct a multidimensional feature vector for the oil and gas pipeline. This feature fusion utilizes normalization and vector concatenation. For the example data above, the pressure fluctuation characteristic of 0.0225 is normalized to 0.45, the temperature gradient characteristic of 0.005°C / m is normalized to 0.33, and the environmental impact factor of 65.5 is normalized to 0.655.

[0091] These normalized features are concatenated to form a feature vector [0.45, 0.33, 0.655, ...]. This vector can be expanded to include other feature dimensions, such as flow rate change 0.28 and pipeline vibration amplitude 0.15. The resulting multi-dimensional feature vector comprehensively characterizes the operational status of the oil and gas pipeline, providing a data foundation for subsequent status assessment and fault warning.

[0092] In an optional embodiment, performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; and constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector in combination with a graph neural network includes:

[0093] The multidimensional feature vector is mapped into a state feature using a state mapping weight matrix; a time feature is constructed based on a timestamp, a period code, and a feature change rate; a spatial feature is constructed based on a position coordinate, an azimuth angle, and a feature space gradient; and the state feature, the time feature, and the space feature are combined to form a spatiotemporal knowledge node;

[0094] Extracting graph features based on the spatiotemporal knowledge nodes through node feature aggregation operations, wherein the node feature aggregation operations use attention weights to perform weighted combination of adjacent node features, wherein the attention weights are calculated through spatiotemporal attention;

[0095] Using a graph neural network to process the graph features to obtain enhanced features, and combining the enhanced features with the multidimensional feature vector to obtain a fused feature vector;

[0096] Performing causal chain discovery on the fused feature vector, including: calculating the causal relationship between features through a conditional independence test, obtaining causal strength based on a causal gradient; and constructing a causal attention mechanism based on the causal relationship and the causal strength;

[0097] The causal attention mechanism is combined with the fusion feature vector, and an oil and gas pipeline fault evolution prediction model is constructed through the graph neural network.

[0098] The multi-dimensional feature vectors are fused, and then an oil and gas pipeline fault evolution prediction model is constructed based on the fused feature vectors and combined with graph neural network.

[0099] To perform feature fusion on multidimensional feature vectors, the system collects sensor data from multiple monitoring points along the oil and gas pipeline, including parameters such as pressure, temperature, flow rate, and vibration frequency. For example, a 500-meter-long oil and gas pipeline has 10 monitoring points distributed throughout, each collecting data with a dimension of 20. This data is preprocessed to form a multidimensional feature vector.

[0100] To map multidimensional feature vectors into state features using a state-mapping weight matrix, the system constructs a 20×15 weight matrix. The matrix's elements are determined based on expert experience and correlation analysis of historical data. For example, if the correlation between pressure features and leaks is 0.85, the corresponding weight element is set to 0.85. By multiplying this weight matrix with the original feature vector, the 20-dimensional original feature vector is mapped into a 15-dimensional state feature vector, thereby extracting the core information directly related to the pipeline status.

[0101] When constructing temporal features based on timestamps, period codes, and feature change rates, the system converts the timestamps of the collected data into time of day (0-24 hours), day of the week (1-7), and season (1-4). The feature change rate is also calculated. For example, if the pressure drops from 3.5 MPa to 3.2 MPa in 10 minutes, the rate of change is -0.03 MPa / minute. This time-related information is combined into an 8-dimensional temporal feature vector, which is used to capture the temporal changes in pipeline status.

[0102] When constructing spatial features based on location coordinates, azimuth, and characteristic spatial gradients, the system records the three-dimensional coordinates (x, y, z) of each monitoring point and the azimuth of the pipeline at that point. For example, the coordinates of a monitoring point are (125.6, 43.2, 15.3) and the azimuth is 37.5 degrees. Simultaneously, the characteristic spatial gradient between adjacent monitoring points is calculated. For example, if the temperature between two points drops from 45°C to 42°C at a distance of 50 meters, the temperature gradient is -0.06°C / meter. This spatial information is combined into a 10-dimensional spatial feature vector, which is used to characterize the spatial distribution characteristics of the pipeline status.

[0103] After combining state, time, and space features to form a spatiotemporal knowledge node, each monitoring point forms a 33-dimensional (15+8+10) feature vector, which serves as a node representation in the graph network. For example, a spatiotemporal knowledge node for a monitoring point contains state information such as pressure of 3.5 MPa and temperature of 45°C, time information such as 10:00 AM, weekday, and summer, as well as spatial information such as the specific location of the point and surrounding environmental characteristics.

[0104] When extracting graph features through node feature aggregation based on spatiotemporal knowledge nodes, the system defines a receptive field for each node, which is the set of nodes connected to it. For two adjacent nodes, the system calculates spatiotemporal attention weights between them. Temporal attention is calculated based on the difference in the time between the data collection of the two nodes. For example, a time difference of 5 minutes corresponds to a temporal attention weight of 0.9. Spatial attention is calculated based on the physical distance between the nodes. For example, a distance of 50 meters corresponds to a spatial attention weight of 0.85. Temporal and spatial attention are combined to form a comprehensive attention weight, which is used for weighted aggregation of node features. For example, if the feature vector of the central node is [3.5, 45, ...] and the feature vectors of the adjacent nodes are [3.4, 44, ...], the corresponding attention weight is 0.8. The aggregated feature components contain 0.2 × 3.5 + 0.8 × 3.4 = 3.42.

[0105] To enhance the features of the graph using a graph neural network, the system constructs a three-layer graph convolutional network, with output dimensions of 40, 30, and 25, respectively. The network continuously updates node representations through message passing, capturing high-order connections between nodes. The resulting 25-dimensional enhanced feature vector is concatenated with the original 20-dimensional feature vector to form a 45-dimensional fused feature vector.

[0106] When discovering causal chains for fused feature vectors, the system calculates the causal relationship between features through a conditional independence test. For example, the system discovered a causal relationship between pressure reduction and pipe wall thickness reduction, and the p-value of the conditional independence test was 0.02 (less than the significance level of 0.05). At the same time, the causal strength is estimated by calculating the intervention effect. For example, a 1MPa pressure reduction leads to a 25% increase in leakage risk, and the corresponding causal strength is 0.25. Based on the discovered causal relationship and strength, a causal graph is constructed, and a causal attention mechanism is designed to give higher weights to features in the causal chain. For example, when predicting pipeline leakage risk, the attention weights of the pressure and pipe wall thickness features are 0.3 and 0.25, respectively, while the weight of the ambient temperature feature, which has a weaker relationship with leakage, is only 0.05.

[0107] By combining the causal attention mechanism with the fused feature vectors, the system uses a graph neural network with a causal attention layer to construct a final oil and gas pipeline fault evolution prediction model. This model, consisting of two graph convolutional layers and three fully connected layers, can predict the failure risk probability and possible fault type for each monitoring point within the next 24 hours. In practical application, the model monitored a section of pipeline and successfully predicted that a corrosion point on the pipe wall would develop into a leak within 72 hours, with an 85% probability. This provided a 36-hour advance warning, enabling maintenance personnel to take timely action and avoid a potential safety incident.

[0108] Figure 2 This is a bar chart comparing the performance of the oil and gas pipeline fault evolution prediction model according to an embodiment of the present invention:

[0109] The figure compares the performance of three different models (the basic model, the spatiotemporal feature fusion model, and the graph neural network causal model) across five evaluation metrics. The graph neural network causal model performs best across all metrics, achieving a prediction accuracy of 94.3%, a fault detection rate of 92.1%, an advance warning time of 86.7%, a causal chain identification rate of 89.5%, and a model convergence rate of 91.2%. The spatiotemporal feature fusion model performs second best across all metrics, with a prediction accuracy of 88.7%, a fault detection rate of 84.6%, an advance warning time of 75.9%, a causal chain identification rate of 81.3%, and a model convergence rate of 83.6%. The basic model performs relatively poorly, with a prediction accuracy of 82.5%, a fault detection rate of 76.8%, an advance warning time of only 65.3%, a causal chain identification rate of 70.2%, and a model convergence rate of 78.4%. Data shows that after introducing graph neural networks and causal reasoning mechanisms, the model has achieved significant improvements in prediction accuracy, fault detection capabilities, warning timeliness, causal relationship identification and computational efficiency, with an average improvement of about 15-20 percentage points compared to the basic model, confirming the superiority of this method.

[0110] In an optional embodiment, calculating the causal relationship between features through a conditional independence test and obtaining the causal strength based on the causal gradient; and constructing a causal attention mechanism based on the causal relationship and the causal strength includes:

[0111] Acquire time series feature data and state feature data, calculate conditional entropy of the time series feature data and the state feature data, perform a conditional independence test based on the conditional entropy, and obtain a conditional independence test result between features;

[0112] Performing conditional probability modeling on the time series feature data, calculating the probabilistic dependency between features using a Bayesian network, and determining the causal relationship between features based on the result of the conditional independence test and the probabilistic dependency;

[0113] Calculating the causal gradient between features based on the causal relationship, and obtaining the direct causal strength according to the causal gradient; calculating the indirect causal strength between features along the causal transmission path; and combining the direct causal strength and the indirect causal strength to obtain the comprehensive causal strength;

[0114] The comprehensive causal strength is combined with the feature representation to generate a causal query vector and a causal key-value vector; a causal attention matrix is ​​constructed based on the causal query vector and the causal key-value vector; and the feature representation is reconstructed according to the causal attention matrix to obtain a causal attention mechanism.

[0115] like Figure 2 As shown, the method includes:

[0116] The causal relationship between features is calculated through conditional independence test, the causal strength is obtained based on the causal gradient, and a causal attention mechanism is constructed based on the causal relationship and the causal strength.

[0117] Acquire time series feature data and state feature data. Time series feature data can be data collected at multiple consecutive time points, such as the temperature, pressure, humidity, and other indicators recorded every hour for a device over the past 24 hours. State feature data can be discrete or continuous values ​​such as the device's current operating mode and operating status. For example, for a piece of production equipment, time series feature data includes motor temperature (35°C-45°C) and vibration frequency (20Hz-30Hz) recorded every hour for the past 10 days. State feature data includes the current load rate (85%) and operating mode (heavy load mode).

[0118] Calculate the conditional entropy of time series feature data and state feature data. Conditional entropy represents the uncertainty of one feature given the known value of the other. For each pair of features X and Y, calculate H(X|Y) and H(Y|X). For example, for the relationship between motor temperature (X) and load factor (Y), statistical analysis can yield H(X|Y) = 0.32 and H(Y|X) = 0.45.

[0119] The conditional independence test is performed based on the conditional entropy to obtain the conditional independence test results between features. The conditional independence test determines whether two features are independent of each other under the condition of a given third feature Z. 2 Tests are performed using statistical methods such as the χ2 test or chi-square test, with a significance level of 0.05. For example, if the motor temperature (X) and vibration frequency (Y) are tested under a given load factor (Z), and the p-value is 0.03 < 0.05, the independence hypothesis is rejected, and it is considered that X and Y are dependent under the given Z condition.

[0120] Conditional probability modeling is performed on time series feature data, using a Bayesian network to calculate the probabilistic dependencies between features. A Bayesian network is a directed acyclic graph, where nodes represent features and edges represent conditional dependencies. Scoring functions such as BIC (Bayesian Information Criterion) or MDL (Minimum Description Length) are used to evaluate the network structure. In the above example, the dependencies between load factor and motor temperature and load factor and vibration frequency are obtained, with BIC scores of -120.5 and -135.8, respectively.

[0121] The causal relationship between features is determined based on the results of the conditional independence test and probabilistic dependencies. Following the principle of causal discovery: if X and Y are dependent, X occurs before Y, and there is no feature Z that makes X and Y conditionally independent, then X is considered the cause of Y. Combining time series information, the causal direction between features is determined. For example, it can be determined that the load rate is the cause of motor temperature, which in turn is the cause of bearing wear, thus forming a causal graph structure.

[0122] Based on the causal relationship, the causal gradient between features is calculated, that is, the degree of influence of a feature change on its causal consequences. For directly connected features X and Y, the causal gradient is calculated by intervening in X and observing the change in Y. This can be achieved through counterfactual analysis: change X by ±10% from its current value, record the corresponding change in Y, and calculate the rate of change. For example, when the load factor increases from 85% to 93.5% (+10%), the motor temperature rises from 42°C to 45°C (+7.1%), and its direct causal strength can be calculated as 0.71.

[0123] Calculate the indirect causal strength between features along the causal transmission path. For features connected through intermediate nodes, calculate their indirect causal effects. Assuming feature A affects feature C through feature B, the indirect causal strength of feature A on feature C is the product of the direct causal strength of feature A on feature B and the direct causal strength of feature B on feature C. For example, the indirect causal strength of load factor on bearing wear is 0.71 (load factor on motor temperature) × 0.65 (motor temperature on bearing wear) = 0.46.

[0124] The combined causal strength is calculated by combining the direct and indirect causal strengths. A weighted summation approach can be used, with a weight of 0.7 for the direct causal strength and 0.3 for the indirect causal strength. For example, if the load factor has both a direct effect (0.25) and an indirect effect (0.46) on bearing wear through motor temperature, the combined causal strength is 0.7 × 0.25 + 0.3 × 0.46 = 0.313.

[0125] The comprehensive causal strength is combined with the feature representation to generate a causal query vector and a causal key-value vector. The feature representation can be a vector obtained through an embedding layer. Assuming the feature representation dimension is 64, the comprehensive causal strength is fused with the feature representation through a linear transformation. For example, for the load rate feature, its original representation vector is [0.12, 0.25, ..., 0.08]. Combining it with the comprehensive causal strength of other features [0.313, 0.28, ..., 0.15] yields a causal query vector [0.16, 0.26, ..., 0.09].

[0126] Construct a causal attention matrix based on the causal query vector and the causal key-value vector. Calculate the dot product of the query vector and the key-value vector, and perform softmax normalization. For example, generate an 8×8 attention matrix, where the value in the i-th row and j-th column represents the attention weight of the i-th feature on the j-th feature.

[0127] The causal attention mechanism is constructed by reconstructing the feature representation based on the causal attention matrix. The attention matrix is ​​multiplied by the eigenvalue vector to generate a new feature representation. For example, for the load rate feature, a weighted combination with the attention weights yields a new feature representation of [0.18, 0.27, ..., 0.11], which better incorporates causal information.

[0128] The causal attention mechanism constructed through the above steps can effectively capture the causal relationships and strengths between features, improving the model's understanding and predictive capabilities for complex systems. In practical applications, this mechanism can be integrated into deep learning models for scenarios such as device failure prediction and system anomaly detection.

[0129] In an optional embodiment, calculating the predicted remaining life of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model includes:

[0130] Decoupling the output results of the oil and gas pipeline fault evolution prediction model according to the failure mechanism to construct a crack evolution feature vector and a corrosion degradation feature vector;

[0131] Establishing a Wiener random process model based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, combining the drift coefficient, the diffusion coefficient and the measurement noise to construct a degradation state evolution equation; introducing a nonlinear compensation term into the degradation state evolution equation;

[0132] Calculating fatigue damage accumulation based on the Wiener random process model, and calculating plastic strain in combination with the corrosion rate in the corrosion degradation characteristic vector and the crack evolution characteristic vector; performing a weighted combination of the fatigue damage accumulation, the corrosion rate, and the plastic strain to obtain a physical life consumption rate;

[0133] The physical life consumption rate is input into the degradation state evolution equation to obtain a data-driven life prediction value; the weight coefficients of the physical life consumption rate and the data-driven life prediction value are adaptively adjusted according to the prediction error to output the remaining life prediction value of the oil and gas pipeline.

[0134] The output of the oil and gas pipeline fault evolution prediction model is decoupled according to the failure mechanism to construct crack evolution feature vectors and corrosion degradation feature vectors. Specifically, a non-negative matrix factorization algorithm is used to decompose the feature matrix output by the model, separating features related to crack evolution and features related to corrosion degradation.

[0135] The crack evolution feature vector includes parameters such as crack size, crack growth rate, and stress intensity factor; the corrosion degradation feature vector includes parameters such as corrosion depth, corrosion rate, and corrosion area. For example, for a 1000-meter-long X65 steel oil and gas pipeline, through feature decoupling, the crack evolution feature vector [2.3mm, 0.05mm / year, 25MPa·m^(1 / 2)] and the corrosion degradation feature vector [1.5mm, 0.08mm / year, 150cm 2 ].

[0136] Based on the eigenvectors obtained from the above decoupling, a Wiener stochastic process model is established. The drift coefficient, diffusion coefficient, and measurement noise are combined to construct the degradation state evolution equation. A nonlinear compensation term is introduced into this equation to improve the model's accuracy in describing the actual degradation process. Specifically, the drift coefficient represents the average rate of change of the degradation process, the diffusion coefficient represents the randomness of the degradation process, and the measurement noise represents the uncertainty of the monitoring data. For the oil and gas pipeline in the above example, the drift coefficient can be set to 0.12, the diffusion coefficient can be set to 0.05, and the measurement noise standard deviation can be set to 0.02. The nonlinear compensation term can be constructed using a cubic spline interpolation function to capture the nonlinear characteristics of the degradation process.

[0137] Once the degradation state evolution equation is constructed, fatigue damage accumulation is calculated based on the Wiener stochastic process model. This calculation takes into account the cyclic loading experienced by the oil and gas pipeline during service and quantifies it using cumulative damage theory. For the example oil and gas pipeline, assuming it has been in service for 10 years, the calculated fatigue damage accumulation is 0.35 (a cumulative value of 1.0 indicates the fatigue life limit has been reached).

[0138] Plastic strain is calculated by combining the corrosion rate and crack evolution eigenvectors from the corrosion degradation eigenvector. The calculation of plastic strain requires consideration of the stress-strain relationship of the pipeline material and the effects of corrosion and cracking on the pipeline's load-bearing capacity. For the oil and gas pipeline in this example, the calculated plastic strain is 0.027.

[0139] The physical life consumption rate is calculated by weighting the cumulative fatigue damage, corrosion rate, and plastic strain. Weighting coefficients can be set based on the importance of different failure mechanisms to the pipeline life. In this example, the cumulative fatigue damage weight is set to 0.4, the corrosion rate weight is set to 0.35, and the plastic strain weight is set to 0.25. The calculated physical life consumption rate is 0.068 / year, indicating that the pipeline life is consumed by approximately 6.8% each year.

[0140] The physical life consumption rate is input into the previously constructed degradation state evolution equation, and a data-driven life prediction is obtained through iterative calculation. This prediction is based on historical monitoring data and the established stochastic process model, accounting for the randomness and uncertainty of the system. For the oil and gas pipeline in this example, the data-driven life prediction obtained through iterative calculation is 12.3 years.

[0141] The weight coefficients for the physical life consumption rate and the data-driven life prediction value are adaptively adjusted based on the prediction error. During the weight adjustment process, a sliding window method can be used to evaluate historical prediction errors and dynamically adjust the weight coefficients based on the magnitude of the error. When the physical model prediction error is small, the weight of the physical life consumption rate is increased; when the data-driven model prediction error is small, the weight of the data-driven life prediction value is increased. For the oil and gas pipeline in the example, assuming the average relative error of the physical model in historical predictions is 7% and the average relative error of the data-driven model is 5%, the weight of the physical life consumption rate can be set to 0.42, and the weight of the data-driven life prediction value can be set to 0.58.

[0142] By weighting the physical life consumption rate and the data-driven life prediction, the remaining life prediction for the oil and gas pipeline is output. For the oil and gas pipeline in this example, the calculated remaining life prediction is 11.8 years. This prediction, which comprehensively considers physical failure mechanisms and statistical data characteristics, can provide a reliable basis for maintenance decisions for oil and gas pipelines.

[0143] In practical applications, pipeline monitoring data can be updated regularly (e.g., every six months) and the above prediction process can be re-executed to obtain more accurate RLL predictions. Furthermore, these predictions can be compared and verified with regular inspection data to further optimize model parameters and prediction strategies. This continuous iterative approach can continuously improve the accuracy and reliability of RLL predictions for oil and gas pipelines.

[0144] Figure 4 This is a schematic diagram of the relationship between fatigue damage accumulation and prediction error according to an embodiment of the present invention:

[0145] This figure shows a comparison of the prediction error rates of three different models (this technical solution, traditional fatigue model and Miner linear cumulative model) under different fatigue damage accumulation amounts. The horizontal axis represents the fatigue damage accumulation amount, ranging from 0.1 to 0.8; the vertical axis represents the prediction error rate (%). It can be seen from the figure that with the increase of fatigue damage accumulation amount, the prediction errors of the three models all show an upward trend, but the prediction error of this technical solution increases the most slowly. Specifically, when the fatigue damage accumulation amount is 0.1, the prediction errors of the three models are 2.3%, 3.8% and 3.1% respectively; when the accumulation amount increases to 0.4, the errors rise to 3.6%, 6.7% and 5.5% respectively; when the accumulation amount reaches 0.8, the errors further increase to 5.2%, 10.2% and 8.7%. Overall, the prediction error of this technical solution in each damage accumulation stage is significantly lower than that of the other two models. The average error rate is about 50% lower than that of the traditional fatigue model and about 40% lower than that of the Miner linear accumulation model, indicating that this solution has better prediction accuracy and stability, especially in the high fatigue damage accumulation range. The prediction performance advantage is more obvious.

[0146] In an optional embodiment, a Wiener random process model is established based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, and the drift coefficient, diffusion coefficient and measurement noise are combined to construct a degradation state evolution equation; introducing a nonlinear compensation term into the degradation state evolution equation includes:

[0147] Obtaining degradation state observation values ​​at multiple time points, wherein the degradation state observation values ​​include state measurement data and timestamp information;

[0148] Calculating the difference between state observation values ​​at adjacent time points and dividing the difference by the corresponding time interval to obtain a degradation rate; processing the degradation rate using a maximum likelihood estimation method to obtain a drift coefficient; performing a square sum operation on the deviation between the degradation rate and the drift coefficient to obtain a diffusion coefficient;

[0149] Performing a time domain analysis on the measurement noise in the degradation state observation value, calculating the autocorrelation function of the measurement noise under different time delays; performing a Fourier transform on the autocorrelation function to obtain a frequency domain characteristic;

[0150] Based on the frequency domain characteristics, the state is predicted, and the difference between the predicted value and the observed value is dynamically corrected in combination with the Kalman gain to eliminate the influence of measurement noise to obtain the state estimation value;

[0151] Performing local linearization processing on the state estimate, calculating a nonlinear approximation term through a second-order Taylor expansion; constructing a dynamic compensation term based on the nonlinear approximation term combined with a historical state data sequence and a historical prediction error;

[0152] The drift coefficient, the diffusion coefficient, the state estimation value and the dynamic compensation term are combined to construct a degradation state evolution equation, wherein the degradation state evolution equation includes a Brownian motion term, a filtered noise term and a nonlinear compensation term, and is used to predict the degradation state at a future moment.

[0153] Obtain crack evolution feature vectors and corrosion degradation feature vectors that reflect the health status of the structure. These feature vectors are obtained by preprocessing and feature extraction of the raw data collected by the sensor network. For example, for a certain steel structure component, the feature vector composed of crack length, width, depth and other parameters [12.5mm, 0.8mm, 3.2mm] can be obtained through the ultrasonic detection system, and the feature vector composed of corrosion potential, corrosion current density and other parameters [-650mV, 25μA / cm 2 ].

[0154] Based on the acquired feature vectors, the system obtains degradation state observations at multiple time points. In practical applications, data can be collected at preset time intervals, such as every 24 hours. For example, in a metal component monitoring system, the system recorded 30 consecutive days of state measurement data, forming a dataset containing degradation state values ​​and corresponding timestamps. Specific data examples include: state value 0.05 on day 1, 0.12 on day 5, 0.25 on day 10, 0.41 on day 15, 0.63 on day 20, 0.89 on day 25, and 1.22 on day 30.

[0155] The degradation rate is calculated by calculating the difference between the state observations at adjacent time points and dividing the difference by the corresponding time interval. Taking the above data as an example, from day 1 to day 5, the state difference is 0.07, and the time interval is 4 days, so the degradation rate is 0.0175 / day; from day 5 to day 10, the degradation rate is 0.026 / day; from day 10 to day 15, the degradation rate is 0.032 / day; and so on, the complete degradation rate sequence is [0.0175, 0.026, 0.032, 0.044, 0.052, 0.066].

[0156] The degradation rate data was processed using the maximum likelihood estimation method, and the drift coefficient was calculated through iterative calculation. In this embodiment, the drift coefficient calculated by processing the degradation rate series was 0.039 / day. This coefficient reflects the average change trend of the degradation process.

[0157] To calculate the diffusion coefficient, the system squares the difference between the degradation rate and the drift coefficient. Using the above data, the square of the difference between each degradation rate and the drift coefficient of 0.039 is calculated to obtain [(0.0175-0.039) 2 ,(0.026-0.039)2 ,(0.032-0.039) 2 ,(0.044-0.039) 2 ,(0.052-0.039) 2 ,(0.066-0.039) 2 ], summed and divided by the number of data points minus one, to obtain a diffusion coefficient of 0.0003 / day.

[0158] The system performs time-domain analysis of the measurement noise in the degradation state observations, calculating the autocorrelation function of the measurement noise at different time delays. Taking 100 noise samples collected in a particular experiment as an example, the system extracts trends using the moving average method to obtain a noise series. The autocorrelation coefficients for time delays of 1, 2, 3, 4, and 5 days are 0.82, 0.65, 0.47, 0.31, and 0.18, respectively.

[0159] A Fourier transform of the autocorrelation function was performed to obtain the frequency domain characteristics of the noise. This transformation revealed that the noise was primarily concentrated in the low-frequency band, with energy primarily distributed within the 0-0.2Hz frequency range and relatively low energy in the high-frequency components. Based on this frequency domain characteristic, a low-pass filter was designed with a cutoff frequency of 0.25Hz and a roll-off rate of 24dB / octave.

[0160] Based on these frequency-domain features, the state is predicted and the difference between the predicted and observed values ​​is dynamically corrected using the Kalman gain. In one test case, the predicted state value on day 31 was 1.32, while the actual observed value was 1.29, a difference of 0.03. Using the calculated Kalman gain of 0.65, the predicted value was corrected to a state estimate of 1.297, effectively eliminating the effects of measurement noise.

[0161] The state estimate is locally linearized, and the nonlinear approximation is calculated using a second-order Taylor expansion. Five days of historical data are selected as the reference window, and the calculated second-order differential term is 0.0032, which represents the dynamic characteristics of the nonlinear system.

[0162] A dynamic compensation term is constructed based on the nonlinear approximation term, combined with the historical state data series and historical forecast errors. The average forecast error over the past 10 days is 0.018. Combining the second-order approximation term with the current state value, the calculated dynamic compensation term is 0.0076.

[0163] A degradation state evolution equation is constructed by combining a drift coefficient of 0.039 / day, a diffusion coefficient of 0.0003 / day, a state estimate of 1.297, and a dynamic compensation term of 0.0076. This equation, which includes a Brownian motion term, a filtered noise term, and a nonlinear compensation term, can more accurately predict the degradation state at future moments. For example, this equation predicts a degradation state value of 1.78 on day 40. Compared to the actual observed value of 1.81, the prediction error is only 1.7%, significantly better than the 9.3% prediction error of the traditional linear model.

[0164] In an optional embodiment, generating a pipeline safety status assessment report and pipeline maintenance decision recommendations based on the optimized safety status assessment accuracy includes:

[0165] The optimized safety status assessment accuracy includes the assessment indicator accuracy and the assessment model confidence; a correction coefficient is calculated based on the assessment model confidence, and the initial weight of the assessment indicator is linearly corrected based on the correction coefficient to obtain a dynamic correction weight; the assessment indicator and the dynamic correction weight are weighted and summed to obtain a comprehensive safety status score;

[0166] Performing K-means cluster analysis on historical assessment data to obtain initial risk level classification criteria; using the assessment model confidence as an adaptive factor, updating the initial classification criteria using an exponential sliding weighted average method; matching the comprehensive safety status score with the updated classification criteria to determine the risk level classification criteria;

[0167] A maintenance priority assessment model is constructed based on the risk level classification standard, and the risk level, maintenance difficulty, and time urgency are quantified and input into the maintenance priority assessment model; the assessment factor weights are set using the assessment indicator accuracy, and the maintenance priority score is calculated by weighted summation;

[0168] Writing the comprehensive safety status score, the risk level classification standard, and the maintenance priority score into a preset assessment report template in order of importance to generate a pipeline safety status assessment report;

[0169] A word extraction algorithm is used to identify assessment information from the pipeline safety status assessment report, and the assessment information is matched with a maintenance decision rule library to generate a pipeline maintenance decision suggestion including maintenance content, maintenance sequence and resource allocation.

[0170] The optimized safety status assessment accuracy consists of two key parameters: the accuracy of the assessment indicator and the confidence of the assessment model. In practical applications, the accuracy of the assessment indicator is usually expressed as a percentage, such as 92.3% for corrosion detection and 88.7% for structural strength assessment. The confidence of the assessment model reflects the degree of confidence in its own prediction results, such as 0.85 for the corrosion prediction model and 0.78 for the structural safety model.

[0171] A nonlinear mapping function is used to calculate correction coefficients based on the confidence level of the assessment model. For example, when the confidence level of the pipeline corrosion detection model is 0.85, the correction coefficient obtained through the preset mapping relationship is 1.12; when the confidence level of the structural safety model is 0.78, the corresponding correction coefficient is 0.95. These correction coefficients are used to adjust the initial weights of the assessment indicators. Assuming the initial weight of the corrosion indicator is 0.3 and the initial weight of the structural strength indicator is 0.25, the dynamically corrected weights are 0.336 (0.3 × 1.12) and 0.2375 (0.25 × 0.95), respectively.

[0172] To calculate the comprehensive safety status score, each evaluation indicator is multiplied by its corresponding dynamic correction weight and the resulting sum is calculated. For example, if a pipeline section has a corrosion score of 78, a structural strength score of 85, and an environmental risk score of 65, with corresponding dynamic correction weights of 0.336, 0.2375, and 0.22, respectively, the overall safety status score is 75.86.

[0173] When performing K-means cluster analysis on historical assessment data, 8,000 pipeline assessment records from three years were selected as the sample data. The number of clusters, K, was set to 5, representing five risk levels: extremely high risk, high risk, medium risk, low risk, and extremely low risk. The cluster analysis yielded the following initial classification criteria: extremely high risk (0-60 points), high risk (60-70 points), medium risk (70-80 points), low risk (80-90 points), and extremely low risk (90-100 points).

[0174] When using the assessment model confidence as an adaptive factor to update the initial classification criteria, an exponential sliding weighted average method was used to update the initial classification criteria when the overall model confidence reached 0.82. The calculation process involves multiplying the initial threshold by a weighting factor, which is then exponentially adjusted based on the confidence level. The updated risk classification criteria are: extremely high risk (0-58 points), high risk (58-69 points), medium risk (69-81 points), low risk (81-91 points), and very low risk (91-100 points).

[0175] Match the comprehensive safety status score with the updated classification criteria. For example, when the comprehensive score of a pipeline section is 75.86 points, the corresponding risk level is "medium risk".

[0176] When constructing a maintenance priority assessment model based on risk grading criteria, three key factors are quantified: risk severity, maintenance difficulty, and time urgency. Risk severity is directly determined by the risk level; for example, a medium risk score is 60. Maintenance difficulty considers human resource needs, equipment requirements, and environmental constraints; for example, a pipeline section's maintenance difficulty score is 75. Time urgency is based on the fault's development rate and impact range; for example, a pipeline section's time urgency score is 80.

[0177] When using the accuracy of the evaluation indicator to set the weight of the evaluation factor, the higher the accuracy, the larger the corresponding weight. For example, if the accuracy of the risk assessment is 92%, the accuracy of the maintenance difficulty assessment is 85%, and the accuracy of the time urgency assessment is 88%, the corresponding weights are 0.35, 0.3, and 0.35, respectively. The weighted sum calculation results in a maintenance priority score of 71.25.

[0178] When generating the pipeline safety status assessment report, the comprehensive safety status score (75.86 points), risk classification (medium risk), and maintenance priority score (71.25 points) are entered into a pre-set assessment report template in order of importance. The report template consists of four sections: basic information, assessment results, risk analysis, and recommendations. The generated report adopts a structured format to facilitate subsequent analysis and decision-making.

[0179] When using a word extraction algorithm to identify assessment information from the assessment report, a rule-based word extraction method was employed to identify key information such as "medium risk," "corrosion index 78 points," and "maintenance priority 71.25 points." This assessment information was then matched against a maintenance decision rule base containing maintenance strategies tailored to different risk levels and fault types. For example, for a pipeline section with a "medium risk" and a low "corrosion index," the matching maintenance decision would be "repair the anti-corrosion coating within three months, deploy a standard anti-corrosion tool set, and employ three specialized technicians."

[0180] The final pipeline maintenance decision recommendation includes maintenance content (anti-corrosion coating repair), maintenance sequence (within three months) and resource allocation (standard anti-corrosion tool set and 3 professional technicians), providing intuitive and clear decision support for the pipeline operation management department.

[0181] Figure 5 This is a bar chart comparing the performance of the pipeline safety status assessment and maintenance decision-making method according to an embodiment of the present invention:

[0182] The figure shows the comparative results of three different methods (this method, a traditional evaluation method, and a static weighting method) across five key performance indicators. This method demonstrates significant advantages across all evaluation indicators: Its evaluation accuracy reached 92.7%, 14.4 percentage points higher than the traditional method (78.3%) and the static weighting method (84.5%), respectively. Its decision rationality reached 89.5%, compared to 72.6% for the traditional method and 79.2% for the static weighting method. Its risk identification rate reached the highest level at 94.3%, compared to 81.5% for the traditional method and 87.1% for the static weighting method. Its maintenance priority accuracy reached 91.8%, significantly higher than the 75.2% for the traditional method and 81.3% for the static weighting method. Its resource allocation efficiency remained high at 88.6%, compared to 70.8% for the traditional method and 76.9% for the static weighting method. Overall, this method achieved an average improvement of 15-20 percentage points across all indicators, with the most significant gains in two key indicators: risk identification rate and maintenance priority accuracy. This fully demonstrates the method's advanced nature and practical value in supporting equipment maintenance decisions. This method not only improves assessment accuracy but also significantly enhances the scientific nature of decision-making and resource efficiency.

[0183] A second aspect of an embodiment of the present invention provides an intelligent monitoring system for the safety status of an oil and gas pipeline throughout its entire life cycle, comprising:

[0184] The first unit is used to collect oil and gas pipeline operation status data, wherein the oil and gas pipeline operation status data includes pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data;

[0185] The second unit is configured to establish a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; and construct a multi-dimensional feature vector based on the digital twin model of the oil and gas pipeline, wherein the multi-dimensional feature vector includes a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector;

[0186] The third unit is configured to perform feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; and construct an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network;

[0187] The fourth unit is used to calculate the remaining service life prediction value of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model;

[0188] The fifth unit is used to continuously optimize the safety status assessment accuracy through online iterative training based on the remaining life prediction value of the oil and gas pipeline and the pipeline safety status assessment rule base; and generate an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations based on the optimized safety status assessment accuracy.

[0189] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0190] processor;

[0191] a memory for storing processor-executable instructions;

[0192] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0193] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0194] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0195] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent monitoring method for the safety status of oil and gas pipelines throughout their entire life cycle, characterized in that: include: Collecting oil and gas pipeline operation status data, the oil and gas pipeline operation status data including pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data; Establishing a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; constructing a multi-dimensional feature vector based on the oil and gas pipeline digital twin model, the multi-dimensional feature vector including a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector; Performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network; Calculating a predicted value of the remaining service life of the oil and gas pipeline based on the output result of the oil and gas pipeline fault evolution prediction model; Based on the predicted value of the remaining life of the oil and gas pipeline and in combination with the pipeline safety status assessment rule base, the accuracy of the safety status assessment is continuously optimized through online iterative training; Based on the optimized safety status assessment accuracy, an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations are generated.

2. The method according to claim 1, characterized in that Establishing a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; Based on the oil and gas pipeline digital twin model, constructing a multi-dimensional feature vector includes: constructing a physical-virtual mapping matrix based on the oil and gas pipeline operation status data, wherein the physical-virtual mapping matrix includes a plurality of mapping relationship coefficients; Establishing a digital twin model of the oil and gas pipeline according to the physical-virtual mapping matrix to realize virtual simulation of the operating status data of the oil and gas pipeline; Obtaining input variables and model parameters, constructing a state update equation based on the input variables and the model parameters, and dynamically updating the oil and gas pipeline digital twin model through the state update equation; Extracting multi-dimensional features based on the oil and gas pipeline digital twin model includes: calculating pressure fluctuation features based on the variance of pressure sampling values ​​and average pressure, calculating temperature gradient features based on the partial derivative of the temperature field in three-dimensional space, and calculating environmental impact factors based on weight coefficients of various environmental parameters, wherein the weight coefficients are determined through historical data analysis; The pressure fluctuation characteristics, the temperature gradient characteristics and the environmental influencing factors are fused to construct a multi-dimensional feature vector of the oil and gas pipeline.

3. The method according to claim 1, characterized in that Performing feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; constructing an oil and gas pipeline fault evolution prediction model based on the fused feature vector and combining it with a graph neural network includes: The multidimensional feature vector is mapped into a state feature using a state mapping weight matrix; a time feature is constructed based on a timestamp, a period code, and a feature change rate; a spatial feature is constructed based on a position coordinate, an azimuth angle, and a feature space gradient; and the state feature, the time feature, and the space feature are combined to form a spatiotemporal knowledge node; Extracting graph features based on the spatiotemporal knowledge nodes through node feature aggregation operations, wherein the node feature aggregation operations use attention weights to perform weighted combination of adjacent node features, wherein the attention weights are calculated through spatiotemporal attention; Using a graph neural network to process the graph features to obtain enhanced features, and combining the enhanced features with the multidimensional feature vector to obtain a fused feature vector; Performing causal chain discovery on the fused feature vector, including: calculating the causal relationship between features through a conditional independence test, obtaining causal strength based on a causal gradient; and constructing a causal attention mechanism based on the causal relationship and the causal strength; The causal attention mechanism is combined with the fusion feature vector, and an oil and gas pipeline fault evolution prediction model is constructed through the graph neural network.

4. The method according to claim 3, characterized in that The causal relationship between features is calculated through conditional independence testing, and the causal strength is obtained based on the causal gradient. The causal attention mechanism is constructed based on the causal relationship and the causal strength, including: Acquire time series feature data and state feature data, calculate conditional entropy of the time series feature data and the state feature data, perform a conditional independence test based on the conditional entropy, and obtain a conditional independence test result between features; Performing conditional probability modeling on the time series feature data, calculating the probabilistic dependency between features using a Bayesian network, and determining the causal relationship between features based on the result of the conditional independence test and the probabilistic dependency; Calculating the causal gradient between features based on the causal relationship, and obtaining the direct causal strength according to the causal gradient; calculating the indirect causal strength between features along the causal transmission path; and combining the direct causal strength and the indirect causal strength to obtain the comprehensive causal strength; The comprehensive causal strength is combined with the feature representation to generate a causal query vector and a causal key-value vector; a causal attention matrix is ​​constructed based on the causal query vector and the causal key-value vector; and the feature representation is reconstructed according to the causal attention matrix to obtain a causal attention mechanism.

5. The method according to claim 1, wherein Calculating the predicted value of the remaining service life of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model includes: Decoupling the output results of the oil and gas pipeline fault evolution prediction model according to the failure mechanism to construct a crack evolution feature vector and a corrosion degradation feature vector; Establishing a Wiener random process model based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, combining the drift coefficient, the diffusion coefficient and the measurement noise to construct a degradation state evolution equation; introducing a nonlinear compensation term into the degradation state evolution equation; Calculating fatigue damage accumulation based on the Wiener random process model, and calculating plastic strain in combination with the corrosion rate in the corrosion degradation characteristic vector and the crack evolution characteristic vector; performing a weighted combination of the fatigue damage accumulation, the corrosion rate, and the plastic strain to obtain a physical life consumption rate; The physical life consumption rate is input into the degradation state evolution equation to obtain a data-driven life prediction value; the weight coefficients of the physical life consumption rate and the data-driven life prediction value are adaptively adjusted according to the prediction error to output the remaining life prediction value of the oil and gas pipeline.

6. The method according to claim 5, characterized in that Establishing a Wiener random process model based on the crack evolution characteristic vector and the corrosion degradation characteristic vector, and constructing a degradation state evolution equation by combining the drift coefficient, the diffusion coefficient and the measurement noise; The nonlinear compensation term introduced into the degradation state evolution equation includes: Obtaining degradation state observation values ​​at multiple time points, wherein the degradation state observation values ​​include state measurement data and timestamp information; Calculating the difference between state observation values ​​at adjacent time points and dividing the difference by the corresponding time interval to obtain a degradation rate; processing the degradation rate using a maximum likelihood estimation method to obtain a drift coefficient; performing a square sum operation on the deviation between the degradation rate and the drift coefficient to obtain a diffusion coefficient; Performing a time domain analysis on the measurement noise in the degradation state observation value, calculating the autocorrelation function of the measurement noise under different time delays; performing a Fourier transform on the autocorrelation function to obtain a frequency domain characteristic; Based on the frequency domain characteristics, the state is predicted, and the difference between the predicted value and the observed value is dynamically corrected in combination with the Kalman gain to eliminate the influence of measurement noise to obtain the state estimation value; Performing local linearization processing on the state estimate, calculating a nonlinear approximation term through a second-order Taylor expansion; constructing a dynamic compensation term based on the nonlinear approximation term combined with a historical state data sequence and a historical prediction error; The drift coefficient, the diffusion coefficient, the state estimation value and the dynamic compensation term are combined to construct a degradation state evolution equation, wherein the degradation state evolution equation includes a Brownian motion term, a filtered noise term and a nonlinear compensation term, and is used to predict the degradation state at a future moment.

7. The method according to claim 1, characterized in that Based on the optimized safety status assessment accuracy, a pipeline safety status assessment report and pipeline maintenance decision recommendations are generated, including: The optimized safety status assessment accuracy includes the assessment indicator accuracy and the assessment model confidence; a correction coefficient is calculated based on the assessment model confidence, and the initial weight of the assessment indicator is linearly corrected based on the correction coefficient to obtain a dynamic correction weight; the assessment indicator and the dynamic correction weight are weighted and summed to obtain a comprehensive safety status score; Performing K-means cluster analysis on historical assessment data to obtain initial risk level classification criteria; using the assessment model confidence as an adaptive factor, updating the initial classification criteria using an exponential sliding weighted average method; matching the comprehensive safety status score with the updated classification criteria to determine the risk level classification criteria; A maintenance priority assessment model is constructed based on the risk level classification standard, and the risk level, maintenance difficulty, and time urgency are quantified and input into the maintenance priority assessment model; the assessment factor weights are set using the assessment indicator accuracy, and the maintenance priority score is calculated by weighted summation; Writing the comprehensive safety status score, the risk level classification standard, and the maintenance priority score into a preset assessment report template in order of importance to generate a pipeline safety status assessment report; A word extraction algorithm is used to identify assessment information from the pipeline safety status assessment report, and the assessment information is matched with a maintenance decision rule library to generate a pipeline maintenance decision suggestion including maintenance content, maintenance sequence and resource allocation.

8. An intelligent monitoring system for the safety status of oil and gas pipelines throughout their life cycle, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to collect oil and gas pipeline operation status data, wherein the oil and gas pipeline operation status data includes pipeline internal pressure data, pipeline internal temperature data and pipeline surrounding environment data; The second unit is configured to establish a digital twin model of the oil and gas pipeline based on the oil and gas pipeline operation status data; and construct a multi-dimensional feature vector based on the digital twin model of the oil and gas pipeline, wherein the multi-dimensional feature vector includes a pipeline structure feature vector, a pipeline environment feature vector, and a pipeline historical fault feature vector; The third unit is configured to perform feature fusion on the multi-dimensional feature vector to obtain a fused feature vector; and construct an oil and gas pipeline fault evolution prediction model based on the fused feature vector and a graph neural network; The fourth unit is used to calculate the remaining service life prediction value of the oil and gas pipeline according to the output result of the oil and gas pipeline fault evolution prediction model; A fifth unit is configured to continuously optimize the accuracy of safety status assessment through online iterative training based on the predicted value of the remaining life of the oil and gas pipeline and in combination with a pipeline safety status assessment rule base; Based on the optimized safety status assessment accuracy, an oil and gas pipeline safety status assessment report and pipeline maintenance decision recommendations are generated.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images