Method for evaluating operating condition vulnerability of oil tube, identification method and device

Abstract

This invention discloses a vulnerability assessment method, identification method, and apparatus for oil pipeline operating conditions. The vulnerability assessment method includes: segmenting target operating condition data to obtain multiple multidimensional time series; performing correlation analysis and feature extraction on the multiple multidimensional time series to obtain corresponding adjacency matrices and input features; combining the adjacency matrix and input features of each multidimensional time series to obtain subgraphs of each time series; using each subgraph as a node, calculating the conditional probability between any two adjacent nodes to obtain the weight of the edge between any two adjacent nodes; obtaining a target network model based on the nodes and the weights of the edges between any two adjacent nodes; calculating the evaluation index of each node based on the target network model and the node data; and selecting the operating conditions with the highest vulnerability based on the evaluation index of each node. This method can accurately identify vulnerable operating conditions of oil pipelines.

Description

Technical Field

[0001] This invention relates to the fields of complex pipeline systems and intelligent operating condition identification, and particularly to a vulnerability assessment method, identification method, and apparatus for pipeline operating conditions. Background Technology

[0002] Refined oil pipelines serve as a bridge connecting upstream refineries and downstream markets, responsible for transporting 70% of refined products such as gasoline, diesel, and kerosene from refineries to the market. Along the pipelines are input stations, receiving stations, oil depots, and pumping stations. Therefore, refined oil pipelines transport a wide variety of oil products, have a broad monitoring scope, and require tracking of numerous oil product interfaces. As of the end of 2022, the total length of refined oil pipelines in my country was approximately 32,000 kilometers. As the primary mode of refined oil transportation, the rapid development of refined oil pipelines will bring unprecedented challenges to future pipeline management. Therefore, vulnerability assessments of various operating conditions of refined oil pipelines are extremely important, enabling rapid identification of the most vulnerable operating conditions and helping on-site personnel make quick judgments and decisions regarding events posing potential safety hazards. Summary of the Invention

[0003] In order to accurately and quickly identify vulnerable operating conditions in refined oil pipeline systems, and thus help on-site personnel make rapid judgments and decisions on events with potential safety hazards, this invention provides a vulnerability assessment method, identification method, and device for the operating conditions of oil pipelines.

[0004] In a first aspect, embodiments of the present invention provide a vulnerability assessment method for the operating conditions of oil pipelines, comprising:

[0005] Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, wherein each multidimensional time series corresponds to a certain operating condition.

[0006] Correlation analysis and feature extraction are performed on the multiple multidimensional time series respectively to obtain the adjacency matrix and input features corresponding to each multidimensional time series;

[0007] By combining the adjacency matrix and the input features corresponding to each of the multidimensional time series, each time series subgraph is obtained;

[0008] Using each of the aforementioned time series subgraphs as nodes, calculate the conditional probability between any two adjacent nodes to obtain the weight of the edge between any two adjacent nodes;

[0009] Based on the weights of the nodes and the edges between any two adjacent nodes, the target network model is obtained;

[0010] Based on the target network model and the data of the nodes, the evaluation index of each node is calculated.

[0011] Based on the evaluation metrics of each node, the operating conditions with the highest vulnerability are selected.

[0012] Optionally, the process of selecting the most vulnerable operating conditions based on the evaluation metrics of each node includes:

[0013] Based on the evaluation indicators of each node, the operating conditions are sorted according to the value of the evaluation indicators, and the operating conditions that occur most frequently and are ranked first are selected as the operating conditions with the highest vulnerability.

[0014] Optionally, the evaluation metrics include at least degree centrality, eigenvector centrality, clustering coefficient, tight centrality, and betweenness centrality;

[0015] The evaluation metrics for each node are calculated based on the target network model and the node data, including:

[0016] For each node, the degree centrality is calculated based on the node's adjacency matrix.

[0017] Based on the eigenvalues ​​of the adjacency matrix of the node and the adjacency matrix of the node, the eigenvector centrality is calculated.

[0018] Based on the target network model, the number of triangles formed by the nodes is determined, and the clustering coefficient is calculated based on the number of triangles formed by the nodes and the degree centrality.

[0019] Based on the target network model, the compact centrality is calculated.

[0020] Based on the target network model, the betweenness centrality is calculated.

[0021] Optionally, for each node, the degree centrality is calculated based on the node's adjacency matrix according to the following formula:

[0022]

[0023] Among them, D i Let A represent the degree centrality of node i, N represent the total number of nodes, and A''''''''''''''''''''''''''''''''''''''''''''''""","' ... i Let i represent the adjacency matrix of node i.

[0024] Optionally, calculating the eigenvector centrality based on the eigenvalues ​​of the obtained adjacency matrix of the node and the adjacency matrix of the node includes:

[0025] Based on the adjacency matrix of the nodes, the eigenvalues ​​of the adjacency matrix are calculated;

[0026] Based on the eigenvalues ​​of the adjacency matrix and the adjacency matrix itself, the eigenvector centrality is calculated according to the following formula:

[0027]

[0028] Where k is the eigenvalue of the adjacency matrix; This indicates the centrality of the eigenvectors.

[0029] Optionally, based on the target network model, the number of triangles formed by the nodes is determined, and the clustering coefficient is calculated based on the number of triangles formed by the nodes and the degree centrality, according to the following formula:

[0030]

[0031] Among them, C i The degree centrality of node i is represented by t. i D represents the number of triangles that node i can form; i C represents the degree centrality of node i; when node i is not connected to any of its neighbors, C is... i =0; C = 0 when node i is connected to all its neighboring nodes. i =1.

[0032] Optionally, calculating the compact centrality based on the target network model includes:

[0033] For each node, based on the target network model, the number of nodes traversed and the total number of edges on the path from that node to all other nodes are calculated, and the compact centrality is calculated according to the following formula:

[0034] Cl i =n iv / ∑ v d iv

[0035] Among them, Cl i Represents the compact centrality of node i; n iv d represents the number of nodes traversed on the path from node i to node v; iv This represents the total number of edges traversed on the path from node i to node v.

[0036] Optionally, calculating the betweenness centrality based on the target network model includes:

[0037] Based on the target network model, for any two remaining nodes excluding the stated node, the number of all shortest paths between the two remaining nodes is calculated, and the number of shortest paths passing through the stated node is also calculated. The betweenness centrality is then calculated using the following formula:

[0038]

[0039] Among them, B i The degree of centrality of node i is represented by l. st This represents the number of all shortest paths between the remaining node s and the remaining node t; This represents the number of shortest paths that pass through node i.

[0040] Optionally, the step of performing correlation analysis and feature extraction based on the multiple multidimensional time series to obtain the adjacency matrix and features corresponding to each multidimensional time series includes:

[0041] Based on the multiple multidimensional time series, the weights between adjacent multidimensional time series are calculated using the mutual information method;

[0042] Obtain all weights to form the adjacency matrix;

[0043] For each of the multiple multidimensional time series, a preset LSTM model is used to extract features to obtain the input features corresponding to each multidimensional time series.

[0044] Secondly, embodiments of the present invention provide a method for identifying the operating conditions of oil pipelines, including:

[0045] Obtain the actual operating data of the target pipeline;

[0046] The actual operating data is input into the operating condition identification model to obtain the operating condition identification result of the target pipeline;

[0047] The method for constructing the working condition identification model includes:

[0048] Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, wherein each multidimensional time series corresponds to a certain operating condition.

[0049] Correlation analysis and feature extraction are performed on the multiple multidimensional time series respectively to obtain the adjacency matrix and input features corresponding to each multidimensional time series;

[0050] By combining the adjacency matrix and the input features corresponding to each of the multidimensional time series, each time series subgraph is obtained;

[0051] Each of the aforementioned time series subgraphs is used as a training set to train a preset graph neural network model, and the trained graph neural network model is used as the working condition recognition model.

[0052] Thirdly, embodiments of the present invention provide a vulnerability assessment device for the operating conditions of oil pipelines, comprising:

[0053] The first data processing module is used to segment the acquired target operating condition data using a time series segmentation method to obtain multiple multi-dimensional time series, wherein each multi-dimensional time series corresponds to an operating condition.

[0054] The second data processing module is used to perform correlation analysis and feature extraction based on the multiple multidimensional time series respectively, to obtain the adjacency matrix and input features corresponding to each multidimensional time series;

[0055] The image output module is used to combine the adjacency matrix corresponding to each of the multidimensional time series and the input features to obtain each time series sub-image;

[0056] The calculation module is used to take each of the time series subgraphs as nodes, calculate the conditional probability between any two adjacent nodes, and obtain the weight of the edge between any two adjacent nodes.

[0057] The network model generation module is used to obtain the target network model based on the weights of the edges between the node and any two adjacent nodes.

[0058] The data calculation module is used to calculate the evaluation index of each node based on the target network model and the data of the nodes;

[0059] The filtering module is used to filter out the operating conditions with the highest vulnerability based on the evaluation indicators of each node.

[0060] Fourthly, embodiments of the present invention provide an operating condition identification device for oil pipelines, comprising:

[0061] The data acquisition module is used to acquire the actual operating data of the target pipeline;

[0062] The data input module is used to input the actual operating data into the operating condition identification model to obtain the operating condition identification result of the target pipeline.

[0063] Fifthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, characterized in that, when executed by a processor, the program implements the vulnerability assessment method for operating conditions of oil pipelines as described in the first aspect, and / or the identification method for operating conditions of oil pipelines as described in the second aspect.

[0064] In a sixth aspect, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the vulnerability assessment method for operating conditions of oil pipelines as described in the first aspect, and / or the identification method for operating conditions of oil pipelines as described in the second aspect.

[0065] In a seventh aspect, embodiments of the present invention provide a computer program product containing instructions that, when the computer program product is run on a computer device, cause the computer device to execute the vulnerability assessment method for operating conditions of oil pipelines as described in the first aspect, and / or the identification method for operating conditions of oil pipelines as described in the second aspect.

[0066] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following:

[0067] This invention provides a vulnerability assessment method for the operating conditions of oil pipelines. By segmenting the target operating condition data, a multidimensional time series is obtained. Correlation analysis and feature extraction are performed on the multidimensional time series to obtain the corresponding adjacency matrix and input features. A time series subgraph is obtained based on the adjacency matrix and input features. Based on the time series subgraph, the weights of edges between adjacent nodes are calculated to obtain the target network model. Evaluation indicators for each node are calculated, and the trend of these indicators over time is monitored. This method can identify vulnerable operating conditions in refined oil pipeline systems, thereby helping on-site personnel make rapid judgments and decisions regarding events with potential safety hazards. It features high timeliness and accuracy.

[0068] From the perspective of graph theory, a relational network based on multiple operating conditions of pipelines, namely the target network model, is constructed for the first time. Using complex network indicators, an evaluation method for vulnerable operating conditions of pipelines is proposed, realizing the identification of pipeline vulnerable operating conditions. This advances the basic research on pipeline operating condition vulnerability identification and can provide decision support for managers in pipeline system operation and maintenance management. It has high application value and can be extended to oil pipelines and gas pipelines.

[0069] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0070] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0071] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0072] Figure 1 This is a flowchart of a vulnerability assessment method for the operating conditions of oil pipelines provided in an embodiment of the present invention;

[0073] Figure 2 This is a flowchart of constructing a time series subgraph (i.e., a dynamic graph network model) provided in an embodiment of the present invention;

[0074] Figure 3 This is a flowchart of constructing a target network model provided in an embodiment of the present invention;

[0075] Figure 4 This is a schematic diagram of the target oil pipe topology provided in an embodiment of the present invention;

[0076] Figure 5 This is a graph showing the calculation results of the evaluation indicators provided in this embodiment of the invention, wherein, Figure 5 (a) is a graph showing the results of the degree centrality calculation. Figure 5 (b) is a graph showing the calculation results of the clustering coefficients. Figure 5 (c) is a graph showing the results of calculating the betweenness centrality. Figure 5 (d) is a graph showing the results of the compactness centrality calculation. Figure 5 (e) is the graph showing the results of eigenvector centrality calculation;

[0077] Figure 6 This is a flowchart illustrating the training process of the graph neural network model provided in this embodiment of the invention.

[0078] Figure 7 This is a flowchart of the construction of the working condition identification model provided in the embodiments of the present invention;

[0079] Figure 8 This is an explanation model of the topology diagrams for each working condition provided in the embodiments of the present invention;

[0080] Figure 9 This is a flowchart of constructing the SPipeNet model provided in an embodiment of the present invention;

[0081] Figure 10 This is a comparison chart of the recognition results of different neural network models provided in the embodiments of the present invention, wherein, Figure 10 (a) A comparison chart of results for identifying four operating conditions: pipeline shutdown, pipeline startup, intermediate station pump start-up, and intermediate station pump shutdown. Figure 10 (b) Comparison chart of the results of identifying oil product switching (0#D-92#G, 92#G-95#G, 95#G-0#G);

[0082] Figure 11 This is a schematic diagram of a vulnerability assessment device for the operating conditions of oil pipelines provided in an embodiment of the present invention. Detailed Implementation

[0083] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0084] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," "far," "near," "front," and "rear," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0085] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0086] The inventors discovered that in the existing technology, the evaluation methods for vulnerable operating conditions of oil pipelines have not formed a complete system. Most of the data used are static historical data, which cannot observe the trend of changes in operating condition parameters over time. The evaluation indicators for vulnerable operating conditions obtained are static data and do not have the characteristics of timeliness and high accuracy.

[0087] To address the aforementioned issues, the inventors, through further research, proposed a vulnerability assessment method, identification method, and apparatus for the operating conditions of oil pipelines, in order to accurately assess the vulnerability of oil pipelines in their operating conditions.

[0088] Example 1

[0089] This embodiment proposes a vulnerability assessment method for the operating conditions of oil pipelines. (See reference...) Figure 1Specifically, this may include the following steps:

[0090] Step S1: Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, where each multidimensional time series corresponds to a certain operating condition.

[0091] Step S2: Perform correlation analysis and feature extraction on multiple multidimensional time series to obtain the adjacency matrix and input features corresponding to each multidimensional time series.

[0092] Step S3: Combine the adjacency matrix and input features corresponding to each multidimensional time series to obtain each time series subgraph.

[0093] Step S4: Treat each time series subgraph as a node, calculate the conditional probability between any two adjacent nodes, and obtain the weight of the edge between any two adjacent nodes.

[0094] Step S5: Based on the weights of the nodes and the edges between any two adjacent nodes, the target network model is obtained.

[0095] Step S6: Based on the target network model and node data, calculate the evaluation index for each node.

[0096] Step S7: Based on the evaluation indicators of each node, select the operating conditions with the highest vulnerability.

[0097] To provide a clearer explanation of the vulnerability assessment method for the operating conditions of oil pipelines described above, each step will be explained in detail below.

[0098] In step S1 above, sample data of the target pipeline (including all oil stations) within a preset time period is obtained. This operational data includes at least: ① Operating conditions: start-up and shutdown of the entire pipeline, start-up and shutdown of intermediate pump stations, and oil product switching; ② Pipeline operating parameters: inlet and outlet pressure, inlet and outlet flow rate, downstream flow rate, inlet and outlet pressure of the main pump and feed pump, and oil density.

[0099] In most cases, pipelines operate in a stable state, resulting in relatively stable data fluctuations without extreme maximum or minimum values. However, considering the varying magnitudes of pipeline operation data, normalization is necessary to limit the preprocessed data to the range [0, 1], thereby eliminating the adverse effects of excessive sample differences and obtaining the target operating condition data. The following formula (1) can be used to normalize the operation data:

[0100]

[0101] In equation (1) above, x represents the running data; x' represents the normalized value, i.e. the target running condition data; max(x) and min(x) represent the maximum and minimum values ​​in the running data, respectively.

[0102] See Figure 2 This example displays a trend chart of n target operating condition data over time. In this embodiment, autocorrelation analysis can be performed on the n target operating condition data to analyze the autocorrelation coefficients at different lag periods, obtaining the analysis results and indirectly revealing the correlation of the target operating condition data at different time intervals. During the autocorrelation analysis of the target operating condition data, it is necessary to select the largest lag order among all the target operating condition data to ensure the consistency of subsequent data segmentation results. The autocorrelation coefficient for a certain time step is:

[0103]

[0104] In equation (2) above, (x1,x2,...,x N The numbers () represent the corresponding target operating condition data; K represents the number of lag periods; and N represents the number of target operating condition data. This represents the average value of all target operating condition data.

[0105] Then, the target operating condition data is segmented using a time series segmentation method to obtain multiple multidimensional time series, which are arranged in chronological order. Here, the time series segmentation method can employ existing techniques such as sliding data window reconstruction, top-down, and bottom-up approaches. In this embodiment, the time series segmentation method uses sliding data window reconstruction, a method that directly segments the data. It generates new data series by moving a fixed-size window and calculating statistics within the window, resulting in multiple multidimensional time series.

[0106] In step S2 above, the specific process of obtaining the adjacency matrix and input features corresponding to each multidimensional time series may include the following steps:

[0107] Step S21: Based on multiple multidimensional time series, the weights between adjacent multidimensional time series are calculated using the mutual information method.

[0108] In step S21 above, the pipeline system has a wide geographical distribution, and the response times of upstream and downstream operational data differ, resulting in a certain delay between different multidimensional time series. The importance of this delay effect to the identification results is adjusted by a correlation mechanism. Therefore, see [reference needed]. Figure 2Based on multiple multidimensional time series, this study utilizes the correlation mechanism of multivariate time series and employs the mutual information method to calculate the weights of each edge in the graph network. Specifically, it calculates the weights between adjacent multidimensional time series, yielding results such as... Figure 2 The MI shown 12 MI 13 MI 14 Equal weight, MI 12 MI represents the weights between the first and second multidimensional time series. 13 MI represents the weights between the first and third multidimensional time series. 14 This represents the weight between the first and fourth multidimensional time series. If the weight between two multidimensional time series exceeds a preset threshold, then an edge exists between them; otherwise, the two multidimensional time series are not connected, and the edge between them will be removed. This method removes connections between two multidimensional time series with low dependencies, reducing the complexity of the graph and computation. The mutual information method can dynamically capture the correlation between different multidimensional time series, and can dynamically capture the spatiotemporal correlation between different oil pumping stations or different equipment. The formula for calculating the weight between two adjacent multidimensional time series is:

[0109] I(X,Y)=H(X)+H(Y)-H(X,Y))(3)

[0110] In equation (3) above, I(X,Y) represents the weight between two adjacent multidimensional time series X and Y; H(X) and H(Y) represent the entropy of multidimensional time series X and multidimensional time series Y, respectively; H(X,Y) represents the mutual entropy between multidimensional time series X and multidimensional time series Y.

[0111] Step S22: Obtain all weights and form an adjacency matrix.

[0112] In step S22 above, see [reference] Figure 2 The weights are summed to form an adjacency matrix, which is... Figure 2 The MI shown 12 MI 13 MI 14 ...MI 35 MI 45 Summarized into an adjacency matrix {W 12 W 13 W 14 ...W 35 W 45}

[0113] Step S23: For each multidimensional time series, use a preset LSTM model to extract features and obtain the input features corresponding to each multidimensional time series.

[0114] In step S23 above, the LSTM model, as an improved recurrent neural network (RNN), not only solves the problem of long-range dependencies that RNNs cannot handle, but also solves common problems in neural networks such as gradient explosion or gradient vanishing, making it very effective in processing time series data. Therefore, this embodiment uses a pre-defined LSTM model to solve the feature extraction problem of multidimensional time series, obtaining the input features corresponding to each multidimensional time series. These input features can reflect the key information and patterns of the multidimensional time series. By extracting input features through LSTM, the dynamic behavior of each multidimensional time series can be better understood. For details, please refer to... Figure 2 The input to the LSTM model is a multidimensional time series: T = {t1, t2, ..., t} n}, where n is the number of multidimensional time series, is transformed into a high-level feature vector through the LSTM model, and the output is the input features corresponding to the multidimensional time series: F={F1,F2,F3,…,F…} n}

[0115] In step S3 above, see [reference] Figure 2 By combining the adjacency matrix and input features corresponding to each multidimensional time series, a subgraph of each time series is obtained, which is the dynamic graph network model (DPipeNet). In the obtained subgraph of time series, the number of nodes is determined by the number of time series, that is, each node represents a time series. The subgraph of time series can reflect the dynamic changes of each time series (i.e., node) over time, including its state, features, and interactions with other time series. At the same time, it can reflect the global structure of the entire graph network, including the distribution of nodes, the density of edges, and the connectivity of the network.

[0116] In step S4 above, see [reference] Figure 3 Based on multiple time series subgraphs, corresponding nodes are set for each, i.e., the time series subgraphs are used as nodes. It's important to note that when the input time series subgraphs have the same graph structure, they are represented by the same node number, and the nodes are connected sequentially according to the chronological order in which each time series subgraph appears. For example... Figure 3 The node S1 shown in the diagram is connected to nodes S2, S3 and S4 in sequence.

[0117] The conditional probability between any two adjacent nodes is calculated using the following formula (4), thus obtaining the weight of the edge between any two adjacent nodes:

[0118] W s =P(node) n |node sum )·P(node m |node n(4)

[0119] In equation (4) above, W s P(node) represents conditional probability; n |node sum P(node ​​n) represents the probability of node n occurring given that all nodes have occurred; m |node n Let f(x) represent the probability of node m occurring given that node n has occurred.

[0120] See Figure 3 For example, the weight between node S1 and node S2 is represented as: P(S1|S sum )·P(S2|S1), where P(S1|S sum P(S1) represents the probability of node S1 occurring when all nodes have occurred, and P(S2|S1) represents the probability of node S2 occurring when node S1 has occurred.

[0121] A random walk model can be used to simulate the connection between nodes, and the above equation (4) can be used to calculate the conditional probability between any two adjacent nodes.

[0122] In step S5 above, see [reference] Figure 3 Based on the weights of the edges between any two adjacent nodes and the nodes, a network is constructed to obtain the target network model.

[0123] In step S6 above, the evaluation metrics include at least degree centrality, eigenvector centrality, clustering coefficient, tight centrality, and betweenness centrality. The specific process for calculating the evaluation metrics for each node may include the following steps:

[0124] Step S61: For each node, calculate the degree centrality based on the node's adjacency matrix.

[0125] In step S61 above, the degree centrality represents the number of edges connected to a node, and is calculated according to the following formula (5):

[0126]

[0127] In equation (5) above, D i Let A represent the degree centrality of node i, N represent the total number of nodes, and A''''''''''''''''''''''''''''''''''''''''''''''""","' ... i Let i represent the adjacency matrix of node i.

[0128] Step S62: Based on the eigenvalues ​​of the obtained adjacency matrix of the nodes and the adjacency matrix of the nodes, calculate the eigenvector centrality.

[0129] In step S62 above, the specific process of calculating the eigenvector centrality may include the following steps:

[0130] Step S621: Calculate the eigenvalues ​​of the adjacency matrix based on the node adjacency matrix.

[0131] Step S622: Calculate the eigenvector centrality based on the eigenvalues ​​and adjacency matrix of the adjacency matrix.

[0132] In step S622 above, eigenvector centrality not only measures whether a node is connected to more than one neighboring node, but also detects whether its neighboring nodes contain important nodes. Compared with degree centrality, it can more comprehensively observe the importance of a node. Eigenvector centrality is calculated according to the following formula (6):

[0133]

[0134] In equation (6) above, k is the eigenvalue of the adjacency matrix; This indicates the centrality of the eigenvectors.

[0135] Step S63: Based on the target network model, determine the number of triangles formed by the nodes, and calculate the clustering coefficient based on the number of triangles and degree centrality of the nodes.

[0136] In step S63 above, the importance of a node is directly related to the clustering coefficient. Furthermore, the clustering coefficient can also describe the propagation of the network. For nodes with the same degree centrality, a higher clustering coefficient indicates stronger propagation ability. The clustering coefficient is calculated by quantifying the number of triangles connected to the node, according to the following formula (7):

[0137]

[0138] In equation (7) above, C i The degree centrality of node i is represented by t. i D represents the number of triangles that node i can form; i C represents the degree centrality of node i; when node i is not connected to any of its neighbors, C is... i =0; C = 0 when node i is connected to all its neighboring nodes. i =1.

[0139] All edges in the target network model can be enumerated using an enumeration method. For each edge, the common neighbors of the nodes corresponding to its two endpoints can be checked. Specifically, for each edge (u,v), all neighbor nodes of node u (except v) can be traversed, and it can be checked whether these neighbor nodes are also neighbors of node v. If so, a triangle is formed. The number of triangles of the nodes can be counted in this way, and the clustering coefficient can be calculated according to the above formula (7).

[0140] The number of triangles in a node can also be calculated using a graph algorithm library. Many graph algorithm libraries (such as NetworkX, igraph, etc.) provide built-in functions for calculating the number of triangles. By processing the target network model through the graph algorithm library, the number of triangles in the node can be obtained, and the clustering coefficient can be calculated using the above formula (7).

[0141] Step S64: Calculate the compact centrality based on the target network model.

[0142] In step S64 above, compact centrality is used to evaluate the average distance between a node and other nodes. The higher the compact centrality of a node, the more direct its influence on other nodes, which is beneficial to the propagation of node-related information. The process of calculating compact centrality may include:

[0143] For each node, based on the target network model, the number of nodes and the total number of edges along the path from the node to all other nodes are calculated, and the compact centrality is calculated according to the following formula (8):

[0144] Cl i =n iv / Σ v d iv (8)

[0145] In equation (8) above, Cl i Represents the compact centrality of node i; n iv d represents the number of nodes traversed on the path from node i to node v; iv This represents the total number of edges traversed on the path from node i to node v.

[0146] Step S65: Calculate the betweenness centrality based on the target network model.

[0147] In step S65 above, betweenness centrality is used to evaluate the influence of a node on the information flow between other node pairs. A higher betweenness centrality indicates that the node has a controlling effect on information transmission between other nodes and has a greater influence in the entire target network model. The specific process of calculating betweenness centrality may include the following steps:

[0148] Based on the target network model, for any two remaining nodes excluding the node, calculate the number of all shortest paths between the two remaining nodes, and obtain the number of shortest paths that pass through the nodes. Calculate the betweenness centrality according to the following formula (9):

[0149]

[0150] In equation (9) above, B i The degree of centrality of node i is represented by l.st This represents the number of all shortest paths between the remaining node s and the remaining node t; This represents the number of shortest paths that pass through node i.

[0151] In step S7 above, the specific process of selecting the operating conditions with the highest vulnerability may include the following steps:

[0152] Based on the evaluation indicators of each node, the operating conditions are sorted according to the value of the evaluation indicators, and the operating conditions that occur most frequently and are ranked first are selected as the operating conditions with the highest vulnerability.

[0153] Since each node corresponds to a certain operating condition, based on the evaluation indicators of each node, the value of the corresponding evaluation indicator for each operating condition can be determined. Then, the operating conditions are sorted according to the value of the evaluation indicator, and the operating condition that appears most frequently and is ranked first is selected as the operating condition with the highest vulnerability.

[0154] For example, a case study analysis was conducted using two months of sample data collected from March to April 2020 and October to November 2020 for a certain oil pipeline in southern China (495.1 km long, including 9 oil pumps: specifically 4 pumping stations, 3 distribution stations, and 2 terminal stations). The data includes ① operational conditions of each station: start-up and shutdown of the entire pipeline, start-up and shutdown of intermediate pumping stations, and oil product switching; and ② pipeline operating parameters: inlet and outlet pressure, inlet and outlet flow rate, downstream flow rate, inlet and outlet pressure of the main pump and feed pump, oil density, etc. The sampling frequency was 1 second, and the sample time span for each operating parameter was 5,184,000 seconds. Table 1 below shows the statistical results of the signals collected at each station. (See reference...) Figure 4 The diagram shows the topology of the target oil pipeline. The supplier delivers oil sequentially to the first station A, pump station B, distribution station C, pump station D, distribution station E, pump station F, and distribution station G. The oil is then delivered to the demand side through the terminal stations H and I, which are connected to distribution station G. There are shut-off valve chambers between any two adjacent oil pumps, such as between the supplier and the first station A, between the first station A and pump station B, and between pump station D and distribution station E.

[0155] Table 1 Statistical results of signals collected at each station

[0156]

[0157] Based on the sample data of a certain oil pipeline, the evaluation indicators for each node are calculated using the methods described in steps S1-S6 above. These evaluation indicators include degree centrality, eigenvector centrality, clustering coefficient, compact centrality, and betweenness centrality. Operating conditions include full-line start-up, full-line shutdown, oil product switching, and intermediate station pump start-up / shutdown. Each node corresponds to one operating condition. The calculation results of the evaluation indicators for each operating condition and the corresponding number of nodes are statistically analyzed under different evaluation indicators, and plotted as follows: Figure 5 The results are shown in the calculation graph.

[0158] Based on the evaluation metrics for each node, the operating conditions with the highest vulnerability are selected. Based on the calculation results of each evaluation metric, the operating conditions are ranked according to the numerical value of the metric, and the operating condition corresponding to the maximum value of each evaluation metric is determined. For details, please refer to [link / reference]. Figure 5 (a) The operating condition corresponding to the maximum degree centrality is the full-line start-up operating condition; see reference Figure 5 (b) The operating condition corresponding to the maximum clustering coefficient is the start-up and shutdown condition of the intermediate station pump; see reference Figure 5 (c) The operating condition corresponding to the maximum intermediation centrality is the start-up and shutdown condition of the intermediate station pump; see reference Figure 5 (d) The operating condition corresponding to the maximum tightness of the centrality is the start-up and shutdown condition of the intermediate station pump; see reference Figure 5 (e) The operating condition corresponding to the maximum value of the eigenvector centrality is the start-up and shutdown condition of the intermediate station pump. Based on the above analysis results, the operating condition that appears most frequently and is ranked first is selected as the start-up and shutdown condition of the intermediate station pump. That is, the start-up and shutdown condition of the intermediate station pump is the operating condition that appears most frequently and corresponds to the maximum value of each evaluation index. The start-up and shutdown condition of the intermediate station pump is regarded as the operating condition with the highest vulnerability.

[0159] By comprehensively analyzing the evaluation indicators of each node, the mutual influence between various operating conditions can be obtained. Figure 5 (a) Figure 5 (b) and Figure 5 As can be seen from (d), the start-up condition of the entire pipeline has a high degree centrality, clustering coefficient and compact centrality, indicating that the condition has a more direct impact on other conditions in the pipeline. This also confirms that the conditions after the pipeline starts up will change according to the actual site conditions, which is in line with the actual engineering situation.

[0160] from Figure 5 (a) Figure 5 (c) and Figure 5As can be seen from (e), the start-up and shutdown of intermediate station pumps occur most frequently in the lower degree centrality, while the intermediate centrality and eigenvector centrality are higher, meaning that there are fewer other nodes connected to this condition, and it usually enters the next condition directly through a single transmission. This indicates that this condition has high information transmission capacity, which means that in practical engineering applications, the start-up and shutdown of intermediate station pumps has a significant impact on the changes in the overall operation of adjacent stations or pipeline systems.

[0161] from Figure 5 (a) Figure 5 (b) Figure 5 (d) and Figure 5 As shown in (e), the degree centrality of most oil product switching conditions is low, with only 3 nodes exhibiting relatively high degree centrality. This condition's clustering coefficient and compact centrality are second only to the full-line start-up condition, and its eigenvector centrality value is also low. This is because oil product changes are generally carried out according to the monthly work plan, occurring periodically. Adjustments to the operating conditions arise based on the changes in oil product shipment batches, but these adjustments do not affect the normal operation of the pipeline.

[0162] from Figure 5 It can be seen that all evaluation indicators for the full-line shutdown condition are low. This is because the pipeline's operating status does not change after the shutdown, and the full-line shutdown condition only appeared in the system during the fifth week of the data collection, thus its impact is relatively small. This condition may occur for pipelines with intermittent transportation, but it is rare for other types of pipeline transportation processes.

[0163] In this embodiment, the target operating condition data is segmented to obtain a multidimensional time series. Correlation analysis and feature extraction are performed on the multidimensional time series to obtain the corresponding adjacency matrix and input features. A time series subgraph is obtained based on the adjacency matrix and input features. Based on the time series subgraph, the weights of edges between adjacent nodes are calculated to obtain the target network model. Evaluation indicators for each node are calculated, and the trend of evaluation indicators over time is monitored. This method can identify vulnerable operating conditions in refined oil pipeline systems, thereby helping on-site personnel to make rapid judgments and decisions on events with potential safety hazards. It features high timeliness and accuracy. From a graph theory perspective, a relationship network based on multiple operating conditions of oil pipelines, i.e., the target network model, is constructed for the first time. Using complex network indicators, an evaluation method for vulnerable operating conditions of oil pipelines is proposed, realizing the identification of vulnerable operating conditions of pipelines. This advances the basic research on vulnerability identification of pipeline operating conditions and can provide decision support for managers in the operation and maintenance management of pipeline systems. It has high application value, and the method can be extended to oil pipelines and gas pipelines.

[0164] Example 2

[0165] The inventors discovered that currently, the identification of most pipeline operating conditions relies primarily on the experience of dispatchers. Dispatchers combine hydraulic and thermal models of oil and gas pipelines to dynamically track the trends of monitoring point data, thereby determining whether abnormal operating conditions have occurred. However, with the rapid development of refined oil pipelines and the increasing scale of the pipeline network, frequent changes in operating conditions and complex hydraulic variations have made manual monitoring increasingly difficult. It is impossible to dynamically track the pipeline's operating status and address the multiple challenges currently facing refined oil pipelines in terms of operation, safety, and supply. Furthermore, with the increase in pipeline monitoring points, refined oil pipeline systems generate a large amount of data daily. This data is scattered across different information systems or professional systems, forming a massive accumulation of multi-source heterogeneous data. The refined oil pipeline data acquisition and monitoring control system suffers from problems such as heavy storage and insufficient analysis, wasting a large amount of potential data.

[0166] With the development of computer science, machine learning and deep learning models are gradually being applied. In other industrial fields, intelligent identification methods based on machine learning can greatly improve the efficiency of traditional operating condition identification methods. Therefore, many scholars are striving to introduce this work into the field of detection and identification of operating conditions of oil and gas pipelines.

[0167] Based on the published research results, the intelligent identification of operating conditions of oil and gas pipelines mainly relies on two methods: 1. Classification methods based on time-frequency domain and statistical features; 2. Deep learning methods based on convolutional neural networks, recurrent neural networks, etc.

[0168] The limitations of the two methods mentioned above are: 1. These methods are only applicable to univariate time series, and their accuracy depends more on the representation of time series features. They lack consideration for multidimensional operating parameters in improving the accuracy of multi-class classification; 2. Both methods neglect the correlation of nonlinear operating data and the spatial correlation of signals from different sensors during pipeline operation.

[0169] Therefore, in order to solve the above problems, based on the same inventive concept as Embodiment 1, this embodiment proposes a method for identifying the operating conditions of oil pipelines, which specifically includes the following steps:

[0170] Obtain the actual operating data of the target pipeline;

[0171] The actual operating data is input into the operating condition identification model to obtain the operating condition identification results of the target pipeline.

[0172] The construction methods for the working condition identification model include:

[0173] Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, where each multidimensional time series corresponds to a certain operating condition.

[0174] Correlation analysis and feature extraction are performed on multiple multidimensional time series to obtain the adjacency matrix and input features for each multidimensional time series.

[0175] By combining the adjacency matrix and input features corresponding to each multidimensional time series, we obtain the subgraphs of each time series.

[0176] Each time series subgraph is used as a training set to train a pre-defined graph neural network model, and the trained graph neural network model is used as a working condition recognition model.

[0177] The specific process of constructing the above-mentioned working condition identification model may include the following steps:

[0178] Obtain the target operating condition data, which can be split into a training dataset and a test set in an 8:2 ratio. The process of obtaining each time series subplot based on the target operating condition data can be found in Example 1 and will not be repeated here. (See also...) Figure 6 A multi-layer graph convolutional network (GCN) model for supervised learning of multivariate time series data is constructed, i.e., a pre-defined graph neural network model, including: two graph convolutional layers G(X,A), pooling layers, two fully connected layers, and an output layer. Specifically, the adjacency matrix describing the graph structure is retained in the time series subgraph and used as the input of this pre-defined graph neural network model. The two graph convolutional layers are used to perform graph-level classification on the adjacency matrix A; each convolutional layer has a 64-dimensional output feature map and uses a non-linear activation function. In this embodiment, the non-linear activation function used is ReLU, and the definition method of ReLU is as follows:

[0179]

[0180] In equation (10) above, x is the input value of the nonlinear activation function, that is, the output value of the previous graph convolutional layer.

[0181] The layer following the two graph convolutional layers is a pooling layer, which performs mean pooling on the output values ​​of the two graph convolutional layers. This pooling layer aggregates the learned node representations to create a graph representation. The graph representation is input to two fully connected layers, each with 32 and 16 units respectively, activated by a non-linear activation function; in this embodiment, ReLU is used. The final layer is an output layer with a single unit and sigmoid activation for output.

[0182] For example, two months of sample data from March to April 2020 and October to November 2020 were used to analyze the data of a certain oil pipeline in southern China (495.1 km long, including 9 oil pumps: specifically 4 pumping stations, 3 distribution stations, and 2 terminal stations). The dataset was split into training and testing sets in an 8:2 ratio. The data includes ① operating conditions of each station: start-up and shutdown of the entire line, start-up and shutdown of intermediate pumping stations, and oil product switching; and ② pipeline operating parameters: inlet and outlet pressure, inlet and outlet flow rate, downstream flow rate, inlet and outlet pressure of the main pump and feed pump, oil density, etc. The sampling frequency was 1 second, and the sample time span for each operating parameter was 5,184,000 seconds. Table 1 of the above Example 1 shows the statistical results of the signals collected at each station. Figure 4 The diagram shows the topology of the target oil pipeline. For an explanation of the topology of the target oil pipeline, please refer to Example 1, which will not be repeated here.

[0183] See Figure 8 The model is explained using topology diagrams for each operating condition. Figure 8 (a) is a topology diagram showing the entire line is shut down, with the first station A, pumping station B, and distribution station C all in a closed state; Figure 8 (b) is the topology diagram of the entire line in operation, with the first station A, pumping station B and distribution station C all in the open state; Figure 8 (c) is a topology diagram of the intermediate station pump start-up. There are two intermediate stations B between the first station A and the distribution station C. The first station A, the distribution station C and the two intermediate stations B are all in the open state. Figure 8 (d) is the topology diagram of the intermediate station pump stoppage. There are two intermediate stations B between the first station A and the distribution station C. Both the first station A and the distribution station C are in the open state, and one of the intermediate stations B is in the closed state. Figure 8 (e) is the topology diagram of oil product switching (0#D-92#G), that is, the supplier inputs oil product 92#G, which switches to 0#D after passing through the first station A; Figure 8 (f) is the topology diagram of oil product switching (92#G-95#G), that is, the supplier inputs oil product 95#G, and after passing through the first station A, the oil product is switched to 92#G; Figure 8 (g) is the topology diagram for oil product switching (95#G-0#G), where the supplier inputs oil product 0#G, which switches to 95#G after passing through the first station A. Considering the different operating parameters that are of focus under different on-site conditions, the input data for the operating conditions of full-line shutdown, full-line startup, intermediate station pump startup, and intermediate station pump shutdown are the outlet pressure of station A, the inlet and outlet pressures of station B, and the inlet pressure of station C; for the three operating conditions corresponding to oil product switching, the input data are the oil product density at the inlet and outlet of station A and the flow rate of all download lines at station A.

[0184] See Figure 7The entire process of constructing the operating condition identification model includes: acquiring the operating dataset of the refined oil pipeline, namely two months of sample data of a certain oil pipeline in the south, performing data standardization on the operating dataset, then using the sliding data window reconstruction method to segment the data to obtain multiple multi-dimensional time series, performing correlation mechanism and LSTM feature extraction on the multi-dimensional time series respectively to obtain the adjacency matrix and input features respectively, and constructing a subgraph (dynamic), namely the time series subgraph.

[0185] The graph neural network model is trained using time-series subgraphs as the training set. After training, parameters are initialized, and the performance of the work condition recognition is evaluated using four evaluation parameters: accuracy (A), precision (P), recall (R), and F1 score. If the evaluation results meet the preset requirements, the graph neural network model is considered a successfully trained model. If the evaluation results do not meet the preset requirements, the training loss is calculated, and the hyperparameters of the graph neural network model are updated using trial and error or network search. Multi-objective optimization is then used to update other parameters in the graph neural network model, resulting in an updated model. The training and validation processes are then repeated on the updated model. After obtaining the trained graph neural network model (i.e., the work condition recognition model), it can be tested and validated using a test set.

[0186] In this embodiment, refer to Figure 9 A static graph network model (SPipeNet) was constructed as a control group. In the SPipeNet model, weights are determined by the real geographical locations of the pumping stations: when a weight exceeds a threshold, an edge exists between two nodes; otherwise, the two nodes are not connected. All weights form an adjacency matrix. If all examined time series belong to the same pumping station, a fully connected network is used by default. An LSTM model is used for feature extraction to obtain input features. Based on the input features and the adjacency matrix, a subgraph (static) is constructed. The number of nodes in the SPipeNet model is determined by the number of pumping stations; each node represents one pumping station. (See also...) Figure 7 The above subgraph (static) is used as the training set to train the graph neural network model. The training and verification process of the graph neural network model is the same as that of the above working condition recognition model, and will not be repeated here.

[0187] See Figure 10 The classification performance of different neural network models for various operating conditions was compared. The neural network model built based on DPipeNet outperformed other neural network classification models in all four metrics. Specifically, for the identification of operating conditions such as pipeline shutdown, pipeline startup, intermediate station pump startup and shutdown, and oil product switching, please refer to [reference needed]. Figure 10(a) and Figure 10 (b) The neural network model built on the DPipeNet model achieved an accuracy of 88.2% and a precision of 86.5%, demonstrating better performance than the neural network model built on the SPipeNet model. For other neural networks, ResCNN outperformed other models of the same type in all four metrics. The single LSTM model had the worst classification accuracy. The classification performance of ResNet and FCN was not significantly different. Although the classification accuracy and precision of ResCNN and SPipeNet models were similar, ResCNN performed better in recall and F1 score, making it somewhat competitive.

[0188] Ablation experiments were conducted to examine the importance of mutual information, LSTM, and GCN to the performance of the condition recognition model. Table 2 summarizes the results of these models in terms of precision, recall, and F1 score.

[0189] Table 2 Performance Comparison of Different Model Configurations

[0190]

[0191] Table 2 reveals the following: First, removing mutual information leads to a significant decrease in all metrics (precision: 83.5%; recall: 82.4%; F1: 82.9%). This indicates that mutual information is crucial for capturing nonlinear relationships between different oil pumping stations or equipment. Second, without LSTM, the model's ability to capture temporal dependencies weakens, resulting in a significant performance drop (precision: 80%; recall: 78.8%; F1: 79.4%). Finally, removing GCN causes the most significant performance decrease (precision: 77.2%; recall: 75.9%; F1: 76.5%), highlighting the critical role of spatial feature extraction in accurate state identification. Therefore, ablation experiments clearly validate the necessity of each component in improving the overall performance of the working condition identification model.

[0192] In practical applications, optimizing the parameters during the DPipeNet model construction process can improve the operational condition identification model. When the weight between two multidimensional time series is calculated to be 0.5, the sequence window length for segmenting the target operational condition data using the time series segmentation method is 1800s, and the delay time for input data from each oil station is set to 90-100s, the identification accuracy of the neural network model built based on the DPipeNet model is 6 percentage points higher than that of the neural network model built based on the DPipeNet model, and up to about 15 percentage points higher than other classic neural network models. Using the actual distance between pipeline stations as weights, the SPipeNet model, with a sequence window length of 480s and a delay time for input data from each station set to 90-100s, achieves an accuracy of 80% in identifying three oil product switching conditions, and over 85% accuracy in identifying pipeline shutdown, pipeline startup, intermediate station pump startup, and intermediate station pump shutdown.

[0193] The neural network model based on the DPipeNet model in this embodiment can be used to identify unsteady operating conditions in pipeline systems. Based on pipeline production operation data and daily operation report event records, a dynamic graph network identification method for unsteady operating conditions in pipelines is established. Compared with traditional neural network classification models and neural network models based on the DPipeNet model, the neural network model based on the DPipeNet model proposed in this embodiment can significantly improve the accuracy of identifying unsteady operating conditions and can be applied to online classification and identification of multiple operating conditions. The operating condition identification method in this embodiment overcomes the problems of insufficient representation of data space features and poor multi-condition classification performance of traditional models, achieving the goal of detecting and identifying unsteady operating conditions in pipeline operation based solely on collected operation data and events. This method can be extended to oil pipelines and gas pipelines.

[0194] Example 3

[0195] Based on the same inventive concept, see [reference] Figure 11 This embodiment proposes a vulnerability assessment device for the operating conditions of oil pipelines, comprising:

[0196] The first data processing module 101 is used to segment the data based on the acquired target operating condition data using a time series segmentation method to obtain multiple multi-dimensional time series, wherein each multi-dimensional time series corresponds to an operating condition.

[0197] The second data processing module 102 is used to perform correlation analysis and feature extraction based on multiple multidimensional time series to obtain the adjacency matrix and input features corresponding to each multidimensional time series.

[0198] The image output module 103 is used to combine the adjacency matrix and input features corresponding to each multidimensional time series to obtain each time series sub-image;

[0199] The calculation module 104 is used to take each time series subgraph as a node, calculate the conditional probability between any two adjacent nodes, and obtain the weight of the edge between any two adjacent nodes.

[0200] The network model generation module 105 is used to obtain the target network model based on the weights of the nodes and the edges between any two adjacent nodes.

[0201] The data calculation module 106 is used to calculate the evaluation index of each node based on the target network model and node data.

[0202] The screening module 107 is used to screen out the operating conditions with the highest vulnerability based on the evaluation indicators of each node.

[0203] The vulnerability assessment device for the operating conditions of oil pipelines provided in this embodiment of the invention has a similar implementation principle and technical effect to that of Embodiment 1, and will not be repeated here.

[0204] Example 4

[0205] Based on the same inventive concept, this embodiment proposes an operating condition identification device for oil pipelines, comprising:

[0206] The data acquisition module is used to acquire the actual operating data of the target pipeline;

[0207] The data input module is used to input actual operating data into the operating condition identification model to obtain the operating condition identification results of the target pipeline.

[0208] The operating condition identification device for oil pipelines provided in this embodiment of the invention has a similar implementation principle and technical effect to that of Embodiment 2, and will not be repeated here.

[0209] Example 5

[0210] Based on the same inventive concept, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the vulnerability assessment method for the operating conditions of oil pipelines as described in Embodiment 1, and / or the identification method for the operating conditions of oil pipelines as described in Embodiment 2.

[0211] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments; or it may exist independently and not assembled into the device / apparatus. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to Embodiment 1 of the present invention.

[0212] According to embodiments of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0213] Example 6

[0214] Based on the same inventive concept, this application also proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the vulnerability assessment method for the operating conditions of oil pipelines as in Embodiment 1, and / or the identification method for the operating conditions of oil pipelines as in Embodiment 2.

[0215] Example 7

[0216] Based on the same inventive concept, this application proposes a computer program product containing instructions that, when the computer program product is run on a computer device, cause the computer device to execute the vulnerability assessment method for the operating conditions of oil pipelines in Embodiment 1, and / or the identification method for the operating conditions of oil pipelines in Embodiment 2.

[0217] The principles by which the above-mentioned devices, clients, media, and related equipment in this embodiment of the invention solve the problem are similar to those of the aforementioned methods. Therefore, their implementation can refer to the implementation of the aforementioned methods, and repeated details will not be repeated.

[0218] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0219] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0220] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0221] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0222] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A vulnerability assessment method for the operating conditions of oil pipelines, characterized in that, include: Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, wherein each multidimensional time series corresponds to a certain operating condition. Correlation analysis and feature extraction are performed on the multiple multidimensional time series respectively to obtain the adjacency matrix and input features corresponding to each multidimensional time series; By combining the adjacency matrix and the input features corresponding to each of the multidimensional time series, each time series subgraph is obtained; Using each of the aforementioned time series subgraphs as nodes, calculate the conditional probability between any two adjacent nodes to obtain the weight of the edge between any two adjacent nodes; Based on the weights of the nodes and the edges between any two adjacent nodes, the target network model is obtained; Based on the target network model and the data of the nodes, the evaluation index of each node is calculated. Based on the evaluation metrics of each node, the operating conditions with the highest vulnerability are selected.

2. The vulnerability assessment method for operating conditions of oil pipelines according to claim 1, characterized in that, The operating conditions with the highest vulnerability are selected based on the evaluation indicators of each node, including: Based on the evaluation indicators of each node, the operating conditions are sorted according to the value of the evaluation indicators, and the operating conditions that occur most frequently and are ranked first are selected as the operating conditions with the highest vulnerability.

3. The vulnerability assessment method for the operating conditions of oil pipelines according to claim 2, characterized in that, The evaluation metrics include at least degree centrality, eigenvector centrality, clustering coefficient, tight centrality, and betweenness centrality; The evaluation metrics for each node are calculated based on the target network model and the node data, including: For each node, the degree centrality is calculated based on the node's adjacency matrix. Based on the eigenvalues ​​of the adjacency matrix of the node and the adjacency matrix of the node, the eigenvector centrality is calculated. Based on the target network model, the number of triangles formed by the nodes is determined, and the clustering coefficient is calculated based on the number of triangles formed by the nodes and the degree centrality. Based on the target network model, the compact centrality is calculated. Based on the target network model, the betweenness centrality is calculated.

4. The vulnerability assessment method for pipeline operating conditions according to claim 3, characterized in that, For each node, the degree centrality is calculated based on the node's adjacency matrix according to the following formula: Among them, D i Let A represent the degree centrality of node i, N represent the total number of nodes, and A''''''''''''''''''''''''''''''''''''''''''''''""","' ... i Let i represent the adjacency matrix of node i.

5. The vulnerability assessment method for operating conditions of oil pipelines according to claim 3, characterized in that, The calculation of the eigenvector centrality based on the eigenvalues ​​of the obtained adjacency matrix of the node and the adjacency matrix of the node includes: Based on the adjacency matrix of the nodes, the eigenvalues ​​of the adjacency matrix are calculated; Based on the eigenvalues ​​of the adjacency matrix and the adjacency matrix itself, the eigenvector centrality is calculated according to the following formula: Where k is the eigenvalue of the adjacency matrix; This indicates the centrality of the eigenvectors.

6. The vulnerability assessment method for operating conditions of oil pipelines according to claim 3, characterized in that, Based on the target network model, the number of triangles formed by the nodes is determined. Based on the number of triangles formed by the nodes and the degree centrality, the clustering coefficient is calculated according to the following formula: Among them, C i The degree centrality of node i is represented by t. i D represents the number of triangles that node i can form; i C represents the degree centrality of node i; when node i is not connected to any of its neighbors, C is... i =0; C = 0 when node i is connected to all its neighboring nodes. i =1.

7. The vulnerability assessment method for pipeline operating conditions according to claim 3, characterized in that, The calculation of the compact centrality based on the target network model includes: For each node, based on the target network model, the number of nodes traversed and the total number of edges on the path from that node to all other nodes are calculated, and the compact centrality is calculated according to the following formula: Cl i =n iv / ∑ v d iv Among them, Cl i Represents the compact centrality of node i; n iv d represents the number of nodes traversed on the path from node i to node v; iv This represents the total number of edges traversed on the path from node i to node v.

8. The vulnerability assessment method for operating conditions of oil pipelines according to claim 3, characterized in that, The calculation of the betweenness centrality based on the target network model includes: Based on the target network model, for any two remaining nodes excluding the stated node, the number of all shortest paths between the two remaining nodes is calculated, and the number of shortest paths passing through the stated node is also calculated. The betweenness centrality is then calculated using the following formula: Among them, B i The degree of centrality of node i is represented by l. st This represents the number of all shortest paths between the remaining node s and the remaining node t; This represents the number of shortest paths that pass through node i.

9. The vulnerability assessment method for operating conditions of oil pipelines according to claim 1, characterized in that, The process involves performing correlation analysis and feature extraction on the multiple multidimensional time series to obtain the adjacency matrix and features corresponding to each multidimensional time series, including: Based on the multiple multidimensional time series, the weights between adjacent multidimensional time series are calculated using the mutual information method; Obtain all weights to form the adjacency matrix; For each of the multidimensional time series, a preset LSTM model is used to extract features to obtain the input features corresponding to each multidimensional time series.

10. A method for identifying the operating conditions of oil pipelines, characterized in that, include: Obtain the actual operating data of the target pipeline; The actual operating data is input into the operating condition identification model to obtain the operating condition identification result of the target pipeline; The method for constructing the working condition identification model includes: Based on the acquired target operating condition data, the data is segmented using a time series segmentation method to obtain multiple multidimensional time series, wherein each multidimensional time series corresponds to a certain operating condition. Correlation analysis and feature extraction are performed on the multidimensional time series to obtain the adjacency matrix and input features corresponding to each multidimensional time series. By combining the adjacency matrix and the input features corresponding to each of the multidimensional time series, each time series subgraph is obtained; Each of the aforementioned time series subgraphs is used as a training set to train a preset graph neural network model, and the trained graph neural network model is used as the working condition recognition model.

11. A vulnerability assessment device for the operating conditions of oil pipelines, characterized in that, include: The first data processing module is used to segment the acquired target operating condition data using a time series segmentation method to obtain multiple multi-dimensional time series, wherein each multi-dimensional time series corresponds to an operating condition. The second data processing module is used to perform correlation analysis and feature extraction based on the multiple multidimensional time series respectively, to obtain the adjacency matrix and input features corresponding to each multidimensional time series; The image output module is used to combine the adjacency matrix corresponding to each of the multidimensional time series and the input features to obtain each time series sub-image; The calculation module is used to take each of the time series subgraphs as nodes, calculate the conditional probability between any two adjacent nodes, and obtain the weight of the edge between any two adjacent nodes. The network model generation module is used to obtain the target network model based on the weights of the edges between the node and any two adjacent nodes. The data calculation module is used to calculate the evaluation index of each node based on the target network model and the data of the nodes; The filtering module is used to filter out the operating conditions with the highest vulnerability based on the evaluation indicators of each node.

12. An operating condition identification device for oil pipelines, characterized in that, include: The data acquisition module is used to acquire the actual operating data of the target pipeline; The data input module is used to input the actual operating data into the operating condition identification model to obtain the operating condition identification result of the target pipeline.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the vulnerability assessment method for operating conditions of tubing as described in any one of claims 1-9, and / or the identification method for operating conditions of tubing as described in claim 10.

14. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the vulnerability assessment method for operating conditions of oil pipes as described in any one of claims 1-9, and / or the identification method for operating conditions of oil pipes as described in claim 10.

15. A computer program product comprising instructions, which, when run on a computer device, causes the computer device to perform the vulnerability assessment method for operating conditions of a pipeline as described in any one of claims 1-9, and / or the identification method for operating conditions of a pipeline as described in claim 10.

Images