Pipeline risk grade evaluation method and system

Abstract

The invention discloses a pipeline risk grade evaluation method and system, and belongs to the technical field of pipeline detection. The method comprises the following steps: constructing a pipeline risk grade evaluation model, wherein the pipeline risk grade evaluation model comprises a first prediction branch, a second prediction branch and a fusion branch; training the first prediction branch based on the historical accident data to obtain an optimized first training model; inputting the real-time measurement data into a first training model to obtain a first operation state of the to-be-measured pipeline; the second prediction branch obtains a second operation state of the to-be-measured pipeline according to the historical state data and the real-time measurement data; and the fusion branch outputs a risk grade evaluation result based on the first operation state and the second operation state. According to the method, multiple factors of pipeline operation can be comprehensively considered, so that the risk evaluation is more comprehensive, the deviation possibly caused by a single data source is avoided, and the risk point of pipeline inspection is accurately identified according to risk early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline detection, and more particularly to a pipeline risk level evaluation method and system. Background Art

[0002] As an important part of modern industrial and livelihood infrastructure, pipelines are widely used in many fields such as petroleum, natural gas, chemical industry, electricity, metallurgy, etc. Pipeline risk assessment refers to the analysis process of identifying the hazards that have an adverse effect on the safe operation of pipelines, evaluating the possibility and consequences of accidents, and then calculating the risk size and proposing risk control measures. This process is of great significance for pipeline managers to fully understand pipeline risks and formulate scientific and effective maintenance and repair strategies.

[0003] At present, traditional pipeline companies have many valuable accident case resources, but they have not deeply explored the value of accident cases, have not fully utilized accident information resources, and the sharing and prevention of safety risk information have not been effectively disseminated and promoted. In addition, due to differences in years of work and professional capabilities, on-site employees lack effective understanding of on-site risks, resulting in inaccurate risk assessment.

[0004] Therefore, it is an urgent problem for technical personnel in this field to propose a pipeline risk level evaluation method and system to improve the accuracy of pipeline risk level evaluation results. Summary of the invention

[0005] In view of this, the present invention provides a pipeline risk level evaluation method and system, which is based on the pipeline risk level evaluation model, takes into account historical accident data and historical status data of the pipeline to be tested, and integrates multi-source data to make the pipeline risk level evaluation result more accurate.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] In one aspect, the present invention discloses a pipeline risk level evaluation method, comprising the following steps:

[0008] Collecting historical pipeline data, wherein the historical pipeline data includes historical accident data of all pipelines and historical status data of the pipeline to be tested;

[0009] Constructing a pipeline risk level evaluation model, the pipeline risk level evaluation model comprising a first prediction branch, a second prediction branch and a fusion branch; the first prediction branch and the second prediction branch are both connected to the fusion branch;

[0010] Training the first prediction branch based on the historical accident data to obtain an optimized first training model;

[0011] Acquire real-time measurement data of the pipeline to be tested, input the real-time measurement data into a first training model, and obtain a first operating state of the pipeline to be tested;

[0012] The second prediction branch obtains a second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data;

[0013] The fusion branch outputs a risk level evaluation result based on the first operating state and the second operating state.

[0014] Preferably, the historical accident data includes historical measurement data and corresponding operating status;

[0015] The historical data of the pipeline to be tested is the measurement data of the pipeline to be tested over a period of time in the past;

[0016] The measurement data includes pipeline pressure, temperature, flow rate, and composition data.

[0017] Preferably, the risk prediction branch includes an attention network and a residual network connected in sequence.

[0018] Preferably, the second prediction branch uses a k-nearest neighbor algorithm to obtain a second operating state of the pipeline to be tested;

[0019] The second prediction branch obtains the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data, including:

[0020] Calculate the Euclidean distance between real-time measurement data and historical status data;

[0021] The second operating state of the pipeline to be tested is determined according to the Euclidean distance.

[0022] Preferably, the fusion branch outputs a risk level evaluation result based on the first operating state and the second operating state, including:

[0023] Determine the weight coefficients of historical accident data and historical status data based on the analytic hierarchy process;

[0024] Calculate the grade evaluation index according to the weight coefficient;

[0025] The risk level is output based on the mapping relationship between the risk assessment index and the risk level.

[0026] Preferably, the risk level includes none, low, medium, and high;

[0027] When the output risk level is medium or high, a risk warning is issued and the risk monitoring point is located.

[0028] Preferably, the method further comprises: setting a first threshold and a second threshold;

[0029] When the first operating state exceeds a first threshold and / or the second operating state exceeds a second threshold, a risk warning is issued and a risk monitoring point is located at the same time.

[0030] On the other hand, the present invention also discloses a pipeline risk level evaluation system, comprising:

[0031] A data acquisition module, used to collect historical pipeline data, wherein the historical pipeline data includes historical accident data of all pipelines and historical status data of the pipeline to be tested;

[0032] Model building module, used to build pipeline risk level evaluation model;

[0033] The model building module includes a first prediction branch unit, a second prediction branch unit and a fusion branch unit;

[0034] The first prediction branch unit includes a training subunit and a first prediction subunit:

[0035] The training subunit trains the first prediction branch based on the historical accident data to obtain an optimized first training model;

[0036] The first prediction subunit is used to input the real-time measurement data into the first training model to obtain a first operating state of the pipeline to be tested;

[0037] The second prediction branch unit is used to obtain the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data;

[0038] The fusion branch unit obtains a grade evaluation index based on the first operating state and the second operating state;

[0039] The risk level output module is used to output the risk level evaluation results according to the level evaluation index.

[0040] Preferably, the second prediction branch unit includes:

[0041] A calculation subunit, used for calculating the Euclidean distance between the real-time measurement data and the historical status data;

[0042] The second prediction subunit is used to determine a second operating state of the pipeline to be tested according to the Euclidean distance.

[0043] Preferably, the fusion branch unit includes:

[0044] A weight allocation subunit is used to determine the weight coefficients of historical accident data and historical status data based on the analytic hierarchy process;

[0045] The index calculation subunit is used to calculate the grade evaluation index according to the weight coefficient.

[0046] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a pipeline risk level evaluation method and system, constructs a pipeline risk level evaluation model, and the pipeline risk level evaluation model includes a first prediction branch, a second prediction branch and a fusion branch; the first prediction branch is trained based on historical accident data to obtain an optimized first training model; the real-time measurement data is input into the first training model to obtain the first operating state of the pipeline to be tested; the second prediction branch obtains the second operating state of the pipeline to be tested based on the historical state data and the real-time measurement data; the fusion branch outputs the risk level evaluation result based on the first operating state and the second operating state. The present invention can comprehensively consider various factors of pipeline operation, making the risk evaluation more comprehensive, avoiding the deviation that may be caused by a single data source, and accurately identifying the risk points of pipeline inspection according to risk warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flow chart of the method provided by the present invention;

[0048] Figure 2 This is a system architecture diagram provided by the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 creative work are within the scope of protection of the present invention.

[0050] On the one hand, the embodiment of the present invention discloses a pipeline risk level evaluation method, referring to Figure 1 , the method comprises the following steps:

[0051] S1. Collect historical pipeline data, which includes historical accident data of all pipelines and historical status data of pipelines to be tested.

[0052] Specifically, the historical accident data includes historical measurement data and corresponding operating status, and the operating status includes abnormal (indicated by the number 1) and normal (indicated by the number 0);

[0053] The historical data of the pipeline to be tested is the measurement data of the pipeline to be tested over a period of time in the past;

[0054] The measured data include pipeline pressure, temperature, flow, composition and other data.

[0055] S2. Construct a pipeline risk level evaluation model, which includes a first prediction branch, a second prediction branch and a fusion branch; the first prediction branch and the second prediction branch are both connected to the fusion branch.

[0056] S21. The risk prediction branch includes an attention network and a residual network connected sequentially.

[0057] The attention network is used to extract the target feature vector of the measurement data based on the multi-head attention mechanism and the channel attention mechanism.

[0058] The multi-head attention mechanism can learn the association between measurement data based on multiple scaled dot product attentions and build a complete global dependency, while the channel attention mechanism can determine the importance of the measurement data in each channel dimension to the output result of the first running state.

[0059] If the attention network unit is θ α is the structural parameter of the attention network unit, and the measurement data is X = [x1, x2, ..., x c ], c is the measurement data dimension, x c represents the c-th dimension measurement data, then the target feature vector of the measurement data is Then there is

[0060] The residual network is used to perform analytical calculation on the target feature vector based on multiple residual blocks, and output a first operating state.

[0061] The residual network can be composed of multiple residual blocks, each of which consists of two convolutional layers and a skip connection. The input feature of the residual block is defined as X r , the output feature of the residual block is Z r , the analytical calculation process of each residual block can be described as:

[0062] Z r =F res (X r )+X r ;

[0063] F res (X r )=Conv2(δ(BN(Conv1δ(BN(X r ))))).

[0064] Among them, Conv1 and Conv2 represent two convolutional layers respectively, BN represents batch normalization, and δ represents the ReLU activation function.

[0065] When the input feature X r And the output feature Z rWhen the channel dimensions of Z do not match, the calculation process of the two changes slightly: r =F res (X r )+W s X r .

[0066] Among them, W s Represents a linear projection, which is only used as input feature X r The dimension of is changed, and a convolution with a convolution kernel size of 1×1 is used in actual construction.

[0067] In an embodiment of the present invention, the residual network may be composed of four different residual layers, each residual layer having a plurality of residual blocks, and the target feature vector is analytically calculated through the residual blocks of the four residual layers to obtain a first motion state evaluation result.

[0068] S22. The second prediction branch uses a k-nearest neighbor algorithm to obtain a second operating state of the pipeline to be tested.

[0069] S3. Train the first prediction branch based on historical accident data to obtain an optimized first training model.

[0070] S4. Acquire real-time measurement data of the pipeline to be tested, input the real-time measurement data into the first training model, and obtain a first operating state of the pipeline to be tested.

[0071] S5. The second prediction branch obtains the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data, including:

[0072] S51. Remove outliers from the collected real-time measurement data according to corresponding parameter standards.

[0073] S52. The real-time measurement data with outliers removed is normalized to make the data weights equal and converted into values ​​of 0 to 1. The real-time feature data after the normalization process is the original data.

[0074] S53. Calculate the Euclidean distance between the original data and the historical state data. The smaller the distance, the closer the current operating state of the pipeline to be tested is to the operating state identified by the historical data.

[0075] The Euclidean distance calculation formula is as follows:

[0076]

[0077] Where: x i Represents the real-time measurement data after normalization; y i Represents measurement data under known operating conditions.

[0078] S54. Determine the second operating state of the pipeline to be tested according to the Euclidean distance.

[0079] The distances calculated in S53 are sorted in ascending order, the first k distances are selected, and the frequencies of the k distances are counted. The maximum value in the frequencies is the mode, which corresponds to the operating state represented by the real-time measurement data. That is, if most of the historical operating states corresponding to the first k distances are a certain operating state, then the current operating state of the pipeline is this operating state.

[0080] S6. The fusion branch outputs a risk level evaluation result based on the first operating state and the second operating state, including:

[0081] S61. Determine the weight coefficients of historical accident data and historical status data based on the hierarchical analysis method.

[0082] The judgment matrix is ​​constructed by experts evaluating the relative importance of historical accident data and historical status data.

[0083] Calculate the maximum eigenvalue and eigenvector of the judgment matrix. The elements in the eigenvector are the weights of each criterion.

[0084] Calculate the consistency index CI and consistency ratio CR to determine whether the consistency of the judgment matrix is ​​acceptable. If CR is less than 0.1, the consistency of the judgment matrix is ​​considered acceptable; otherwise, the judgment matrix needs to be readjusted.

[0085] S62. Calculate the grade evaluation index according to the weight coefficient, the formula is as follows:

[0086] Grade evaluation index = w1*first operating state + w2*second operating state;

[0087] Where w1 and w2 are the weight coefficients of historical accident data and historical status data respectively.

[0088] S63. Output the risk level according to the mapping relationship between the risk assessment index and the risk level.

[0089] According to the size of the risk assessment index, the risks are divided into different levels. The level classification standard can be determined according to the actual situation of the pipeline.

[0090] In this embodiment, the risk levels include none, low, medium, and high;

[0091] When the output risk level is medium or high, a risk warning is issued and the risk monitoring point is located.

[0092] Preferably, the method further comprises: setting a first threshold and a second threshold;

[0093] When the first operating state exceeds a first threshold and / or the second operating state exceeds a second threshold, a risk warning is issued and a risk monitoring point is located at the same time.

[0094] On the other hand, the embodiment of the present invention also discloses a pipeline risk level evaluation system, such as Figure 2 As shown, the system includes:

[0095] The data collection module is used to collect historical pipeline data, which includes historical accident data of all pipelines and historical status data of pipelines to be tested;

[0096] Model building module, used to build pipeline risk level evaluation model;

[0097] The model building module includes a first prediction branch unit, a second prediction branch unit and a fusion branch unit;

[0098] The first prediction branch unit includes a training subunit and a first prediction subunit:

[0099] The training subunit trains the first prediction branch based on the historical accident data to obtain an optimized first training model;

[0100] The first prediction subunit is used to input the real-time measurement data into the first training model to obtain the first operating state of the pipeline to be tested;

[0101] The second prediction branch unit is used to obtain the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data; the real-time measurement data is obtained based on the real-time data acquisition module.

[0102] The fusion branch unit obtains a grade evaluation index based on the first operating state and the second operating state;

[0103] The risk level output module is used to output the risk level evaluation result according to the level evaluation index, and specifically output the risk level according to the mapping relationship between the risk evaluation index and the risk level.

[0104] Preferably, the second prediction branch unit includes:

[0105] A calculation subunit, used for calculating the Euclidean distance between the real-time measurement data and the historical status data;

[0106] The second prediction subunit is used to determine a second operating state of the pipeline to be tested according to the Euclidean distance.

[0107] Preferably, the fusion branch unit includes:

[0108] A weight allocation subunit is used to determine the weight coefficients of historical accident data and historical status data based on the analytic hierarchy process;

[0109] The index calculation subunit is used to calculate the grade evaluation index according to the weight coefficient.

[0110] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0111] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A pipeline risk level evaluation method, It is characterized in that The following steps are involved: Collecting historical pipeline data, wherein the historical pipeline data includes historical accident data of all pipelines and historical status data of the pipeline to be tested; Constructing a pipeline risk level evaluation model, the pipeline risk level evaluation model comprising a first prediction branch, a second prediction branch and a fusion branch; the first prediction branch and the second prediction branch are both connected to the fusion branch; Training the first prediction branch based on the historical accident data to obtain an optimized first training model; Acquire real-time measurement data of the pipeline to be tested, input the real-time measurement data into a first training model, and obtain a first operating state of the pipeline to be tested; The second prediction branch obtains a second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data; The fusion branch outputs a risk level evaluation result based on the first operating state and the second operating state.

2. A pipeline risk level evaluation method and system according to claim 1, It is characterized in that The historical accident data includes historical measurement data and corresponding operating status; The historical data of the pipeline to be tested is the measurement data of the pipeline to be tested over a period of time in the past; The measurement data includes pipeline pressure, temperature, flow rate, and composition data.

3. A pipeline risk level evaluation method according to claim 1, It is characterized in that The risk prediction branch includes an attention network and a residual network connected in sequence.

4. A pipeline risk level evaluation method according to claim 1, It is characterized in that The second prediction branch uses a k-nearest neighbor algorithm to obtain a second operating state of the pipeline to be tested; The second prediction branch obtains the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data, including: Calculate the Euclidean distance between real-time measurement data and historical status data; The second operating state of the pipeline to be tested is determined according to the Euclidean distance.

5. A pipeline risk level evaluation method according to claim 1, It is characterized in that The fusion branch outputs a risk level evaluation result based on the first operating state and the second operating state, including: Determine the weight coefficients of historical accident data and historical status data based on the analytic hierarchy process; Calculate the grade evaluation index according to the weight coefficient; The risk level is output based on the mapping relationship between the risk assessment index and the risk level.

6. A pipeline risk level evaluation method according to claim 5, It is characterized in that The risk levels include none, low, medium, and high; When the output risk level is medium or high, a risk warning is issued and the risk monitoring point is located.

7. A pipeline risk level evaluation method according to claim 1, It is characterized in that Also includes: Setting a first threshold and a second threshold; When the first operating state exceeds a first threshold and / or the second operating state exceeds a second threshold, a risk warning is issued and a risk monitoring point is located at the same time.

8. A pipeline risk level assessment system, It is characterized in that include: A data acquisition module, used to collect historical pipeline data, wherein the historical pipeline data includes historical accident data of all pipelines and historical status data of the pipeline to be tested; Model building module, used to build pipeline risk level evaluation model; The model building module includes a first prediction branch unit, a second prediction branch unit and a fusion branch unit; The first prediction branch unit includes a training subunit and a first prediction subunit: The training subunit trains the first prediction branch based on the historical accident data to obtain an optimized first training model; The first prediction subunit is used to input the real-time measurement data into the first training model to obtain a first operating state of the pipeline to be tested; The second prediction branch unit is used to obtain the second operating state of the pipeline to be tested according to the historical state data and the real-time measurement data; The fusion branch unit obtains a grade evaluation index based on the first operating state and the second operating state; The risk level output module is used to output the risk level evaluation results according to the level evaluation index.

9. A pipeline risk level evaluation system according to claim 8, It is characterized in that The second prediction branch unit comprises: A calculation subunit, used for calculating the Euclidean distance between the real-time measurement data and the historical status data; The second prediction subunit is used to determine a second operating state of the pipeline to be tested according to the Euclidean distance.

10. A pipeline risk level evaluation system according to claim 8, It is characterized in that The fusion branch unit comprises: A weight allocation subunit is used to determine the weight coefficients of historical accident data and historical status data based on the analytic hierarchy process; The index calculation subunit is used to calculate the grade evaluation index according to the weight coefficient.

Images