Gas pipeline safety monitoring and early warning system based on multi-modal data fusion

Abstract

The invention provides a gas pipeline safety monitoring and early warning system based on multi-modal data fusion, which relates to the technical field of gas pipeline monitoring and comprises a distributed optical fiber sensor, a potential measurement module and a monitoring optical camera. The distributed optical fiber sensors are arranged on the outer wall of the pipeline, distributed optical fiber sensors for measuring strain data, temperature data and vibration data are arranged at the same time, and the data measured by the distributed optical fiber sensors are transmitted to the data control center through the demodulator; the potential measurement module comprises a voltmeter, a communication module, a battery module and a positioning module, and stray current measured by the voltmeter is transmitted to the data control center through the communication module; and the monitoring optical camera is arranged above the key area and is used for continuously shooting the ground condition above the pipeline for 24 hours and transmitting the measured data to the data control center. The operation safety of the gas pipeline can be effectively guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of gas pipeline monitoring, and in particular to a gas pipeline safety monitoring and early warning system based on multimodal data fusion. Background Art

[0002] Gas pipelines are the lifeline of the country and bear the important responsibility of transmitting gas. Due to the influence of various human and natural factors, gas pipelines have long been facing the risk of damage and leakage. Among them, human factors mainly include mechanical overload, improper construction, human damage and other factors, and natural factors mainly include pipeline corrosion, settlement, ground fissures, landslides, mudslides and other factors. At present, the existing gas pipeline safety monitoring system mainly identifies pipeline leaks based on the changes in temperature, strain and vibration transmitted by distributed optical fiber sensors. However, this method is greatly affected by environmental and human interference, has low accuracy, and often causes false alarms. Summary of the invention

[0003] In response to the above-mentioned prior art, the present invention proposes a gas pipeline safety monitoring and early warning system based on multimodal data fusion. The algorithm integrates multiple sensor data parameters such as distributed optical fiber sensors, potential measurement modules, and monitoring optical cameras, which improves the accuracy of gas pipeline safety monitoring and ensures the operational safety of the gas pipeline.

[0004] The present invention provides a gas pipeline safety monitoring and early warning system based on multimodal data fusion, including a distributed optical fiber sensor, a potential measurement module, and a monitoring optical camera; The distributed optical fiber sensor is arranged on the outer wall of the pipeline, and a distributed optical fiber sensor for measuring three kinds of data, namely strain, temperature and vibration, is arranged at the same time. The data measured by the distributed optical fiber sensor is transmitted to the data control center through a demodulator; The potential measurement module includes a voltmeter, a communication module, a battery module and a positioning module, and the stray current measured by the voltmeter is transmitted to the data control center through the communication module; The monitoring optical camera is deployed above the key area to continuously photograph the ground conditions above the pipeline 24 hours a day and transmit the measured data to the data control center.

[0005] As a further technical solution, the monitoring method includes the following steps: Step 1: Obtain a large number of training samples of strain, temperature, vibration and potential of the potential measurement module of the distributed optical fiber sensor under the influence of human factors and natural factors from the Internet or experiments; the human factors include but are not limited to mechanical overload, improper construction, and human damage, and the natural factors include but are not limited to pipeline corrosion, settlement, ground fissures, landslides, and mud-rock flows.

[0006] Step 2: Classify and store the training samples according to different types of human factors and natural factors; the following data analysis process takes mechanical overload as an example, and the data processing methods of other influencing factors are the same; Step 3: Preprocess the data of training samples under mechanical overload; the training sample data of strain, temperature, vibration, and potential are text data, and Pandas is used to preprocess the noise, anomalies, missing, and duplication in the training sample data; specifically, taking strain as an example, the global maximum value in the training sample data of each strain is taken out and set as Value, from Take 100 data points forward and backward respectively and record them as , which is the training sample data of n strains; similarly, the temperature takes the global minimum value, the vibration and potential take the global maximum value, and the training sample data of temperature, vibration and potential are processed to obtain the training sample data of n temperatures ; n vibration training sample data ; Training sample data of n potentials ; The training sample data of strain, temperature, vibration and potential are spliced ​​together in sequence to form n 201 rows and 4 columns of matrices Training samples of arrays , recorded as ; Substitute n training sample matrices in turn, and obtain a new matrix with 201 rows and 4 columns through the training of MLP layer and Softmax layer , recorded as , which is the characteristic matrix of strain, temperature, vibration, and potential under mechanical overload; by the same method, the characteristic matrix of strain, temperature, vibration, and potential under other influencing factors i can be obtained ; Step 4: Download a large number of pictures of vehicles, people, subsidence, ground fissures, landslides, and mudslides from the Internet. Through sample training of the MLP layer and the Softmax layer, the computer can correctly identify people, various vehicles, and ground disaster change information to obtain the image recognition feature matrix ,in, is the character information feature data, is the vehicle information feature data, It is the ground information characteristic data; Step 5: The characteristic matrix of strain, temperature, vibration and potential under various influencing factors And the image recognition feature matrix Splice them separately to form a new risk feature identification matrix ; Step 6: During the operation of the gas pipeline safety monitoring and early warning system, text data of strain, temperature, vibration, potential and image data of the optical camera are collected at the same time. The training sample method is used to obtain data at certain time intervals and stitch them into a real-time data matrix ; Step 7: Using the mutual information principle, calculate the risk characteristic matrix of the gas pipeline safety monitoring and early warning system and real-time data matrix of gas pipeline safety monitoring and early warning system The correlation between them is calculated as ;in, for and The joint probability density function between ; and Are marginal probability density functions, corresponding to and ; Step 8: Based on the calculation results Analyze and determine whether there are risks in the gas pipeline and the risk type.

[0007] As a further technical solution, the early warning method is to use the long short-term memory neural network (LSTM) to predict the data based on the current data collected by the system, and then compare the predicted data with the feature matrix to determine whether a risk accident will occur in the future, so as to improve the safety of the gas pipeline safety monitoring and early warning system. Specifically, it includes the following steps: Step 1: Obtain the current data matrix of the gas pipeline safety monitoring and early warning system , that is, in the LSTM method Input value And the output value at the previous moment ; Step 2: Enter the value And the output value at the previous moment Obtaining the forget gate through the sigmoid function ,in, is the weight coefficient, It is a deviation; Step 3: Enter the value And the output value at the previous moment The input gate is obtained by the sigmoid function and the tanh function ; Step 4: By formula , we can obtain the memory information at time t ; Step 5: Enter Values , the output value at the previous moment and the memory information at time t The output gate is obtained by using the sigmoid function and the tanh function ; Step 6: By formula Calculate the predicted output value at time t ; Step 7: Use the mutual information method to predict the output value and risk signature matrix Compare and get the calculation results , based on the results, determine whether there will be risks in the future and the type of risks that will occur.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a gas pipeline safety monitoring and early warning system based on multimodal data fusion. The entire system integrates multimodal sensor data, which greatly improves the accuracy of risk identification and risk early warning, reduces the incidence of false alarms, and can greatly reduce the number of operation and maintenance personnel, fundamentally ensuring the operation and maintenance safety of the gas pipeline. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 Schematic diagram of the composition of the gas pipeline safety monitoring and early warning system in an embodiment of the present invention.

[0010] Figure 2 This is a flow chart of a data processing method for a gas pipeline safety monitoring and early warning system in an embodiment of the present invention.

[0011] Figure 3 The present invention is a flowchart of a method for calculating the predicted output value of a gas pipeline safety monitoring and early warning system based on LSTM in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below with reference to specific diagrams.

[0013] Example: Figure 1-3 A gas pipeline safety monitoring and early warning system based on multimodal data fusion is shown, including a distributed optical fiber sensor, a potential measurement module, and a monitoring optical camera; The distributed optical fiber sensor is arranged on the outer wall of the pipeline, and a distributed optical fiber sensor for measuring three kinds of data, namely strain, temperature and vibration, is arranged at the same time. The data measured by the distributed optical fiber sensor is transmitted to the data control center through a demodulator; The potential measurement module includes a voltmeter, a communication module, a battery module and a positioning module, and the stray current measured by the voltmeter is transmitted to the data control center through the communication module; The monitoring optical camera is deployed above the key area to continuously photograph the ground conditions above the pipeline 24 hours a day and transmit the measured data to the data control center.

[0014] Furthermore, the monitoring method of the gas pipeline safety monitoring and early warning system includes the following steps: Step 1: Obtain a large number of training samples of strain, temperature, vibration and potential of the potential measurement module of the distributed optical fiber sensor under the influence of human factors and natural factors from the Internet or experiments; the human factors include but are not limited to mechanical overload, improper construction, and human damage, and the natural factors include but are not limited to pipeline corrosion, settlement, ground fissures, landslides, and mud-rock flows.

[0015] Step 2: Classify and store the training samples according to different types of human factors and natural factors; the following data analysis process takes mechanical overload as an example, and the data processing methods of other influencing factors are the same; Step 3: Preprocess the data of training samples under mechanical overload; the training sample data of strain, temperature, vibration, and potential are text data, and Pandas is used to preprocess the noise, anomalies, missing, and duplication in the training sample data; specifically, taking strain as an example, the global maximum value in the training sample data of each strain is taken out and set as Value, from Take 100 data points forward and backward respectively and record them as , which is the training sample data of n strains; similarly, the temperature takes the global minimum value, the vibration and potential take the global maximum value, and the training sample data of temperature, vibration and potential are processed to obtain the training sample data of n temperatures ; n vibration training sample data ; Training sample data of n potentials ; The training sample data of strain, temperature, vibration and potential are spliced ​​together in sequence to form n 201 rows and 4 columns of matrices Training samples of arrays , recorded as ; Substitute n training sample matrices in turn, and obtain a new matrix with 201 rows and 4 columns through the training of MLP layer and Softmax layer , recorded as , which is the characteristic matrix of strain, temperature, vibration, and potential under mechanical overload; by the same method, the characteristic matrix of strain, temperature, vibration, and potential under other influencing factors i can be obtained ; Step 4: Download a large number of pictures of vehicles, people, subsidence, ground fissures, landslides, and mudslides from the Internet. Through sample training of the MLP layer and the Softmax layer, the computer can correctly identify people, various vehicles, and ground disaster change information to obtain the image recognition feature matrix ,in, is the character information feature data, is the vehicle information feature data, It is the ground information characteristic data; Step 5: The characteristic matrix of strain, temperature, vibration and potential under various influencing factors And the image recognition feature matrix Splice them separately to form a new risk feature identification matrix ; Step 6: During the operation of the gas pipeline safety monitoring and early warning system, text data of strain, temperature, vibration, potential and image data of the optical camera are collected at the same time. The training sample method is used to obtain data at certain time intervals and stitch them into a real-time data matrix ; Step 7: Using the mutual information principle, calculate the risk characteristic matrix of the gas pipeline safety monitoring and early warning system and real-time data matrix of gas pipeline safety monitoring and early warning system The correlation between them is calculated as ;in, for and The joint probability density function between ; and Are marginal probability density functions, corresponding to and ; Step 8: Based on the calculation results Analyze and determine whether there are risks in the gas pipeline and the risk type.

[0016] Furthermore, the early warning method of the gas pipeline safety monitoring and early warning system is to use the long short-term memory neural network (LSTM) to predict the data based on the current data collected by the system, and then compare the predicted data with the feature matrix to determine whether a risk accident will occur in the future, so as to improve the safety of the gas pipeline safety monitoring and early warning system, which specifically includes the following steps: Step 1: Obtain the current data matrix of the gas pipeline safety monitoring and early warning system , that is, in the LSTM method Input value And the output value at the previous moment ; Step 2: Enter the value And the output value at the previous moment Obtaining the forget gate through the sigmoid function ,in, is the weight coefficient, It is a deviation; Step 3: Enter the value And the output value at the previous moment The input gate is obtained by the sigmoid function and the tanh function ; Step 4: By formula , we can obtain the memory information at time t ; Step 5: Enter Values , the output value at the previous moment and the memory information at time t The output gate is obtained by using the sigmoid function and the tanh function ; Step 6: By formula Calculate the predicted output value at time t ; Step 7: Use the mutual information method to predict the output value and risk signature matrix Compare and get the calculation results , based on the results, determine whether there will be risks in the future and the type of risks that will occur.

[0017] The gas pipeline safety monitoring and early warning system of this embodiment includes distributed optical fiber sensors, potential measurement modules, and monitoring optical cameras. Through a multimodal data fusion optimization algorithm, it can identify different types of pipeline leakage influencing factors, improve the pipeline safety early warning capability and the accuracy of pipeline leakage identification, and ensure the operational safety of the gas pipeline.

[0018] The above are only implementation modes of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure made using the contents of the present invention's specification and drawings, directly or indirectly used in other related technical fields, are also within the patent protection scope of the present invention.

Claims

1. A gas pipeline safety monitoring and early warning system based on multimodal data fusion, characterized in that: Includes distributed fiber optic sensors, potential measurement modules, and monitoring optical cameras; The distributed optical fiber sensor is arranged on the outer wall of the pipeline, and a distributed optical fiber sensor for measuring three kinds of data, namely strain, temperature and vibration, is arranged at the same time. The data measured by the distributed optical fiber sensor is transmitted to the data control center through a demodulator; The potential measurement module includes a voltmeter, a communication module, a battery module and a positioning module, and the stray current measured by the voltmeter is transmitted to the data control center through the communication module; The monitoring optical camera is deployed above the key area to continuously photograph the ground conditions above the pipeline 24 hours a day and transmit the measured data to the data control center.

2. The gas pipeline safety monitoring and early warning system based on multimodal data fusion according to claim 1, characterized in that: The monitoring method includes the following steps: Step 1: Obtain a large number of training samples of strain, temperature, vibration and potential of the potential measurement module of the distributed optical fiber sensor under the influence of human factors and natural factors from the Internet or experiments; the human factors include but are not limited to mechanical overload, improper construction, and human damage; the natural factors include but are not limited to pipeline corrosion, settlement, ground fissures, landslides, and mud-rock flows; Step 2: Classify and store the training samples according to different types of human factors and natural factors; the following data analysis process takes mechanical overload as an example, and the data processing methods of other influencing factors are the same; Step 3: Preprocess the data of training samples under mechanical overload; the training sample data of strain, temperature, vibration, and potential are text data, and Pandas is used to preprocess the noise, anomalies, missing, and duplication in the training sample data; specifically, taking strain as an example, the global maximum value in the training sample data of each strain is taken out and set as Value, from Take 100 data points forward and backward respectively and record them as , which is the training sample data of n strains; similarly, the temperature takes the global minimum value, the vibration and potential take the global maximum value, and the training sample data of temperature, vibration and potential are processed to obtain the training sample data of n temperatures ; n vibration training sample data ; Training sample data of n potentials ; The training sample data of strain, temperature, vibration and potential are spliced ​​together in order to form n 201 rows and 4 columns of matrices Training samples of arrays , recorded as ; Substitute n training sample matrices in turn, and obtain a new matrix with 201 rows and 4 columns through the training of MLP layer and Softmax layer , recorded as , which is the characteristic matrix of strain, temperature, vibration, and potential under mechanical overload; by the same method, the characteristic matrix of strain, temperature, vibration, and potential under other influencing factors i can be obtained ； Step 4: Download a large number of pictures of vehicles, people, subsidence, ground fissures, landslides, and mudslides from the Internet. Through sample training of the MLP layer and the Softmax layer, the computer can correctly identify people, various vehicles, and ground disaster change information to obtain the image recognition feature matrix ,in, is the character information feature data, is the vehicle information feature data, It is the ground information characteristic data; Step 5: The characteristic matrix of strain, temperature, vibration and potential under various influencing factors And the image recognition feature matrix Splice them separately to form a new risk feature identification matrix ； Step 6: During the operation of the gas pipeline safety monitoring and early warning system, text data of strain, temperature, vibration, potential and image data of the optical camera are collected at the same time. The training sample method is used to obtain data at certain time intervals and stitch them into a real-time data matrix ； Step 7: Using the mutual information principle, calculate the risk characteristic matrix of the gas pipeline safety monitoring and early warning system and real-time data matrix of gas pipeline safety monitoring and early warning system The correlation between them is calculated as ;in, for and The joint probability density function between ; and Are marginal probability density functions, corresponding to and ; Step 8: Based on the calculation results Analyze and determine whether there are risks in the gas pipeline and the risk type.

3. The gas pipeline safety monitoring and early warning system based on multimodal data fusion as claimed in claim 2, characterized in that: The early warning method includes the following steps: Step 1: Obtain the current data matrix of the gas pipeline safety monitoring and early warning system , that is, in the LSTM method Input value And the output value at the previous moment ; Step 2: Enter the value And the output value at the previous moment Obtaining the forget gate through the sigmoid function ,in, is the weight coefficient, It is a deviation; Step 3: Enter the value And the output value at the previous moment The input gate is obtained by the sigmoid function and the tanh function ; Step 4: By formula , we can obtain the memory information at time t ; Step 5: Enter Values , the output value at the previous moment and the memory information at time t The output gate is obtained by using the sigmoid function and the tanh function ; Step 6: By formula Calculate the predicted output value at time t ; Step 7: Use the mutual information method to predict the output value and risk signature matrix Compare and get the calculation results , based on the results, determine whether there will be risks in the future and the type of risks that will occur.

Images