Natural gas pipeline leakage risk monitoring and early warning system and method based on DVS

Abstract

The invention discloses a natural gas pipeline leakage risk monitoring and early warning system and method based on DVS, the monitoring and early warning system is composed of hardware and software, the hardware comprises an upper computer, distributed optical fiber vibration sensing DVS equipment and a sensing optical cable, the software comprises a distributed optical fiber vibration sensing system, a natural gas pipeline leakage identification system and a natural gas pipeline leakage risk monitoring and early warning system; the distributed optical fiber vibration sensing DVS equipment automatically sends one-dimensional vibration data acquired by the sensing optical cable to the upper computer according to a sampling interval through Websocket communication; and the upper computer analyzes the received one-dimensional vibration data through the software part, identifies, judges and visually displays whether the state of the natural gas pipeline is abnormal or not, and outputs a final suspected abnormal result. By integrating hardware equipment and a software algorithm, the intelligent level is high, and real-time monitoring and accurate early warning of the leakage risk of the natural gas pipeline can be achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural gas pipeline leakage monitoring and early warning, and particularly relates to a natural gas pipeline leakage risk monitoring and early warning system and method based on DVS. Background Art

[0002] As a key safety guarantee technology, the natural gas pipeline leakage risk monitoring and early warning system has achieved certain results in practical applications, but there are still some problems to be solved urgently, mainly including imperfect systems, high false alarm rates, inaccurate positioning, etc.: 1. Large system limitations: Some monitoring technologies, such as leakage monitoring technologies based on simulation, do not require new equipment to be added to the pipeline or the existing pressure gauge distribution to be modified, but there are large limitations. For example, once the pipeline is rerouted or a branch line is added, the core software must be adjusted extensively, otherwise there will be false alarms and missed alarms. In addition, the system may have a risk of missed detection when facing small-scale leaks, especially when the leakage volume of the pipeline is small or at the initial stage of leakage, it may be difficult to detect the leakage event in time, resulting in a response delay.

[0003] 2. High false alarm rate: In the aspect of third-party construction damage monitoring, due to the lack of a unified DVS acoustic wave data sample in the industry and the complex background noise around the pipeline, the intelligent recognition algorithm of DVS has a high false alarm rate during application. False alarms of the optical fiber early warning system mainly occur in the area where the optical cable crosses the road. The continuous driving of vehicles causes ground vibrations, triggering the system alarm. In addition, false alarms of the optical fiber also occur when villagers around the pipeline use water pumps to irrigate the land.

[0004] 3. Inaccurate positioning: When the monitoring technology based on optical fiber strain (BOTDA) is used to monitor the leakage of natural gas pipelines, as the diameter of the natural gas pipeline increases and the distance between the detection optical fiber and the leakage point is long, it is necessary to study how the optical fiber effectively transmits the temperature field, transmission time, and how to accurately judge the change of the temperature field formed by the optical cable. Multiple optical cables may be required to complete the monitoring, and the equipment cost will further increase. In addition, the heat transfer effects vary greatly due to various geological conditions in the areas through which the pipeline passes, and the possibilities of missed alarms and delayed alarms are relatively large.

[0005] Therefore, it is necessary to further optimize and improve the existing natural gas pipeline leakage risk monitoring and early warning system to improve the perfection of the system, reduce the false alarm rate, and improve the positioning accuracy, so as to better ensure the safe operation of natural gas pipelines. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a natural gas pipeline leakage risk monitoring and early warning system and method based on DVS, which can realize real-time monitoring and early warning of the natural gas pipeline leakage risk by integrating advanced hardware devices and software algorithms.

[0007] To achieve the above object, the present invention adopts the following technical solutions.

[0008] First of all, the present invention provides a natural gas pipeline leakage risk monitoring and early warning system based on DVS. The monitoring and early warning system consists of two parts: hardware and software. The hardware includes a host computer, a distributed optical fiber vibration sensing DVS device, and a sensing optical cable. The distributed optical fiber vibration sensing DVS device establishes a real-time connection with the host computer through Websocket communication. The distributed optical fiber vibration sensing DVS device is communicatively connected to the sensing optical cable. The sensing optical cable is laid in the same trench as the natural gas pipeline. The sensing optical cable is used to collect the temperature, strain, and vibration information of any point along the natural gas pipeline in real time. The distributed optical fiber vibration sensing DVS device is used to convert the real-time collected vibration information into one-dimensional vibration data and send it to the host computer. The host computer analyzes the received one-dimensional vibration data through the software part to identify, judge, and visually display whether the state of the natural gas pipeline is abnormal; The software includes a distributed optical fiber vibration sensing system, a natural gas pipeline leakage identification system, and a natural gas pipeline leakage risk monitoring and early warning system. The distributed optical fiber vibration sensing system includes a data receiving module and a data output module. The data receiving module receives the vibration information collected by the sensing optical cable in real time and converts it into one-dimensional vibration data, which is then sent to the data output module. The data output module converts the one-dimensional vibration data into a binary form and transmits it to the natural gas pipeline leakage risk monitoring and early warning system of the host computer. The natural gas pipeline leakage identification system includes a data analysis module, a data conversion module, and an identification module. The data analysis module is built-in with an eigenvalue screening algorithm for screening outliers in the one-dimensional vibration data. The data conversion module is used to convert the screened one-dimensional vibration data into a two-dimensional SDP image. The identification module includes a convolutional neural network CNN model and a model training unit. The convolutional neural network CNN model is used to identify, judge, and output the suspected abnormal results of the natural gas pipeline for the converted two-dimensional SDP image. The model training unit is used to train the convolutional neural network CNN model according to the preset training mode and automatically update the trained convolutional neural network CNN model to a new convolutional neural network CNN model. The natural gas pipeline leakage risk monitoring and early warning system includes a visualization module, a display module, and a data storage module. The visualization module is used to convert the output suspected abnormal results of the natural gas pipeline into specific leakage data, leakage warning SDP images, longitude and latitude information, and time domain diagrams. The display module is used to display the information converted by the visualization module. The data storage module is used to classify and store leakage data and non-leakage data.

[0009] Specifically, the network communication between the host computer and the distributed optical fiber vibration sensing DVS device is based on the TCP / IP protocol and is connected using the C / S structure. The data upload of the distributed optical fiber vibration sensing DVS device adopts an automatic sending method, that is, after the host computer issues a collection command, the distributed optical fiber vibration sensing DVS device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensing optical cable to the connected host computer at the sampling interval.

[0010] Specifically, the natural gas pipeline leakage risk monitoring and early warning system adopts a front-end and back-end separation architecture and a three-layer Web application architecture based on the B / S mode. The front end includes a client, that is, the display module, and the back end includes a server, that is, the visualization module and the data storage module. The visualization module uses Vue 2.0 and axios to achieve real-time data display and visualization of the pipeline warning status. The display module can display real-time leakage data, leakage warning SDP images, longitude and latitude information, and time domain diagrams.

[0011] Based on the above natural gas pipeline leakage risk monitoring and early warning system, the present invention also provides an early warning method for the natural gas pipeline leakage risk monitoring system, including the following steps: Step S1: The distributed fiber optic vibration sensing (DVS) device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensing optical cable to the host computer at a sampling interval through Websocket communication; Step S2: The host computer analyzes and secondarily identifies the received vibration information, and outputs the final suspected abnormal result.

[0012] Specifically, the secondary identification includes two parts: eigenvalue screening and buried leakage identification of the natural gas pipeline, which are respectively implemented based on the eigenvalue screening algorithm and the buried leakage identification algorithm of the natural gas pipeline based on wavelet transform and symmetric point pattern (SDP) algorithm. The steps are as follows: Step S21: Narrow the scope of the suspected abnormal one-dimensional vibration data through the eigenvalue screening algorithm to reduce the amount of data recognized by the convolutional neural network (CNN) model; Step S22: Adopt the buried leakage identification algorithm of the natural gas pipeline based on wavelet transform and SDP algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image; The wavelet transform reduces noise by suppressing useless signals and enhancing useful signals, while maintaining the local peak characteristics of the useful signals. To ensure the adaptive characteristics of the wavelet function, the wavelet basis function is determined based on the selection method of wavelet entropy, and the wavelet entropy is calculated by quantifying the energy distribution of the wavelet coefficients at each decomposition level; The symmetric point pattern (SDP) algorithm can convert one-dimensional vibration data into a two-dimensional SDP image, and the conversion formula is as follows: ; In the above formula, is the transpose of ; is the amplitude of the i th point in the time domain signal; and are the maximum and minimum values of the time domain signal respectively; τ is the time interval factor, 1 ≤ τ ≤ 10; θ is the mirror symmetry angle, θ = 60°; ζ represents the gain of the plotting angle, ζ ≤ θ; is the polar coordinate radius of the i th point; and are the clockwise and counterclockwise rotation angles corresponding to the polar axis respectively; Step S23: Based on the convolutional neural network (CNN) model, identify the converted two-dimensional SDP image to identify whether the risk state of the natural gas pipeline is normal or abnormal; The convolutional neural network (CNN) model includes four convolutional layers, four pooling layers, two fully connected layers, and a softmax activation function. The specific formulas for the convolutional layer and the pooling layer are as follows: ; ; In the above formula, is the j -th element of the -th layer; j is the feature map of the -th convolutional region of the -th layer; is the weight matrix of the j -th layer; down is the bias; f(*) is the non-linear activation function; represents the ; In the above formula, is the value of the -th category output by the fully connected layer, and K is the total number of categories in the fully connected layer; Step S24: Based on the visualization module, display the identification result and the information related to the identification result.

[0013] The calculation steps of the eigenvalue screening algorithm described in Step S21 are as follows: Step S211: Grouping Let the set of all channels be , where m is the total number of channels; divide C into k groups, each group contains q channels, that is , where h = 1, 2,..., k; Step S212: Calculate the statistics of each channel For each channel , calculate its mean value , variance , and peak value : ; ; ; In the above formula, is the b -th data point in the channel, ais the total number of data points in the channel; is the maximum value among all data points in the channel; is the minimum value among all data points in the channel; Express the statistics of all channels as a set μ 、 and M: ; ; ; Step S213, Sorting Sort all channels in each group from largest to smallest according to the statistics, and obtain the sorted channel set: ; In the above formula, is the index sorted according to the statistics μ 、 、M; Step S214, Select suspected abnormal channels Let p be the percentage number set by the early warning system, and select the top p% of the channels as suspected abnormal channels to obtain the set of suspected abnormal channels : ; In the above formula, is the suspected abnormal channel, = 1, 2,..., qp , qp round up to a positive integer; Step S215, Merge channels Merge the suspected abnormal channels selected according to the calculated statistics in all groups into a list to obtain the merged channel set A; ; Step S216, Remove duplicates Remove the duplicate channel numbers in the merged list to obtain the de-duplicated channel set : ; Step S217, Return the result Return as a list containing all suspected abnormal channels.

[0014] Furthermore, the natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetric point pattern SDP algorithm has three working modes: a. Real-time identification mode, i.e., the secondary identification process; b. Add / Read data mode Data addition mode: While collecting real-time one-dimensional vibration data, add known leakage sample data, and replace the risk status of the corresponding one-dimensional vibration data collected in real time with abnormal; Data reading mode: Capture the pipeline leakage signal intensity in real time, i.e., the amplitude value of the leakage fluctuation, and accurately segment the leakage and non-leakage data according to the leakage start / end time; c. Training mode Apply the known leakage sample data multiple times to the natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetric point pattern (SDP) algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image for specific classification learning, thereby optimizing the convolutional neural network (CNN) model, and automatically update the trained convolutional neural network (CNN) model as a new identification model.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Integrated data intensification and visualization: The system structure is simple and reasonable, without the need to add additional hardware devices, reducing the construction and maintenance costs. Through the integration and optimization of software algorithms, the intensification of functions is achieved, and at the same time, the visualization degree of data is improved, enabling operators to quickly and intuitively grasp the safety status of the pipeline, improving the usability and operation convenience of the system, enhancing the intuitiveness and comprehensibility of the monitoring results, and making the monitoring work more efficient and accurate.

[0016] 2. Improvement of real-time performance and accuracy: The system integrates advanced algorithms such as Websocket communication technology, eigenvalue screening algorithm, natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetric point pattern (SDP), and convolutional neural network (CNN) model. Through eigenvalue screening and real-time interactive conversion of SDP images, the CNN model is used for identification, and the results are displayed on the interface through visualization methods, realizing real-time monitoring and early warning of natural gas pipeline leakage, improving the real-time performance and accuracy of leakage monitoring, and being able to respond to potential leakage risks faster. 3. High level of intelligence: The training of the convolutional neural network (CNN) model in the training mode enables the system to continuously learn and optimize, improving the accuracy of leakage identification, and automatically updating the model to adapt to new data features, enhancing the intelligence level of the system. Description of the drawings

[0017] Figure 1 It is a schematic diagram of the natural gas pipeline leakage risk monitoring and early warning system based on DVS of the present invention.

[0018] Figure 2 It is a flowchart of the natural gas pipeline leakage risk monitoring and early warning method based on DVS of the present invention.

[0019] Figure 3 This is a flowchart for eigenvalue screening and buried leakage identification of natural gas pipelines in the present invention; Figure 4 This is a schematic structural diagram of the convolutional neural network CNN model of the present invention; Figure 5 This is the SDP image generated in the embodiment of the present invention; Figure 6 These are the time domain graph and frequency domain graph in the embodiment of the present invention; Figure 7 This is a schematic diagram of longitude and latitude early warning positioning in the embodiment of the present invention.

[0020] In the figure: 1, natural gas pipeline; 2, leakage hole; 3, communication optical fiber bundle tube; 4, communication optical fiber; 5, host computer; 6, DVS device; 7, gas gathering station. Specific implementation mode

[0021] To facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the following will detail each step of the method proposed by the present invention. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0022] Embodiment As Figure 1 shown, the present invention discloses a DVS-based natural gas pipeline leakage risk monitoring and early warning system. The monitoring and early warning system consists of two parts: hardware and software. The hardware includes a host computer, a distributed fiber optic vibration sensing DVS device, and a sensing optical cable. The distributed fiber optic vibration sensing DVS device establishes a real-time connection with the host computer through Websocket communication. The distributed fiber optic vibration sensing DVS device is communicatively connected to the sensing optical cable, and the sensing optical cable is laid in the same trench as the natural gas pipeline. The sensing optical cable is used to collect the temperature, strain, and vibration information of any point along the natural gas pipeline in real time. The distributed fiber optic vibration sensing DVS device is used to convert the real-time collected vibration information into one-dimensional vibration data and send it to the host computer. The host computer analyzes the received one-dimensional vibration data through the software part to identify, judge, and visually display whether the state of the natural gas pipeline is abnormal; The software includes a distributed optical fiber vibration sensing system, a natural gas pipeline leakage identification system, and a natural gas pipeline leakage risk monitoring and early warning system. The distributed optical fiber vibration sensing system includes a data receiving module and a data output module. The data receiving module receives the vibration information collected by the sensing optical cable in real time and converts it into one-dimensional vibration data, which is then sent to the data output module. The data output module converts the one-dimensional vibration data into a binary form and transmits it to the natural gas pipeline leakage risk monitoring and early warning system of the host computer. The natural gas pipeline leakage identification system includes a data analysis module, a data conversion module, and an identification module. The data analysis module is built-in with an eigenvalue screening algorithm for screening outliers in the one-dimensional vibration data. The data conversion module is used to convert the screened one-dimensional vibration data into a two-dimensional SDP image. The identification module includes a convolutional neural network CNN model and a model training unit. The convolutional neural network CNN model is used to identify, judge, and output the suspected abnormal results of the natural gas pipeline for the converted two-dimensional SDP image. The model training unit is used to train the convolutional neural network CNN model according to the preset training mode and automatically update the trained convolutional neural network CNN model to a new convolutional neural network CNN model. The natural gas pipeline leakage risk monitoring and early warning system includes a visualization module, a display module, and a data storage module. The visualization module is used to convert the output suspected abnormal results of the natural gas pipeline into specific leakage data, leakage warning SDP images, longitude and latitude information, and time domain diagrams. The display module is used to display the information converted by the visualization module. The data storage module is used to classify and store leakage data and non-leakage data.

[0023] Specifically, the network communication between the host computer and the distributed optical fiber vibration sensing DVS device is based on the TCP / IP protocol and is connected using the C / S structure. The data upload of the distributed optical fiber vibration sensing DVS device adopts an automatic sending method, that is, after the host computer issues a collection command, the distributed optical fiber vibration sensing DVS device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensing optical cable to the connected host computer at the sampling interval.

[0024] Specifically, the natural gas pipeline leakage risk monitoring and early warning system adopts a front-end and back-end separation architecture and a three-layer Web application architecture based on the B / S mode. The front end includes the client, that is, the display module, and the back end includes the server, that is, the visualization module and the data storage module. The visualization module uses Vue 2.0 and axios to achieve real-time data display and visualization of the pipeline warning status. The display module can display real-time leakage data, leakage warning SDP images, longitude and latitude information, and time domain diagrams.

[0025] On the basis of the above natural gas pipeline leakage risk monitoring and early warning system, as Figure 2As shown, in this embodiment, an early warning method for the natural gas pipeline leakage risk monitoring system is also given, including the following steps: Step S1: The distributed fiber optic vibration sensing (DVS) device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensing optical cable to the upper computer at the sampling interval through Websocket communication; Step S2: The upper computer analyzes and performs secondary identification on the received vibration information, and outputs the final suspected abnormal result.

[0026] Specifically, as Figure 3 shown, the secondary identification includes two parts: eigenvalue screening and buried leakage identification of the natural gas pipeline, which are respectively implemented based on the eigenvalue screening algorithm and the buried leakage identification algorithm of the natural gas pipeline based on wavelet transform and symmetric point pattern (SDP) algorithm. The steps are as follows: Step S21: Narrow the scope of the suspected abnormal one-dimensional vibration data through the eigenvalue screening algorithm, and reduce the amount of data recognized by the convolutional neural network (CNN) model; Step S22: Adopt the buried leakage identification algorithm of the natural gas pipeline based on wavelet transform and symmetric point pattern (SDP) algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image; The wavelet transform reduces noise by suppressing useless signals and enhancing useful signals, while maintaining the local peak characteristics of the useful signals. To ensure the adaptive characteristics of the wavelet function, the wavelet basis function is determined based on the selection method of wavelet entropy, and the wavelet entropy is calculated by quantifying the energy distribution of the wavelet coefficients at each decomposition level; The symmetric point pattern (SDP) algorithm can convert the one-dimensional time-domain waveform of the vibration signal into a polar coordinate snowflake image, and the conversion formula is as follows: ; In the above formula, is the transpose of ; is the amplitude of the i th point in the time-domain signal; and are respectively the maximum and minimum values of the time-domain signal; τ is the time interval factor, 1 ≤ τ ≤ 10; θ is the mirror symmetry angle, θ = 60°; ζ represents the gain of the plotting angle, ζ ≤ θ; is the polar coordinate radius of the i th point; and are respectively the clockwise and counterclockwise rotation angles corresponding to the polar axis; Step S23: Based on the convolutional neural network (CNN) model, identify the converted two-dimensional SDP image to identify whether the risk state of the natural gas pipeline is normal or abnormal; As Figure 4As shown, the convolutional neural network (CNN) model includes four convolutional layers, four pooling layers, two fully connected layers, and a softmax activation function. The specific formulas for the convolutional layer and the pooling layer are as follows: ; ; In the above formula, is the j-th element of the l-th layer; is the feature map of the (l - 1)-th layer in the j-th convolutional region; is the weight matrix of the l-th layer; is the bias; f (*) is the non-linear activation function; represents the j-th feature map of the l-th layer of the weight; down (*) is the downsampling function; After multiple convolutional and pooling operations, the fully connected layer converts the multi-dimensional feature representation into a one-dimensional vector; the activation function of the last layer of the fully connected layer is softmax, and the mathematical expression of the softmax activation function is as follows: ; In the above formula, is the value of the n-th category output by the fully connected layer, and K is the total number of categories in the fully connected layer; is the predicted probability of the n-th category; Step S24: Based on the visualization module, display the identification result and the information related to the identification result.

[0027] Specifically, the calculation steps of the eigenvalue screening algorithm in step S21 are as follows: Step S211: Grouping Let the set of all channels be , where m is the total number of channels; divide C into k groups, and each group contains q channels, that is , where h = 1, 2,..., k; Step S212: Calculate the statistics of each channel For each channel , calculate its average value , variance , and peak value : ; ; ; In the above formula, is the b -th data point in the channel; a is the total number of data points in the channel; is the maximum value among all data points in the channel; is the minimum value among all data points in the channel; Express the statistics of all channels as a set μ 、 and M: ; ; ; Step S213, Sorting Sort all channels of each group from largest to smallest according to the statistics to obtain the sorted channel set: ; In the above formula, is the index sorted according to the statistics μ 、 and M; Step S214, Select suspected abnormal channels Let p be the percentage number set by the warning system, and select the top p% of the channels as suspected abnormal channels to obtain the set of suspected abnormal channels : ; In the above formula, is the suspected abnormal channel, = 1, 2,..., qp , qp Round up to a positive integer; Step S215, Merge channels Merge the suspected abnormal channels selected according to the calculated statistics in all groups into a list to obtain the merged channel set A; ; Step S216, Remove duplicates Remove the duplicate channel numbers in the merged list to obtain the set of channels after removing duplicates : ; Step S217, Return the result Return as a list containing all suspected abnormal channels.

[0028] Furthermore, the natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetric point pattern SDP algorithm has three working modes: a. Real-time identification mode, i.e., the secondary identification process; b. Add / Read data mode Add data mode: While collecting real-time one-dimensional vibration data, add known leakage sample data, and replace the risk status of the corresponding one-dimensional vibration data collected in real-time with abnormal; Read data mode: Capture the signal intensity of pipeline leakage in real-time, that is, the amplitude value of leakage fluctuations, and accurately divide the leakage and non-leakage data according to the leakage start / end time; c. Training mode Apply the natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetric point pattern SDP algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image for specific classification learning multiple times for the known leakage sample data, and then optimize the convolutional neural network CNN model, and automatically update the trained convolutional neural network CNN model to a new identification model.

[0029] Working process: Start the distributed optical fiber vibration sensing DVS device and the host computer. The distributed optical fiber vibration sensing DVS device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected in real-time to the host computer. The natural gas pipeline leakage risk monitoring and early warning system of the host computer analyzes the DVS data, and then performs a secondary identification algorithm: first reduce the amount of data recognized by the convolutional neural network CNN model through the eigenvalue screening algorithm, and then use the symmetric point pattern SDP algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image. The converted two-dimensional SDP image is as Figure 5 shown. Use the convolutional neural network CNN model to identify the two-dimensional SDP image to determine whether the pipeline state is abnormal. Finally, visualize the recognition results and information on the interface of the natural gas pipeline leakage risk monitoring and early warning system, as Figure 6 and Figure 7 shown, which are the time-domain diagram, frequency-domain diagram, and longitude and latitude warning positioning of the leakage location shown on the interface of the natural gas pipeline leakage risk monitoring and early warning system respectively. Through the visualization method, the monitoring results and useful information in the recognition process are intuitively displayed on the interface, enabling the operator to quickly master the safety state of the pipeline. By optimizing and updating the convolutional neural network CNN model through the training mode, the system can continuously learn and adapt to new data features, effectively prevent and respond to leakage events in a timely manner, and ensure the safe operation of the pipeline.

[0030] The above is only a preferred embodiment of the present invention, and it is not a limitation of the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still belong to the protection scope of the technical solution of the present invention.

Claims

1. A natural gas pipeline leakage risk monitoring and early warning system based on DVS, characterized in that: The monitoring and early warning system consists of two parts: hardware and software. The hardware includes a host computer, a distributed optical fiber vibration sensor DVS device and a sensor optical cable. The distributed optical fiber vibration sensor DVS device establishes a real-time connection with the host computer through Websocket communication. The distributed optical fiber vibration sensor DVS device is connected to the sensor optical cable for communication. The sensor optical cable is laid in the same trench as the natural gas pipeline. The sensor optical cable is used to collect temperature, strain and vibration information at any point along the natural gas pipeline in real time. The distributed optical fiber vibration sensor DVS device is used to convert the real-time collected vibration information into one-dimensional vibration data and send it to the host computer. The host computer parses the received one-dimensional vibration data through software to identify, judge and visually display whether the natural gas pipeline status is abnormal. The software includes a distributed optical fiber vibration sensing system, a natural gas pipeline leakage identification system and a natural gas pipeline leakage risk monitoring and early warning system. The distributed optical fiber vibration sensing system includes a data receiving module and a data output module. The data receiving module receives the vibration information collected by the sensing optical cable in real time and converts it into one-dimensional vibration data and sends it to the data output module. The data output module converts the one-dimensional vibration data into binary form and transmits it to the natural gas pipeline leakage risk monitoring and early warning system of the upper computer; the natural gas pipeline leakage identification system includes a data analysis module, a data conversion module and an identification module. The data analysis module has a built-in characteristic value screening algorithm for screening outliers in the one-dimensional vibration data. The data conversion module is used to convert the screened one-dimensional vibration data into a two-dimensional SDP image. The identification module includes a convolutional neural network The invention relates to a convolutional neural network (CNN) model and a model training unit, wherein the convolutional neural network (CNN) model is used to identify, judge and output suspected abnormal results of the natural gas pipeline on the converted two-dimensional SDP image, and the model training unit is used to train the convolutional neural network (CNN) model according to a preset training mode, and automatically update the trained convolutional neural network (CNN) model to a new convolutional neural network (CNN) model; the natural gas pipeline leakage risk monitoring and early warning system comprises a visualization module, a display module and a data storage module, wherein the visualization module is used to convert the output suspected abnormal results of the natural gas pipeline into specific leakage data, leakage warning SDP image, longitude and latitude information and time domain diagram, the display module is used to display various information converted by the visualization module, and the data storage module is used to classify and store leakage data and non-leakage data.

2. According to claim 1, a natural gas pipeline leakage risk monitoring and early warning system based on DVS is characterized in that: The network communication between the host computer and the distributed optical fiber vibration sensing DVS device is based on the TCP / IP protocol and is connected using a C / S structure; the data upload of the distributed optical fiber vibration sensing DVS device is automatically sent, that is, after the host computer issues a collection command, the distributed optical fiber vibration sensing DVS device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensor cable to the connected host computer according to the sampling interval.

3. According to claim 1, a natural gas pipeline leakage risk monitoring and early warning system based on DVS is characterized in that: The natural gas pipeline leakage risk monitoring and early warning system adopts a front-end and back-end separated architecture and a three-layer Web application architecture based on the B / S mode. The front-end includes a client, i.e., a display module, and the back-end includes a server, i.e., a visualization module and a data storage module. The visualization module uses Vue 2.0 and axios to realize real-time data display and visualization of pipeline early warning status. The display module can display real-time leakage data, leakage warning SDP images, longitude and latitude information, and time domain graphs.

4. A DVS-based natural gas pipeline leakage risk monitoring and early warning method, based on the natural gas pipeline leakage risk monitoring and early warning system as claimed in any one of claims 1 to 3, characterized in that: The following steps are involved: Step S1, the distributed optical fiber vibration sensor DVS device automatically sends the temperature, strain, and vibration information of any point along the natural gas pipeline collected by the sensor optical cable to the host computer according to the sampling interval through Websocket communication; Step S2: The host computer analyzes and performs secondary identification on the received vibration information, and outputs the final suspected abnormal result.

5. A natural gas pipeline leakage risk monitoring and early warning method based on DVS according to claim 4, characterized in that: The secondary identification includes two parts: eigenvalue screening and natural gas pipeline underground leakage identification, which are implemented based on the eigenvalue screening algorithm and the natural gas pipeline underground leakage identification algorithm based on wavelet transform and symmetric point pattern SDP algorithm respectively. The steps are as follows: Step S21, narrowing the scope of suspected abnormal one-dimensional vibration data through a characteristic value screening algorithm, and reducing the amount of data recognized by the convolutional neural network CNN model; Step S22, using a natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetrical point pattern SDP algorithm to convert the one-dimensional vibration data into a two-dimensional SDP image; The wavelet transform reduces noise by suppressing useless signals and enhancing useful signals, while maintaining the local peak characteristics of useful signals. In order to ensure the adaptive characteristics of the wavelet function, the wavelet basis function is determined based on the selection method of the wavelet entropy, and the wavelet entropy is calculated by quantifying the energy distribution of the wavelet coefficients at each decomposition level; The symmetrical point pattern SDP algorithm can convert one-dimensional vibration data into a two-dimensional SDP image. The conversion formula is as follows: ； In the above formula, Yes The transpose of is the time domain signal i The amplitude of the points; and are the maximum and minimum values ​​of the time domain signal respectively; τ is the time interval factor, 1≤τ≤10; θ is the mirror symmetry angle, θ=60°; ζ represents the gain of the drawing angle, ζ≤θ; It is i The polar coordinate radius of a point; and are the clockwise and counterclockwise rotation angles corresponding to the polar axis respectively; Step S23: Recognize the converted two-dimensional SDP image based on the convolutional neural network (CNN) model to identify whether the risk status of the natural gas pipeline is normal or abnormal; The convolutional neural network CNN model includes four convolutional layers, four pooling layers, two fully connected layers and a softmax activation function, wherein the specific formulas of the convolutional layer and the pooling layer are as follows: ； ； In the above formula, is the first layer j elements; It is j The l-1th layer feature map of the convolution area; is the weight matrix of the lth layer; is bias; f (*) is a nonlinear activation function; represents the weight of the lth layer j feature maps; down (*) is the downsampling function; After multiple convolution and pooling operations, the fully connected layer converts the multi-dimensional feature representation into a one-dimensional vector; the activation function of the last layer of the fully connected layer is softmax, and the mathematical expression of the softmax activation function is as follows: ； In the above formula, is the value of the nth category output by the fully connected layer, and K is the total number of categories in the fully connected layer; is the predicted probability of the nth category; Step S24: display the recognition results and information related to the recognition results based on the visualization module.

6. A natural gas pipeline leakage risk monitoring and early warning method based on DVS according to claim 5, characterized in that: The calculation steps of the characteristic value screening algorithm in step S21 are as follows: Step S211: Grouping Let the set of all channels be , where m is the total number of channels; C is divided into k groups, each group contains q channels, that is, , where h=1,2,...,k; Step S212: Calculate the statistics of each channel For each channel , calculate its average value ,variance , and peak : ； ； ； In the above formula, It is the first b data points; a is the total number of data points in the channel; is the maximum value among all data points in the channel; is the minimum value among all data points in the channel; The statistics of all channels are expressed as a set μ , and M: ； ； ； Step S213: Sorting Sort all channels of each group from large to small according to the statistics to obtain the sorted channel set: ； In the above formula, According to statistics μ , , M sorted index; Step S214: Select suspected abnormal channel Let p be the percentage set by the early warning system, select the first p% of channels as suspected abnormal channels, and get the set of suspected abnormal channels : ； In the above formula, It is a suspected abnormal channel. =1,2,..., qp , qp Round to a positive integer; Step S215: Merge channels Merge the suspected abnormal channels selected according to the calculated statistics in all groups into a list to obtain a merged channel set A; ； Step S216: Deduplication Remove duplicate channel numbers from the merged list to get a deduplicated channel set : ； Step S217: Return result return As a list containing all suspected abnormal channels.

7. A natural gas pipeline leakage risk monitoring and early warning method based on DVS according to claim 5, characterized in that: The natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetrical point pattern SDP algorithm has three working modes: a. Real-time identification mode is the secondary identification process; b. Add / read data mode Add data mode: Add known leakage sample data while collecting real-time one-dimensional vibration data, and replace the risk status of the corresponding one-dimensional vibration data collected in real time with abnormality; Data reading mode: Real-time capture of pipeline leakage signal strength, i.e., the amplitude of leakage fluctuations, and accurate segmentation of leakage and non-leakage data based on the start / end time of the leakage; c. Training mode The known leakage sample data are repeatedly used to convert the one-dimensional vibration data into a two-dimensional SDP image for specific classification learning using the natural gas pipeline buried leakage identification algorithm based on wavelet transform and symmetrical point pattern SDP algorithm, thereby optimizing the convolutional neural network (CNN) model, and automatically updating the trained convolutional neural network (CNN) model to a new identification model.

Images