A natural gas pipeline leakage intelligent detection method, device, medium and product

Abstract

The application discloses a natural gas pipeline leakage intelligent detection method and device, medium and product, relates to the natural gas pipeline detection field, and the method comprises the steps of acquiring multi-modal data of a natural gas pipeline; preprocessing the multi-modal data to obtain preprocessed data; inputting the preprocessed data into a high-dimensional feature extraction network to obtain high-dimensional leakage feature information; and inputting the high-dimensional leakage feature information into a classification network to obtain a classification prediction result of a natural gas pipeline leakage grade. The application can complete self-prediction of pipeline segment leakage conditions and realize real-time intelligent detection of the natural gas pipeline leakage grade.

Description

Technical Field

[0001] This application relates to the field of natural gas pipeline inspection, and in particular to an intelligent method, equipment, medium, and product for detecting natural gas pipeline leaks. Background Technology

[0002] As pipeline mileage increases, the probability of gas leaks also increases. The main causes of natural gas pipeline accidents include pipeline corrosion and aging, wear and tear during on-site operations, and natural disasters such as earthquakes. Once a natural gas pipeline leaks, it not only pollutes the environment and wastes energy, but also endangers people's lives and property. Real-time monitoring of pipeline operation and timely and accurate detection of leaks are crucial for relevant departments to take rapid response measures, thereby improving pipeline transportation capacity. Therefore, developing reliable and rapid detection methods for natural gas pipeline leaks is of great significance.

[0003] Natural gas pipeline leaks cause changes in pipeline operating parameters (sound waves, temperature, pressure, flow rate, etc.). Currently, data-driven natural gas pipeline leak detection can be broadly categorized into: negative pressure wave method, acoustic emission leak detection method, pressure point analysis method, and mass / volume balance method. While these methods have achieved good results in pipeline leak detection, they mostly involve collecting large amounts of data from the field, performing offline analysis, and then implementing leak detection on computers or servers. This approach suffers from problems such as complex model structures, poor noise resistance, and the inability to achieve real-time leak detection. Furthermore, most of these detection methods use a single set of data for pipeline leak state analysis, failing to comprehensively analyze the multimodal parameters of pipeline operating status. Additionally, for long-distance natural gas pipelines, the system typically consists of many pipeline segments of varying sizes and shapes, requiring the deployment of numerous sensors to monitor the health status of different segments. Real-time detection at the edge is hampered by hardware resource constraints. Offline detection methods, on the other hand, require cloud servers to receive and process massive amounts of data over long periods, posing significant challenges to data transmission and processing. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, medium and product for intelligent detection of natural gas pipeline leaks, which aims to intelligently process multimodal data at the edge and predict the leakage situation of pipeline sections.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] Firstly, this application provides an intelligent detection method for natural gas pipeline leaks, including:

[0007] Acquire multimodal data of natural gas pipelines; the multimodal data includes one or more of temperature data, pressure wave data, and acoustic vibration data from different detection nodes;

[0008] The multimodal data is preprocessed to obtain preprocessed data;

[0009] The preprocessed data is input into a high-dimensional feature extraction network to obtain high-dimensional leakage feature information;

[0010] The high-dimensional leakage feature information is input into a classification network to obtain the classification prediction results of the natural gas pipeline leakage level.

[0011] Optionally, the multimodal data is preprocessed to obtain preprocessed data, specifically including:

[0012] The multimodal data is normalized to obtain normalized data;

[0013] The normalized data is divided into frames using a sliding window and reconstructed into samples of a set length.

[0014] The sample of a set length is processed using a Hamming window to obtain the processing result;

[0015] The preprocessed data is obtained by performing a short-time Fourier transform on the processing result.

[0016] Optionally, the high-dimensional feature extraction network consists of two convolutional modules; the output of the first convolutional module serves as the input of the second convolutional module.

[0017] Optionally, each of the convolutional modules includes: a depthwise convolutional layer, a first batch normalization layer, a first ReLU activation layer, a pointwise convolutional layer, a second batch normalization layer, a second ReLU activation layer, a feature fusion layer, and an average pooling layer.

[0018] In each convolutional module, the input data is split into two paths. One path of input data passes sequentially through the depthwise convolutional layer, the first batch processing normalization layer, and the first ReLU activation layer to obtain the spatial features of the single-modal data. The spatial features of the single-modal data then pass sequentially through the pointwise convolutional layer, the second batch processing normalization layer, and the second ReLU activation layer to obtain the channel features of the multimodal data. The other path of input data and the channel features of the multimodal data are fused together in the feature fusion layer. The fused features are then pooled by the average pooling layer to obtain the output.

[0019] Optionally, the classification network includes: two convolutional layers, one global average pooling layer, and one softmin activation layer.

[0020] Secondly, this application provides an intelligent detection device for natural gas pipeline leaks, comprising:

[0021] The sensing module includes multiple sensors for collecting multimodal data from the natural gas pipeline; the multimodal data includes one or more of temperature data, pressure wave data, and acoustic vibration data from different detection nodes.

[0022] The processing module, connected to the sensing module, is used to implement the intelligent detection method for natural gas pipeline leaks provided above, so as to obtain the classification prediction result of the natural gas pipeline leak level based on the multimodal data.

[0023] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent detection method for natural gas pipeline leaks.

[0024] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described intelligent detection method for natural gas pipeline leaks.

[0025] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described intelligent detection method for natural gas pipeline leaks.

[0026] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0027] This application provides a method, device, medium, and product for intelligent detection of natural gas pipeline leaks. By preprocessing multimodal data to obtain preprocessed data, and inputting the preprocessed data into a high-dimensional feature extraction network, high-dimensional leak feature information can be obtained. By adopting a lightweight network (i.e., a classification network), the classification prediction results of the natural gas pipeline leak level can be obtained based on the high-dimensional leak feature information, and the pipeline segment leak situation can be predicted automatically, realizing real-time intelligent detection of natural gas pipeline leak levels. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1A schematic flowchart illustrating an intelligent detection method for natural gas pipeline leaks provided in an embodiment of this application;

[0030] Figure 2 This is a schematic diagram of a high-dimensional feature extraction network structure provided in an embodiment of this application;

[0031] Figure 3 This is a schematic diagram of a classification network structure provided in an embodiment of this application;

[0032] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] In one exemplary embodiment, such as Figure 1 As shown, a smart detection method for natural gas pipeline leaks is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is described using a terminal (or edge device) as an example, including steps 100 to 103. Wherein:

[0036] Step 100: Acquire multimodal data of the natural gas pipeline. Multimodal data includes one or more of the following: temperature data, pressure wave data, and acoustic vibration data from different detection nodes.

[0037] For example, multiple sets of sensors are installed on natural gas pipelines. Each set includes one or more of the following: temperature sensors, overpressure wave sensors, and acoustic vibration sensors. Each set of sensors collects multimodal data on pipeline leaks under different operating conditions and leak levels. Natural gas leaks can be classified into three levels: no leak, minor leak, and major leak, but this classification is not exhaustive and the specific leak level can be set according to the actual situation.

[0038] Step 101: Preprocess the multimodal data to obtain preprocessed data.

[0039] For example, the acquired multimodal data is normalized, and then the time-series signal is converted into a frequency domain energy spectrum signal using a short-time Fourier transform. Specifically, the short-time Fourier transform performs frame segmentation, windowing, and Fourier transform on the multimodal data samples to convert the multimodal data into a frequency domain energy spectrum signal.

[0040] Step 102: Input the preprocessed data into the high-dimensional feature extraction network to obtain high-dimensional leakage feature information.

[0041] Step 103: Input the high-dimensional leakage feature information into the classification network to obtain the classification prediction results of the natural gas pipeline leakage level.

[0042] By implementing steps 100 to 103 above, edge computing can be performed on multimodal data to automatically predict pipeline leakage, thereby solving the problem that existing technologies cannot detect the leakage level of natural gas pipelines in real time.

[0043] In another exemplary embodiment of this application, in order to further improve the accuracy of feature extraction, it is necessary to process the collected data through preprocessing methods and provide a data foundation for high-dimensional feature extraction in step 102. The above step 101 is replaced by the following steps 1 to 4:

[0044] Step 1: Normalize the multimodal data to obtain normalized data.

[0045] For example, after acquiring multimodal data on natural gas pipeline leaks, the data for each modality is normalized and mapped to the (0,1) interval. The calculation formula is as follows:

[0046]

[0047] Where, x i For the i-th modal data, x imin Let x be the minimum value of the i-th modal data. imax The maximum value of the data in the i-th mode is 1, 2, or 3. In this embodiment, i takes the value 1, 2, or 3. This represents the normalized data for the i-th mode.

[0048] Step 2: Divide the normalized data into frames using a sliding window and reconstruct them into samples of a set length.

[0049] For example, if we choose the number of signal points in each frame (i.e., the size of the sliding window) as 'a', the number of overlapping points between adjacent frames (i.e., the sliding step size) as 'b', and the total sample length as 'N', then the number of sample frames F after framing through the sliding window is:

[0050]

[0051] Step 3: Apply Hamming window processing to the samples of the set length to obtain the processing results.

[0052] Step 4: Perform a short-time Fourier transform on the processing results to obtain preprocessed data.

[0053] In the practical application of steps 1 to 4 above, for example, 12 detection nodes are deployed on a natural gas pipeline. Each node is equipped with three types of sensors: temperature, overpressure wave, and acoustic vibration, to collect multimodal state information of the natural gas pipeline. Pipeline leakage is divided into three levels (no leakage, minor leakage, and major leakage), and 3,000 samples are collected for each level, for a total of 9,000 samples. The collected multimodal data is 300,000 × 3 × 12 (time length × number of modal channels × number of detection nodes). A sliding window of a = 2,000 and a sliding step size of b = 1,000 are selected, resulting in samples of a specified length of F = 299 frames. Hamming window processing is applied to the framed multimodal data, and then a short-time Fourier transform is performed using 1,024 sampling points. The preprocessed data is 1024 × 299 × 36 (frequency domain energy spectrum × number of frames × total number of channels).

[0054] In another exemplary embodiment of this application, in order to reduce the computational load of the network and facilitate deployment at the edge, a method is provided such as... Figure 2 The high-dimensional feature extraction network shown consists of two sets of convolutional modules.

[0055] In practical applications, the data processing procedure for high-dimensional feature extraction networks is as follows:

[0056] The input data first passes through a deep convolutional layer, a first batch normalization layer, and a first ReLU activation layer to extract spatial features of the single-modal data in the feature dimension. Then, it passes through a pointwise convolutional layer, a second batch normalization layer, and a second ReLU activation layer to aggregate all channels and extract channel features of the multimodal data. Finally, the channel features of the input data and the multimodal data are concatenated along the feature channels and passed through an average pooling layer to obtain the final output.

[0057] Compared to the commonly used 3×3 convolution, using depthwise convolution and pointwise convolution can reduce the computational cost of the network, which is beneficial for deployment at the edge. The specific structural parameters of the high-dimensional feature extraction network in this embodiment are shown in Table 1.

[0058] Table 1. Structure parameters of the high-dimensional feature extraction network

[0059]

[0060] All programs in this application were developed using Python on the TensorFlow and Keras frameworks, taking into account different hardware and providing compatibility and flexibility for edge computing. Preprocessed data is input into the high-dimensional feature extraction network, and the samples are shuffled using the shuffle-batch method. The high-dimensional feature extraction network is initialized with a learning rate of 0.01, a training batch size of 64, and 300 iterations. During training, the Adam optimization algorithm is used to reduce the loss function value, ultimately obtaining the trained high-dimensional feature extraction network and its parameter configuration. In actual training, the multimodal data obtained in step 100 can be used as training samples input into the high-dimensional feature extraction network, and the parameter configuration of the feature extraction network can be obtained through multiple iterations.

[0061] In another exemplary embodiment of this application, to improve prediction accuracy, a lightweight convolutional neural network and a mapping relationship between multimodal data and natural gas pipeline leakage levels can be constructed. Based on this, this application can construct a network as follows: Figure 3 The classification network shown includes: 2 convolutional layers, 1 global average pooling layer, and 1 softmin activation layer.

[0062] The high-dimensional leakage characteristics of natural gas pipelines are processed through two convolutional layers, one global average pooling layer, and a softmin activation layer to obtain the classification and prediction results of the natural gas pipeline leakage level, corresponding to several leakage levels (no leakage, small leakage, large leakage). Specifically:

[0063] First, the classification network uses two 1×1 convolutional layers for channel dimensionality reduction. Then, global average pooling is used. Finally, a softmin activation layer is used to convert the output data in the global average pooling layer into probability output. The output dimension corresponds to several levels of leakage, namely no leakage, small leakage, and large leakage. The minimum probability term is found to complete the classification of natural gas pipeline leakage levels.

[0064] From the perspective of network lightweighting, this application employs 1×1 convolution and global average pooling to reduce network parameters and enhance network generalization ability. Furthermore, since leakage is a low-probability event, a softmin activation function is used to give higher probabilities to categories with lower scores. The specific structural parameters of the natural gas pipeline leakage level classification network in this embodiment are shown in Table 2.

[0065] Table 2 Classification Network Structure Parameters

[0066]

[0067]

[0068] In practical applications, the network model parameters in Tables 1 and 2 can be deployed to embedded devices with limited storage and computing capabilities (such as ARM processors at the sensor end) to perform high-dimensional feature extraction and leakage level classification on the multimodal data streams collected in real time from the sensors. The output result is the prediction result of the natural gas pipeline leakage level, so as to achieve the purpose of intelligent processing and prediction of multimodal data at the edge.

[0069] Based on the above description, this application designs a high-dimensional feature extraction network based on deep learning and a classification network for natural gas pipeline leakage levels for the acquired multimodal data. Specifically: First, multimodal data at different leakage levels can be acquired by deploying multiple sets of sensors on the natural gas pipeline. Second, the multimodal data is preprocessed and used as input to the high-dimensional feature extraction network. High-dimensional leakage feature information is extracted from the multimodal data through depthwise convolution and pointwise convolution, and this high-dimensional leakage feature information is input into the classification network to output the leakage level of the natural gas pipeline. Finally, the parameters of the high-dimensional feature extraction network and the classification network can be deployed at the edge to achieve intelligent processing and prediction of multimodal data at the edge. The edge can be a portable hardware platform such as an ARM processor or FPGA, which is directly deployed at or near the natural gas pipeline site to process the collected data directly without data transmission to a server.

[0070] Based on the same inventive concept, this application also provides a natural gas pipeline leak intelligent detection device for implementing the above-mentioned intelligent detection method for natural gas pipeline leaks. The solution provided by this device is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more embodiments of the intelligent detection device for natural gas pipeline leaks provided below can be found in the limitations of the intelligent detection method for natural gas pipeline leaks described above, and will not be repeated here.

[0071] In one exemplary embodiment, a smart detection device for natural gas pipeline leaks is provided, comprising a sensing module and a processing module. The processing module is connected to the sensing module.

[0072] The sensing module includes multiple sensors (such as one or more of temperature sensors, overpressure wave sensors, and acoustic vibration sensors) for acquiring multimodal data from the natural gas pipeline. The multimodal data includes one or more of temperature data, pressure wave data, and acoustic vibration data from different detection nodes.

[0073] The processing module is used to implement the intelligent detection method for natural gas pipeline leakage provided in this application to obtain classification and prediction results of natural gas pipeline leakage level based on multimodal data.

[0074] As an optional implementation, an ARM processor can be used as the processing module. In practical applications, the parameters of the trained network model can be deployed on the ARM processor, and edge computing can be used to process the multimodal data collected in real time by various sensors to output the leakage level prediction results of natural gas pipelines.

[0075] This application utilizes multiple sets of temperature, overpressure wave, and acoustic vibration sensors deployed along natural gas pipelines to collect multimodal data. By normalizing and performing short-time Fourier transform, the time-series signals are converted into frequency-domain energy spectrum signals. A lightweight convolutional neural network is then constructed to establish a mapping relationship between the multimodal data and the leakage level of the natural gas pipeline, thereby improving the accuracy of natural gas pipeline leakage prediction. Simultaneously, the trained network model parameters are deployed on an ARM processor at the sensor end, employing edge computing to process the real-time multimodal data collected by the sensors. This enables the network to autonomously predict pipeline leakage conditions, eliminating the need for redundant raw data to be sent to the server and instead reporting only the pipeline's health status. This achieves the goal of real-time detection of natural gas pipeline leakage levels.

[0076] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores natural gas pipeline leak level classification and prediction data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an intelligent detection method for natural gas pipeline leaks.

[0077] Those skilled in the art will understand that Figure 4The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0078] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0079] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0080] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0081] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0082] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0083] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A smart detection method for natural gas pipeline leaks, characterized in that, The intelligent detection method for natural gas pipeline leaks includes: Acquire multimodal data of natural gas pipelines; the multimodal data includes one or more of temperature data, pressure wave data, and acoustic vibration data from different detection nodes; The multimodal data is preprocessed to obtain preprocessed data, including: normalizing the multimodal data to obtain normalized data; dividing the normalized data into frames using a sliding window and reconstructing them into samples of a set length; performing Hamming window processing on the samples of the set length to obtain a processing result; and performing a short-time Fourier transform on the processing result to obtain the preprocessed data. The preprocessed data is input into a high-dimensional feature extraction network to obtain high-dimensional leaked feature information. The high-dimensional feature extraction network consists of two convolutional modules. The output of the first convolutional module is used as the input of the second convolutional module. Each convolutional module includes: a depthwise convolutional layer, a first batch normalization layer, a first ReLU activation layer, a pointwise convolutional layer, a second batch normalization layer, a second ReLU activation layer, a feature fusion layer, and an average pooling layer. In each convolutional module, the input data is divided into two paths. One path of input data passes through the depthwise convolutional layer, the first batch processing normalization layer, and the first ReLU activation layer in sequence to obtain the spatial features of the single-modal data. The spatial features of the single-modal data then pass through the pointwise convolutional layer, the second batch processing normalization layer, and the second ReLU activation layer in sequence to obtain the channel features of the multimodal data. The other path of input data and the channel features of the multimodal data are fused together in the feature fusion layer. The fused features are then pooled by the average pooling layer to obtain the output. The high-dimensional leakage feature information is input into the classification network to obtain the classification prediction result of the natural gas pipeline leakage level; the classification network includes: 2 convolutional layers, 1 global average pooling layer and 1 softmin activation layer.

2. A smart detection device for natural gas pipeline leaks, characterized in that, The intelligent natural gas pipeline leak detection equipment includes: The sensing module includes multiple sensors for collecting multimodal data from the natural gas pipeline; the multimodal data includes one or more of temperature data, pressure wave data, and acoustic vibration data from different detection nodes. The processing module, connected to the sensing module, is used to implement the intelligent detection method for natural gas pipeline leaks as described in claim 1, so as to obtain the classification prediction result of the natural gas pipeline leak level based on the multimodal data.

3. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the intelligent detection method for natural gas pipeline leaks as described in claim 1.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent detection method for natural gas pipeline leaks as described in claim 1.

5. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent detection method for natural gas pipeline leaks as described in claim 1.

Images