Gas extraction pipeline leakage detection method based on optical fiber sensing technology

Abstract

The invention relates to the technical field of gas extraction pipeline leakage detection, and discloses a gas extraction pipeline leakage detection method based on an optical fiber sensing technology, and the method comprises the following steps: synchronously collecting data, and constructing a standardized monitoring matrix through denoising and space-time alignment; extracting space gradient and frequency domain energy characteristics, and screening thermal and dynamic abnormal regions; a multi-dimensional feature vector is constructed, classification model mapping is input, and a leakage judgment result is generated; extracting position information, performing double-source verification on a thermal dynamic abnormal center, and generating a final leakage coordinate; and generating an alarm based on the result and the coordinates, and constructing a control instruction to drive pipeline parameter adjustment. According to the method, distributed temperature and acoustic data are synchronously collected, a standardized monitoring matrix is constructed, multi-dimensional feature fusion judgment is carried out by utilizing spatial gradient features of the temperature data and frequency domain energy features of the acoustic data, and thermal dynamic coupling feature analysis is carried out based on a support vector machine classification model; and the real leakage signal and the background environment noise can be effectively distinguished.

Description

Technical Field

[0001] This invention relates to the field of gas extraction pipeline leakage detection technology, specifically a gas extraction pipeline leakage detection method based on fiber optic sensing technology. Background Technology

[0002] Gas drainage pipelines, serving as the transport medium, often traverse intricate underground roadway networks. These pipelines are laid over long distances, and underground geological conditions and working environments are harsh. Loosening of pipeline joints, corrosion perforation, or deformation due to external forces are common occurrences. Once a leak occurs in the pipeline, it not only directly leads to a drop in drainage negative pressure and a significant reduction in drainage efficiency, but the leaked gas is also more likely to accumulate in local roadway spaces. Excessive concentration poses a significant risk of explosion. Real-time monitoring and fault diagnosis of the entire pipeline network's operational status are essential prerequisites for ensuring safe mine production.

[0003] Current detection methods primarily rely on traditional electrochemical or catalytic combustion sensors. These devices are typically deployed at discrete points along the pipeline. For long-distance pipelines, physical monitoring gaps exist between sensors. Some mines have begun introducing distributed fiber optic sensing technology. Conventional applications often employ straight-line fiber optic cable laying, sensing changes in physical quantities along the route through the principle of optical time-domain reflectometry. Current mainstream solutions focus on single-mode signal demodulation, or separately monitor temperature field distribution to detect leaks based on temperature rise anomalies, or separately monitor acoustic field vibrations to capture the acoustic signals of ejected gas. Alarm logic is often based on threshold comparisons of a single indicator.

[0004] However, existing detection technologies rely solely on temperature signals. The heat exchange process caused by gas leaks is slow, resulting in a significant lag in the thermal signal response. Relying solely on acoustic signals is also problematic, as the noise from underground fans and vehicles can easily obscure the true leak sound signature due to environmental interference. Furthermore, localization algorithms, relying on a single signal peak, are heavily influenced by multipath effects in tunnels and the isotropic nature of heat diffusion, often leading to calculated leak point coordinates that deviate from the actual location. Therefore, this invention provides a gas extraction pipeline leak detection method based on fiber optic sensing technology to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a gas extraction pipeline leakage detection method based on fiber optic sensing technology. This method solves the problems of existing detection technologies, such as high false alarm and false negative rates due to single-mode monitoring being susceptible to underground environmental noise and thermal hysteresis, low positioning accuracy due to multipath effects in roadways affecting single-source positioning algorithms, and false alarms caused by the lack of an adaptive mechanism for operating conditions during normal process adjustments.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology, comprising the following steps: S1. Simultaneously collect distributed temperature data and distributed acoustic data along the gas extraction pipeline, perform noise reduction processing on the collected data, and align and map the noise-reduced data based on a unified spatiotemporal coordinate reference to construct a standardized monitoring matrix. S2. Based on the temperature data components in the standardized monitoring matrix, spatial gradient features are generated, and simultaneously, based on the acoustic data components in the standardized monitoring matrix, frequency domain energy features are generated to screen out candidate regions for thermodynamic anomalies and candidate regions for kinetic anomalies, respectively. S3. Using the statistical parameters of the spatial gradient features and the frequency domain energy features, a multi-dimensional feature vector is constructed and input into the classification model for feature space mapping to generate a determination result indicating whether a leakage event exists at the current monitoring location. S4. In response to the judgment result indicating the existence of a leakage event, extract the spatial location information of the abnormal feature signal, use the thermodynamic anomaly center determined based on temperature data to perform spatial consistency verification on the dynamic anomaly center determined based on acoustic data, and generate the final leakage location coordinates based on the data after the verification is passed. S5. Based on the determination result and the final leak location coordinates, generate an alarm signal, and construct control instructions for the gas extraction pipeline according to the alarm signal to drive the adjustment of pipeline operating parameters.

[0007] Preferably, in step S1, the step of constructing the standardized monitoring matrix further includes: An integrated composite optical cable, laid in a spiral winding manner along the outer wall of the gas extraction pipeline, is used as a signal acquisition carrier to obtain raw optical signals with high spatial resolution. The wavelet transform algorithm is used to perform multi-scale decomposition and reconstruction on the data obtained from the original optical signal to generate a clean data sequence with random noise components removed. Using high-resolution sensor modal data as a benchmark, an interpolation algorithm is used to drive the resampling process of low-resolution modal data, mapping distributed temperature data and distributed acoustic data onto a unified spatial grid node, thereby generating a standardized monitoring matrix aligned in the spatiotemporal dimension.

[0008] Preferably, in step S1, the denoising process on the acquired data further includes an acoustic data reconstruction step based on frequency domain characteristics: Construct a digital bandpass filter whose passband frequency covers the frequency band of gas leakage airflow noise; The digital bandpass filter is used to perform filtering operations on the collected distributed acoustic data to remove mechanical vibration components below the lower limit frequency and background noise components above the upper limit frequency, generating a preprocessed acoustic data sequence that retains only the characteristic frequency band of leakage sound emission, and the preprocessed acoustic data sequence is used as the basic data for constructing the acoustic data components in the standardized monitoring matrix.

[0009] Preferably, in step S2, the screening step for the candidate thermodynamic anomaly region further includes: Based on the temperature data components in the standardized monitoring matrix, the temperature difference between adjacent spatial grid nodes is calculated, and the spatial temperature gradient value distributed along the axial direction of the gas extraction pipeline is calculated based on the temperature difference. The mean and standard deviation are calculated based on the temperature data distribution pattern of the entire monitoring optical cable, and a dynamic judgment baseline that conforms to statistical criteria is constructed. The calculated spatial temperature gradient value is compared with the dynamic judgment baseline, and regions whose gradient values ​​exceed the dynamic judgment baseline are selected and marked as candidate regions of thermodynamic anomalies.

[0010] Preferably, in step S2, the step of generating the frequency domain energy features further includes: Perform a short-time Fourier transform on the acoustic data components in the standardized monitoring matrix to calculate the acoustic power spectral density, and calculate the rate of change of the acoustic power spectral density at the current moment relative to the leak-free reference value. The temporal energy envelope and zero crossover rate of acoustic data are extracted simultaneously. The changing trends of the change ratio, energy envelope and zero crossover rate are jointly analyzed to construct a frequency domain energy feature representing the intensity of airflow turbulence. Based on the frequency domain energy feature, candidate regions of dynamic anomalies are screened out.

[0011] Preferably, step S3 further includes: The variance of the spatial gradient features within the sliding time window is calculated as the temperature dispersion parameter, and the variance of the frequency domain energy features within the sliding time window is calculated as the acoustic dispersion parameter. The temperature dispersion parameter and the acoustic dispersion parameter are combined to construct a multidimensional feature vector representing the stability of signal fluctuations. The classification model employs a support vector machine, using radial basis functions as kernel functions to map the multidimensional feature vectors to a high-dimensional feature space, and searches for the optimal classification hyperplane to separate leaked samples from normal samples, thereby outputting the judgment result containing the leakage probability value.

[0012] Preferably, in step S4, the step of generating the final leak location coordinates further includes: Using a spatial mapping model based on spiral winding laying parameters, the optical cable length coordinates are converted into duct axial length coordinates and three-dimensional spatial coordinates; The maximum temperature gradient point is searched in the candidate region of thermodynamic anomaly as the center of thermodynamic anomaly, and the peak point of acoustic power spectral density is searched in the candidate region of kinetic anomaly as the center of kinetic anomaly. The spatial distance deviation between the center of thermodynamic anomaly and the center of kinetic anomaly is calculated. The spatial distance deviation is compared with a preset positioning convergence threshold. When the spatial distance deviation is less than the positioning convergence threshold, the arithmetic mean of the thermodynamic anomaly center and the kinetic anomaly center is calculated and output as the final leak location coordinates.

[0013] Preferably, step S1 further includes a step of correcting temperature data based on operating condition parameters: Obtain the real-time negative pressure value and mixed gas flow rate value inside the pipeline; Retrieve a pre-built temperature correction model, which defines the mapping relationship between pressure change, flow rate change and pipe wall temperature correction. The real-time negative pressure value and the mixed gas flow rate value are input into the temperature correction model to generate the temperature correction amount under the current operating condition. The temperature correction amount is then used to compensate the collected distributed temperature data to generate a corrected temperature sequence that eliminates the influence of operating condition fluctuations. The corrected temperature sequence is then used to participate in the construction of the standardized monitoring matrix.

[0014] A second aspect of the present invention provides a gas extraction pipeline leakage detection system based on fiber optic sensing technology, comprising: The distributed optical fiber sensing unit includes an integrated composite optical cable laid along the outer wall of the gas extraction pipeline, which is used to construct a physical field signal acquisition link. The signal demodulation and processing unit is connected to the distributed optical fiber sensing unit and is used to perform the construction of the standardized monitoring matrix, the generation of spatial gradient features and frequency domain energy features, the output of the judgment result, and the verification calculation of the final leakage location coordinates. The alarm and linkage control unit is communicatively connected to the signal demodulation and processing unit. It is used to issue graded alarm signals in response to the judgment result and generate control commands based on the alarm signals to drive external equipment to perform pipeline operating parameter adjustments. The ground monitoring and visualization platform is used to receive and display the final leak location coordinates and pipeline status information.

[0015] A third aspect of the present invention provides an electronic 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 gas extraction pipeline leakage detection method based on fiber optic sensing technology described in any of the preceding claims.

[0016] This invention provides a method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology. It has the following advantages: 1. This invention synchronously collects distributed temperature and acoustic data and constructs a standardized monitoring matrix. It uses the spatial gradient characteristics of temperature data and the frequency domain energy characteristics of acoustic data to perform multi-dimensional feature fusion judgment. Compared with single-modal detection, the thermo-dynamic coupling feature analysis based on the support vector machine classification model can effectively distinguish between real leakage signals and background environmental noise, thereby reducing the false alarm rate and false alarm rate in complex downhole environments.

[0017] 2. This invention adopts an integrated composite optical cable spiral laying method to increase the sampling density per unit tube length and introduces a dual-source verification and positioning mechanism; by calculating the spatial consistency between the thermodynamic anomaly center and the dynamic anomaly center, and using the spatial distance deviation for convergence verification, the random error existing in single physical field positioning is effectively eliminated, ensuring that the generated final leak location coordinates have credibility and accuracy, and meeting the needs of emergency repair.

[0018] 3. This invention introduces a temperature correction model based on operating parameters, which can dynamically eliminate the influence of non-leakage temperature fluctuations caused by the adjustment of the extraction process, ensuring the purity of monitoring data. The system can automatically generate control commands based on the judgment results and positioning information, driving the real-time adjustment of the operating parameters of the gas extraction pipeline, thereby improving the intelligence level and safety of the system. Attached Figure Description

[0019] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the method steps of the present invention; Figure 3 This is a schematic diagram of the distributed optical fiber sensing unit of the present invention; Figure 4 This is a schematic diagram of the dual-source verification and positioning process of the present invention; Figure 5 This is a schematic diagram of the distributed optical fiber duct laying and winding of the present invention; Figure 6 This is a schematic diagram of the feature extraction and SVM fusion algorithm of the present invention; Figure 7 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 Appendix Figure 3 and attached Figure 5 , Figure 1 This is a system architecture diagram according to an embodiment of the present invention. The present invention provides a gas extraction pipeline leakage detection system based on fiber optic sensing technology, including a distributed fiber optic sensing unit, a signal demodulation and processing unit, a data transmission and display unit, an alarm and linkage control unit, and a ground monitoring and visualization platform.

[0022] The distributed fiber optic sensing unit, designed to adapt to the unique environment of underground gas extraction pipelines in coal mines, employs an integrated composite optical cable. Its structure comprises an outer flame-retardant sheath (polyurethane material), a reinforcing layer, a tightly packed sensing fiber core, and an acoustic coupling layer. The optical cable is laid along the outer wall of the main and branch gas extraction pipelines. To increase the effective sensing length of the optical fiber per unit pipeline length and improve the detection sensitivity to weak leakage signals, the cable is laid using a spiral winding method with a spiral pitch of 1.0m and a fixed spacing controlled at 1.0±0.2m. It is secured using a combination of 304 stainless steel clamps and flame-retardant rubber gaskets. A fiber optic reflector terminal (FBT) is installed at the cable end to form a closed-loop optical path structure, ensuring rapid location of the break point in the event of a cable fracture.

[0023] In a preferred embodiment of the present invention, fiber Bragg grating (FBG) sensing technology is introduced to monitor the health status of the pipeline structure. FBG sensors are connected in series at key stress concentration points such as valves and elbows in the pipeline to sense changes in pipe wall strain and thermal stress.

[0024] The signal demodulation and processing unit is connected to the beginning of the optical cable. The main unit adopts an explosion-proof housing design to meet the explosion-proof requirements of underground equipment. The main unit has a built-in distributed temperature sensing module (DTS), a distributed acoustic sensing module (DAS), and a fiber optic demodulation module (FBG). These three modules share or time-division multiplex the same composite optical cable to achieve synchronous acquisition and demodulation of temperature, acoustic, and stress signals, and perform signal preprocessing and feature calculations. The data transmission and display unit is connected to the signal demodulation and processing unit through a communication interface, and is used to transmit the demodulated monitoring data to the server and present it visually. The alarm and linkage control unit is equipped with a communication interface for connecting to the automatic extraction control system and sending control commands when a leak is detected. The ground monitoring and visualization platform serves as a comprehensive management carrier, integrating data storage, analysis, and human-computer interaction functions.

[0025] See attached document Figure 2 This invention provides a method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology, comprising the following steps: S1. Use composite fiber optic sensing units to synchronously acquire distributed temperature signals, distributed acoustic signals and discrete point strain signals along the pipeline, and perform noise reduction and spatiotemporal synchronization preprocessing on the acquired signals. S2. Extract the spatial gradient features of the preprocessed temperature signal and the frequency domain energy features of the acoustic signal to identify thermodynamic and kinetic anomalies. S3. Construct a feature vector containing temperature and acoustic features, and use a feature fusion recognition algorithm to analyze the feature vector to determine whether a leakage event exists. S4. Calculate the coordinates of the leak point based on the spatial index of the abnormal signal, and verify the consistency between the location result based on the temperature signal and the location result based on the acoustic signal. S5. Trigger a graded alarm based on the judgment result and adjust the negative pressure control parameters of the extraction pipeline accordingly.

[0026] The above steps will be explained in detail below with reference to specific embodiments.

[0027] In step S1, the synchronous acquisition and preprocessing of physical field signals along the gas extraction pipeline mainly includes the following steps: Using an integrated composite optical cable spirally laid along the outer wall of the gas extraction pipeline as the sensing medium, the optical cable integrates a single-mode optical fiber (G652D type) and an acoustic coupling layer. Through the splitting module in the optical fiber demodulation host, the optical signal transmitted from the same optical fiber is simultaneously introduced into the distributed temperature sensing module (DTS) and the distributed acoustic sensing module (DAS).

[0028] For temperature signal acquisition, the distributed temperature sensing module operates based on the Raman scattering principle. A laser emits pulsed light into an optical fiber; the light scatters as it propagates within the fiber. The demodulation unit collects the Stokes and anti-Stokes beams from the backscattered light. Utilizing the temperature sensitivity of the anti-Stokes beam while the Stokes beam is insensitive, the temperature data distributed along the fiber's length is demodulated. The system is set with a sampling period of 5 seconds, a temperature measurement accuracy controlled within ±0.1℃, and a spatial resolution of 1 meter.

[0029] For acoustic signal acquisition, the distributed acoustic sensing module operates based on the principle of coherent optical interference. It uses a highly coherent laser to emit probe light pulses and senses external vibrations by detecting phase changes in the Rayleigh scattered light within the optical fiber. When a leak in a pipe causes vibrations in the pipe wall or surrounding medium, it induces micro-strain in the optical fiber, thereby altering the phase of the scattered light. The demodulation unit converts these phase changes into acoustic vibration signals through interferometric demodulation, covering a detection frequency band of 10Hz-40kHz with a positioning accuracy better than ±1m.

[0030] For stress signal acquisition, the fiber optic grating demodulation module monitors the center wavelength drift of the FBG sensor using wavelength division multiplexing (WDM) technology, thereby demodulating the micro-strain data of key components.

[0031] The distributed temperature sensing module and the distributed acoustic sensing module share the same clock source to ensure the accuracy of the acquired raw temperature sequence. With the original acoustic sequence Maintain synchronization on the timeline, where Indicates the spatial location index along the optical cable. Indicates the sampling time.

[0032] Due to electromagnetic interference and thermal noise from optoelectronic devices in the downhole environment, the acquired raw temperature signals typically contain high-frequency random noise. The processing unit processes the raw temperature sequence... Wavelet denoising is performed. Specifically, a suitable wavelet basis function (such as the Daubechies wavelet) is selected for non-stationary signal analysis to decompose the temperature signal into multiple scales, obtaining approximation coefficients and detail coefficients. Random noise components are removed by soft-thresholding the high-frequency detail coefficients containing noise, and the denoised temperature signal is then reconstructed through inverse wavelet transform. The specific algorithm implementations for wavelet transform and threshold denoising are conventional signal processing techniques well-known to those skilled in the art and will not be elaborated upon here.

[0033] Furthermore, to address the zero-point drift issue that may arise during long-term operation of fiber optic sensing equipment, baseline drift correction is performed. The system calculates the temperature sliding average over a certain time window as a dynamic baseline. This dynamic baseline is subtracted from the denoised signal to eliminate slow temperature background fluctuations caused by non-leakage factors, resulting in preprocessed temperature data. .

[0034] The raw acoustic signal contains vibration information across the entire frequency range. To separate the specific acoustic features caused by the gas leak, irrelevant interference needs to be filtered out. The processing unit... A digital bandpass filter was applied. Based on the frequency domain characteristics of airflow noise generated by gas extraction pipeline leakage, the passband frequency range of the filter was set to 100Hz to 20kHz. The specific basis for this frequency band selection is as follows: signals with frequencies below 100Hz are mainly caused by mechanical vibrations such as underground motor operation and vehicle passage; signals with frequencies above 20kHz are mainly affected by high-frequency background noise such as airflow friction. Through bandpass filtering, the acoustic emission signal components generated by the leakage fluid impacting the pipe wall and turbulence were effectively preserved, resulting in preprocessed acoustic data. .

[0035] To enable subsequent multi-physical quantity fusion analysis, the preprocessed temperature data Acoustic data Perform standardized alignment in the spatiotemporal dimensions.

[0036] In terms of spatial dimension, since the spatial resolution of DTS and DAS may differ, the system uses the higher resolution as a benchmark and uses an interpolation algorithm to map the two types of data onto a unified spatial coordinate grid to ensure that the same coordinate point corresponds to the same physical location on the pipeline.

[0037] In terms of time, the two independent time series are aligned based on the unified timestamp at the time of data collection to construct a unified monitoring matrix. The monitoring matrix contains the first Temperature and acoustic conditions at all locations along the pipe at each sampling time: ; in, Represents the monitoring matrix. Represents a time series index. Indicates spatial location index, Indicates the first The physical spatial coordinates of each measuring point on the pipeline Indicates the first A unified timestamp corresponding to each sampling moment. This represents the total number of fiber optic measurement points laid along the pipeline. After the above processing, a standardized dataset was formed that is strictly location-corresponding, highly time-synchronized, and free from major environmental interferences, providing a reliable data foundation for subsequent feature extraction and fusion determination.

[0038] See attached document Figure 6 In step S2, feature extraction of the preprocessed signal mainly includes two steps: temperature field anomaly feature extraction and sound field time-frequency feature extraction. When a gas extraction pipeline leaks, the negative pressure inside the pipeline draws in ambient air. The heat exchange effect caused by the mixing of outside air and gas inside the pipeline, as well as the Joule-Thomson effect caused by gas expansion, leads to a localized change in the pipe wall temperature near the leak point, typically manifested as a temperature decrease. To accurately identify this minute change from fluctuations in the ambient background temperature, a method combining differential measurements between adjacent measuring points and spatial gradient statistics is employed.

[0039] Calculate the temperature difference characteristics between adjacent measuring points, for the first point along the optical fiber. The measurement point is extracted. The measuring point and the first Temperature difference at each measuring point at the current moment The calculation formula is as follows: ; in, and They represent the first The first measuring point and the first Real-time temperature values ​​at each measuring point. Set a temperature difference threshold. In this embodiment, the value is taken as 0.3℃. If the calculated... Greater than If so, then the measuring point is marked as a candidate point for temperature anomaly.

[0040] To calculate the characteristics of the space temperature gradient and further distinguish between local abrupt changes caused by leakage and overall gradual changes caused by the environment, a space temperature gradient index is constructed. The spatial sampling interval is set to... The temperature gradient is calculated as follows: ; in, and Representing positions respectively The temperature values ​​at and adjacent intervals, This indicates the temperature difference between adjacent measuring points.

[0041] A dynamic baseline was established using statistical methods, and the average temperature gradient at all measuring points along the entire monitoring optical cable was calculated. and standard deviation ,in accordance with The criterion is that a temperature gradient at a certain location is considered an anomaly if it meets the following conditions: ; This decision logic can adapt to the overall drift of the downhole ambient temperature and only respond to local anomalies that exceed statistical regularities.

[0042] Gas leaks cause airflow pulsations and turbulent impacts at the leak point, and this dynamic process excites acoustic signals in a specific frequency band. This embodiment extracts acoustic features from both the frequency and time domains.

[0043] Extracting frequency domain power spectral density features. A short-time Fourier transform (STFT) is performed on the preprocessed acoustic signal to convert the time-domain signal to a time-frequency domain signal, and the acoustic power spectral density (PSD) is calculated. The average power spectral density under normal operating conditions before leakage occurs is defined as the baseline value. The change in power spectral density calculated at the current moment is Constructing an acoustic anomaly discrimination ratio It is defined as the ratio of the change to the benchmark value: ; in, Indicates the acoustic anomaly discrimination ratio. This represents the reference value of the power spectral density under normal operating conditions.

[0044] Set the acoustic threshold to 1.5. If the calculation results satisfy... If the current sound energy is more than 1.5 times higher than the background noise level, then the area is considered to have an acoustic anomaly.

[0045] Extracting temporal dynamic features. To describe the intensity and frequency characteristics of airflow disturbances, the energy envelope of the signal is extracted. Zero crossover rate (ZCR). Energy envelope. The ZCR (Zero Cross Rate) is used to represent the amplitude and intensity of airflow disturbances. It calculates the Hilbert transform modulus of the signal amplitude within a certain time window; a high-energy envelope corresponds to a strong airflow impact. The zero-crossing rate (ZCR) is used to represent the frequency characteristics of the airflow state, calculating the number of times the signal crosses the zero-level axis per unit time. Because turbulent noise generated by high-pressure gas leaks typically contains abundant high-frequency components, the ZCR increases significantly when a leak occurs. When analyzed in conjunction with ZCR, the signal is confirmed to be generated by continuous airflow disturbance rather than instantaneous mechanical impact interference when both show an upward trend and exceed their respective historical statistical averages.

[0046] The wavelength drift of the FBG sensor is monitored in real time and converted into strain value. When the monitored strain change rate or absolute value exceeds the preset safety threshold, it indicates that there may be a risk of pipe deformation or physical damage at that location, serving as an auxiliary basis for leak detection.

[0047] In step S3, the support vector machine algorithm is used to fuse and identify temperature and acoustic features, specifically including the following sub-steps: A multi-dimensional input feature vector is constructed. To comprehensively reflect the combined thermodynamic and kinetic distribution characteristics of pipeline leakage, a statistical quantity that can represent the degree of signal dispersion is selected as the model input. A sliding time window is set, the length of which includes several sampling points. The variance of the preprocessed temperature signal within this time window is calculated. variance of acoustic signal power spectrum .

[0048] Temperature variance This reflects the stability of temperature data fluctuations over time; a temperature drop caused by leakage will increase this variance. Sound power variance. It reflects the oscillation intensity of sound energy; airflow turbulence causes sound power to exhibit high dispersion in the time domain.

[0049] The two statistical features mentioned above are combined to form a feature vector. : ; This feature vector As sample points, they are mapped into a two-dimensional feature space, where one dimension represents thermodynamic wave characteristics and the other represents kinetic energy characteristics.

[0050] Perform a non-linear classification mapping based on support vector machines. Transfer the feature vectors... The input is fed into a pre-trained Support Vector Machine (SVM) classifier. Considering the nonlinear coupling between temperature and acoustic changes during leak evolution, a Radial Basis Function (RBF) is used as the kernel function to map the low-dimensional feature space to a high-dimensional space to find the optimal classification hyperplane. This classification hyperplane maximizes the margin between leaked sample classes and normal sample classes, thereby achieving effective separation of leak events under complex operating conditions. The specific training process and parameter optimization of the SVM classifier fall within the scope of conventional machine learning techniques in this field and will not be detailed here.

[0051] Output leakage probability and alarm determination. The SVM classification model evaluates the feature vector of the current input. The analysis outputs a value between 0 and 1, which is defined as the leakage probability. This indicates the confidence level that the current pipeline condition constitutes a leak event.

[0052] Set alarm threshold In this embodiment, based on the training results of historical leak data, The value is set to 0.85. The system generates the judgment result based on the following logic. : ; in, Indicates an alarm status. Indicates normal or interference status. This indicates the confidence level of the probability of leakage.

[0053] when If a leak occurs at the current monitoring location, the system will determine that a leak has occurred and trigger the subsequent location and alarm process; otherwise, it will be judged as a normal or interference signal.

[0054] This fusion judgment mechanism utilizes the complementary characteristics of temperature signals, which have strong resistance to transient interference but slow response, and acoustic signals, which have fast response speed but are easily affected by mechanical vibration. An alarm will only be triggered when the thermodynamic and dynamic characteristics simultaneously fall into the high-confidence leakage area in the feature space, thereby effectively suppressing false alarms caused by fluctuations of a single signal source (such as only motor vibration without temperature change, or only ambient temperature difference without airflow sound).

[0055] See attached document Figure 4 In step S4, for events identified as leaks, the spatial distribution characteristics of optical fibers are used to perform physical location inversion and accuracy verification, specifically including the following steps: The system establishes a physical space mapping model for the fiber optic sensing channel. The distributed fiber optic sensor discretizes the continuous physical field along the length of the optical cable into a series of sampling points, each corresponding to a unique spatial index. Based on the time-of-flight principle of optical pulses in optical fibers, a sampling point index and its distance from the beginning of the optical cable are established. A linear correspondence between them.

[0056] When an abnormal signal peak is detected, the sampling point index corresponding to that peak is extracted. Combined with the system's spatial sampling resolution (i.e., the fiber optic distance between two adjacent sampling points), calculate the absolute fiber optic distance from the anomaly point to the fiber optic cable access point. : ; in, This represents the length coordinates along the optical cable laying path. Since the optical cable is laid in a spiral pattern, a preset spiral mapping algorithm is needed to convert the optical cable length coordinates back to the axial length coordinates of the pipeline, and then map them to the three-dimensional spatial coordinates in the well.

[0057] To eliminate potential biases in locating the leak using a single physical quantity, the suspected leak center was calculated independently based on both temperature and acoustic data.

[0058] For temperature data, search for identified temperature anomaly regions and find temperature gradients. The point of maximum temperature drop, or the location of the largest temperature decrease, is defined as the thermodynamic leakage center, denoted as . This location reflects the core area of ​​the cold source caused by the leak.

[0059] For acoustic data, search for the ratio of acoustic power spectral density. The peak position is defined as the dynamic leakage center, denoted as . This location reflects the point where the sound energy of the airflow turbulence source is strongest.

[0060] A dual-source location matching verification strategy is implemented to eliminate false locations caused by environmental interference (such as localized heat sources without sound sources, or misjudgments due to distant mechanical noise). Thermodynamic leakage center is calculated. With the center of dynamic leakage absolute distance deviation between : ; Set the localization convergence threshold In this embodiment The value is set to 1m. The final output logic of the positioning system follows these constraints: Only when If the location result is deemed valid, then the temperature field anomaly and the sound field anomaly are considered to be triggered by the same leak source, and the arithmetic mean of the two is taken as the final precise leak location. : ; like If the abnormal centers of two physical quantities do not coincide spatially and the deviation exceeds the allowable range, the system will determine them as non-homogeneous interference signals and will not perform leakage location output, thus ensuring high confidence and high accuracy of the positioning results. By utilizing the rapid response characteristics of acoustics to capture transient positions and the steady-state characteristics of temperature to correct position drift, a comprehensive positioning accuracy better than ±1m is achieved.

[0061] In step S5, the signal steady-state optimization and intelligent linkage control with the gas extraction system in the complex downhole environment are achieved through the following specific implementation process: To suppress periodic background noise and random broadband interference generated by equipment such as downhole fans and water pumps, and to improve the signal-to-noise ratio of acoustic signals, adaptive noise cancellation (ANC) processing based on the minimum mean square error criterion is implemented. An adaptive filter is constructed, and a reference signal is introduced. The reference signal is derived from the main noise source (such as near the fan) or from background noise in a non-leakage area. The original acquired signal is set to include the desired leakage signal and background noise.

[0062] Adaptive filters adjust the weight vector The reference signal is filtered to output an estimated noise. : ; in, This represents the transpose of the weight vector. This represents a discrete-time index.

[0063] Calculate the system output error signal This error signal is the estimated leakage signal after removing relevant noise: ; in, This represents the original acquired signal. This indicates an estimated noise signal.

[0064] The LMS algorithm is used to iteratively update the filter weights in order to minimize the mean square value of the error signal. ; in, Step size factor For the first The weight vector of the next iteration This represents the filter weight vector for the next time step. Through this processing, the system can dynamically track and cancel out time-varying background noise in the well, highlighting the faint acoustic characteristics of leaks.

[0065] A temperature correction model is established to address temperature baseline drift caused by non-leakage factors (such as diurnal temperature variation of ambient temperature and gas adiabatic expansion caused by fluctuations in extraction negative pressure).

[0066] Collect real-time negative pressure values ​​in the pipeline The flow rate of the mixed gas inside the pipe Define the correction factor and , respectively, correspond to the weights of the thermodynamic effects of pressure changes and flow rate changes on pipe wall temperature. Among them, the correction coefficient... and By selecting historical monitoring data during a non-leakage period, with temperature change as the dependent variable and negative pressure change and flow velocity change as independent variables, the results were obtained using multiple linear regression fitting.

[0067] Construct a corrected equation for the original temperature measured by DTS. Compensation is performed to obtain the corrected temperature. : ; in, and The compensation mechanism uses the baseline negative pressure and baseline flow rate under normal operating conditions to eliminate global temperature fluctuations caused by operating condition adjustments, ensuring that the system only responds to local abnormal cooling caused by the intake of external air, thus reducing the false alarm rate.

[0068] A deep learning model is used to predict the trend of pipeline status, and a CNN-LSTM hybrid neural network model is constructed, based on the standardized monitoring matrix sequence constructed in step S1. As model input, a convolutional neural network (CNN) layer is used to extract the temperature-acoustic coupling feature map along the optical fiber in the spatial dimension, and a long short-term memory network (LSTM) layer is used to capture the signal evolution pattern in the temporal dimension, thus processing past data. The model takes the feature matrix sequence at each time step as input and outputs the future... Predicted leakage probability at time 1 .

[0069] By calculating the trend of leakage probability changes in real time, the system issues an early warning when the predicted leakage probability shows a continuous upward trend and approaches the alarm threshold, allowing maintenance personnel to intervene and investigate.

[0070] Based on the fusion judgment results and the severity of the leak, the ground monitoring platform executes a tiered response and closed-loop control strategy. This is based on the leak probability. Based on signal strength, leakage events are classified into different levels.

[0071] When a leak is determined to be a Level I minor leak, the system highlights the leak location on the 3D GIS interface and pushes an inspection work order containing the location coordinates and abnormal amplitude to the dispatch center.

[0072] When a Level II or higher severe leak is detected, the alarm and linkage control unit automatically triggers a control command. The system sends a control signal to the gas extraction automatic control system via the industrial bus to adjust the opening of the electric valve on the branch where the leak point is located.

[0073] Set the current valve opening to The target adjustment step size is Perform the following negative pressure suppression operation: ; in, This indicates the target valve opening after adjustment.

[0074] By reducing the valve opening, the negative pressure of the extraction in this branch is reduced, thereby reducing the intake of external air and preventing the gas concentration from exceeding the limit. The system continuously monitors the signal feedback after adjustment. If the leakage characteristics do not significantly decrease, the adjustment range is further increased or the branch is cut off until the hidden danger is eliminated.

[0075] Please see the appendix Figure 7 The present invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it is able to perform the above method.

[0076] The present invention also provides a storage medium storing a computer program, which is executed by a processor to perform the method described above.

[0077] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0078] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology, characterized in that, Includes the following steps: S1. Simultaneously collect distributed temperature data and distributed acoustic data along the gas extraction pipeline, perform noise reduction processing on the collected data, and align and map the noise-reduced data based on a unified spatiotemporal coordinate reference to construct a standardized monitoring matrix. S2. Based on the temperature data components in the standardized monitoring matrix, spatial gradient features are generated, and simultaneously, based on the acoustic data components in the standardized monitoring matrix, frequency domain energy features are generated to screen out candidate regions for thermodynamic anomalies and candidate regions for kinetic anomalies, respectively. S3. Using the statistical parameters of the spatial gradient features and the frequency domain energy features, a multi-dimensional feature vector is constructed and input into the classification model for feature space mapping to generate a determination result indicating whether a leakage event exists at the current monitoring location. S4. In response to the judgment result indicating the existence of a leakage event, extract the spatial location information of the abnormal feature signal, use the thermodynamic anomaly center determined based on temperature data to perform spatial consistency verification on the dynamic anomaly center determined based on acoustic data, and generate the final leakage location coordinates based on the data after the verification is passed. S5. Based on the determination result and the final leak location coordinates, generate an alarm signal, and construct control instructions for the gas extraction pipeline according to the alarm signal to drive the adjustment of pipeline operating parameters.

2. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, In step S1, the step of constructing the standardized monitoring matrix further includes: An integrated composite optical cable, laid in a spiral winding manner along the outer wall of the gas extraction pipeline, is used as a signal acquisition carrier to obtain raw optical signals with high spatial resolution. The wavelet transform algorithm is used to perform multi-scale decomposition and reconstruction on the data obtained from the original optical signal to generate a clean data sequence with random noise components removed. Using high-resolution sensor modal data as a benchmark, an interpolation algorithm is used to drive the resampling process of low-resolution modal data, mapping distributed temperature data and distributed acoustic data onto a unified spatial grid node, thereby generating a standardized monitoring matrix aligned in the spatiotemporal dimension.

3. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, In step S1, the denoising process on the acquired data further includes an acoustic data reconstruction step based on frequency domain characteristics: Construct a digital bandpass filter whose passband frequency covers the frequency band of gas leakage airflow noise; The digital bandpass filter is used to perform filtering operations on the collected distributed acoustic data to remove mechanical vibration components below the lower limit frequency and background noise components above the upper limit frequency, generating a preprocessed acoustic data sequence that retains only the characteristic frequency band of leakage sound emission, and the preprocessed acoustic data sequence is used as the basic data for constructing the acoustic data components in the standardized monitoring matrix.

4. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, In step S2, the screening step for the candidate thermodynamic anomaly region further includes: Based on the temperature data components in the standardized monitoring matrix, the temperature difference between adjacent spatial grid nodes is calculated, and the spatial temperature gradient value distributed along the axial direction of the gas extraction pipeline is calculated based on the temperature difference. The mean and standard deviation are calculated based on the temperature data distribution pattern of the entire monitoring optical cable, and a dynamic judgment baseline that conforms to statistical criteria is constructed. The calculated spatial temperature gradient value is compared with the dynamic judgment baseline, and regions whose gradient values ​​exceed the dynamic judgment baseline are selected and marked as candidate regions of thermodynamic anomalies.

5. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, In step S2, the generation step of the frequency domain energy features further includes: Perform a short-time Fourier transform on the acoustic data components in the standardized monitoring matrix to calculate the acoustic power spectral density, and calculate the rate of change of the acoustic power spectral density at the current moment relative to the leak-free reference value. The temporal energy envelope and zero crossover rate of acoustic data are extracted simultaneously. The changing trends of the change ratio, energy envelope and zero crossover rate are jointly analyzed to construct a frequency domain energy feature representing the intensity of airflow turbulence. Based on the frequency domain energy feature, candidate regions of dynamic anomalies are screened out.

6. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, Step S3 further includes: The variance of the spatial gradient features within the sliding time window is calculated as the temperature dispersion parameter, and the variance of the frequency domain energy features within the sliding time window is calculated as the acoustic dispersion parameter. The temperature dispersion parameter and the acoustic dispersion parameter are combined to construct a multidimensional feature vector representing the stability of signal fluctuations. The classification model employs a support vector machine, using radial basis functions as kernel functions to map the multidimensional feature vectors to a high-dimensional feature space, and searches for the optimal classification hyperplane to separate leaked samples from normal samples, thereby outputting the judgment result containing the leakage probability value.

7. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, In step S4, the step of generating the final leak location coordinates further includes: Using a spatial mapping model based on spiral winding laying parameters, the optical cable length coordinates are converted into duct axial length coordinates and three-dimensional spatial coordinates; The maximum temperature gradient point is searched in the candidate region of thermodynamic anomaly as the center of thermodynamic anomaly, and the peak point of acoustic power spectral density is searched in the candidate region of kinetic anomaly as the center of kinetic anomaly. The spatial distance deviation between the center of thermodynamic anomaly and the center of kinetic anomaly is calculated. The spatial distance deviation is compared with a preset positioning convergence threshold. When the spatial distance deviation is less than the positioning convergence threshold, the arithmetic mean of the thermodynamic anomaly center and the kinetic anomaly center is calculated and output as the final leak location coordinates.

8. The method for detecting leaks in gas extraction pipelines based on fiber optic sensing technology according to claim 1, characterized in that, Step S1 also includes a step of correcting temperature data based on operating condition parameters: Obtain the real-time negative pressure value and mixed gas flow rate value inside the pipeline; Retrieve a pre-built temperature correction model, which defines the mapping relationship between pressure change, flow rate change and pipe wall temperature correction. The real-time negative pressure value and the mixed gas flow rate value are input into the temperature correction model to generate the temperature correction amount under the current operating condition. The temperature correction amount is then used to compensate the collected distributed temperature data to generate a corrected temperature sequence that eliminates the influence of operating condition fluctuations. The corrected temperature sequence is then used to participate in the construction of the standardized monitoring matrix.

9. A gas extraction pipeline leakage detection system based on fiber optic sensing technology, applied to the gas extraction pipeline leakage detection method based on fiber optic sensing technology as described in any one of claims 1-8, characterized in that, include: The distributed optical fiber sensing unit includes an integrated composite optical cable laid along the outer wall of the gas extraction pipeline, which is used to construct a physical field signal acquisition link. The signal demodulation and processing unit is connected to the distributed optical fiber sensing unit and is used to perform the construction of the standardized monitoring matrix, the generation of spatial gradient features and frequency domain energy features, the output of the judgment result, and the verification calculation of the final leakage location coordinates. The alarm and linkage control unit is communicatively connected to the signal demodulation and processing unit. It is used to issue graded alarm signals in response to the judgment result and generate control commands based on the alarm signals to drive external equipment to perform pipeline operating parameter adjustments. The ground monitoring and visualization platform is used to receive and display the final leak location coordinates and pipeline status information.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a gas extraction pipeline leakage detection method based on fiber optic sensing technology as described in any one of claims 1-8.

Images