A horizontal well gas-liquid two-phase flow identification and steady flow control method based on acoustic wave feature fusion

Abstract

This invention discloses a method for identifying and controlling horizontal well gas-liquid two-phase flow based on acoustic feature fusion, belonging to the field of multiphase flow detection and control technology in petroleum engineering. The method first uses a non-invasive acoustic sensor to collect pipeline acoustic signals, extracting time-domain energy and frequency-domain spectral density features through bandpass filtering and frame processing. Next, a primary classification logic based on a time-domain energy threshold is constructed to quickly classify the flow pattern into stratified flow, annular flow, and slug flow transition zones. For the transition zone signals, frequency-domain feature analysis is further introduced, using the bimodal structure of the spectrum to finely distinguish between ordinary slug flow and strong slug flow. Finally, a closed-loop feedback control mechanism is established based on the identification results. When a strong slug flow risk condition is identified, step-by-step throttling control is automatically executed until the flow pattern stabilizes. This invention solves the problem of low accuracy of single-feature identification through deep fusion of multi-domain features and eliminates the risks caused by strong slug flow through closed-loop feedback control, significantly improving the safety of horizontal pipeline gas-liquid two-phase flow transportation systems and possessing broad engineering application value.

Description

Technical Field

[0001] This invention relates to a novel method for identifying and actively controlling the gas-liquid two-phase flow pattern in a horizontal pipeline under complex operating conditions, specifically a method for identifying and controlling the gas-liquid two-phase flow in a horizontal well based on acoustic feature fusion. Background Technology

[0002] Horizontal well production technology is a widely used and important development method in the modern petroleum industry, and has been applied on a large scale in shale gas, tight oil, and offshore oil and gas fields. Because horizontal wells typically have long horizontal pipe sections and complex wellbore structures, the fluids within the well often flow in a gas-liquid two-phase mixture, exhibiting significant nonlinear and strong transient characteristics. Under different operating conditions, various flow patterns easily form within the wellbore and horizontal pipe sections, such as stratified flow, slug flow, and annular flow, with frequent transitions between these patterns. The existence and evolution of these complex flow patterns not only affect the pressure drop characteristics and liquid holdup distribution within the pipe, but also directly relate to the operational stability and production safety of the gathering and transportation system. For example, strong slug flow can easily cause drastic fluctuations in flow rate and pressure, adversely affecting the production system. Therefore, achieving accurate identification and proactive control of complex gas-liquid two-phase flow patterns is a key technical issue in oil and gas production operation management.

[0003] Currently, although various technical solutions have been proposed for the identification and control of flow patterns in horizontal pipeline gas-liquid two-phase flow, they are still mainly limited by two problems in complex industrial conditions: limited detection methods and insufficient identification accuracy.

[0004] In terms of detection methods, optical observation methods rely on optical accessibility and are difficult to apply to opaque metal pipes; X-ray detection methods have penetrating ability, but pose radiation safety risks and have high equipment costs; probe methods based on capacitance or conductivity principles are mostly invasive measurements, which are prone to disturbing the flow field and wear during long-term operation, thus affecting measurement stability.

[0005] In terms of recognition accuracy, existing flow pattern recognition methods still face problems such as insufficient model reliability and limited feature representation dimensions. On the one hand, data-driven methods based on deep neural networks lack clear physical constraints, making the decision-making process difficult to interpret under strong noise conditions, resulting in limited reliability of the results. On the other hand, traditional methods often rely on single or limited feature parameters, making it difficult to distinguish between similar flow patterns such as slug flow and strong slug flow, which can easily lead to misjudgments and potential operational risks.

[0006] Currently, research on the above-mentioned problems is insufficient. Existing technical approaches are mostly limited to single-dimensional signal representation, lacking an innovative method that can solve the identification of strong slug flows through deep fusion of multi-domain features, thus failing to provide a reliable decision-making basis for subsequent steady-flow control. Therefore, it is urgent to establish a gas-liquid two-phase flow identification and steady-flow control scheme based on acoustic feature fusion to reduce the uncertainty and technical risks in the operation of oil and gas gathering and transportation systems. Summary of the Invention

[0007] To address the limitations of existing technologies in detecting gas-liquid two-phase flow in horizontal pipelines, the low accuracy of identifying strong slug flows, and the lack of stable flow closed-loop control, this invention aims to provide a method for identifying and controlling gas-liquid two-phase flow in horizontal wells based on acoustic feature fusion. This method systematically fuses the time-domain and frequency-domain features of the signal based on traditional non-invasive acoustic detection, constructing a flow pattern identification framework that integrates time / frequency domain features. Based on the identification results, a valve control logic of "identification-feedback-regulation" is established, thereby constructing a highly robust flow pattern diagnosis and stable flow system. This invention effectively solves the problem of insufficient identification accuracy of single features under complex operating conditions by introducing a multi-domain feature fusion mechanism and a bimodal spectrum criterion for strong slug flows. Furthermore, it utilizes closed-loop feedback control to eliminate the risks posed by strong slug flows, significantly improving the safety of horizontal pipeline gas-liquid two-phase flow transportation systems and possessing broad engineering application value.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a Bayesian manifold identification method for microchannel boiling and fusion noise, the method comprising: Step 1: Non-invasive acoustic signal acquisition and multi-dimensional feature extraction S1. Sensor selection: Select a suitable sensor based on the pipeline material and actual working conditions to ensure that the effective response frequency range covers the entire multiphase flow characteristics, so as to ensure the complete acquisition of wideband signals. S2. Sensor Arrangement: A non-invasive piezoelectric acoustic sensor is installed on the outer wall of a horizontal well or horizontal gathering pipeline. The sensor is tightly attached to the pipe wall with a coupling agent and is used to receive acoustic emission signals generated by the gas-liquid two-phase flow inside the pipe. S3. Data preprocessing: The acquired raw acoustic signal is amplified by a preamplifier; the amplified signal is filtered by a bandpass filter, the passband frequency range is set to the effective frequency band (e.g., 15 kHz-70 kHz), and the non-stationary signal is divided into quasi-stationary segments using rolling time window technology. S4. Time / Frequency Feature Calculation: For each time window, calculate the time domain features including: average absolute energy and root mean square value; calculate the frequency domain features including: perform fast Fourier transform on the signal, calculate the power spectral density, and extract the feature height (the gain of the PSD peak relative to the floor noise) and the main peak frequency.

[0009] Step 2: Primary classification of manifolds based on time-domain energy A multi-level discrimination logic based on the time / frequency domain energy characteristics of acoustic signals is constructed. By using preset energy thresholds, the flow pattern is divided into different energy level intervals to achieve rapid initial screening. S1. If the time-domain average energy is lower than the first preset threshold, it is a stratified flow; S2. If the time-domain average energy is between the first preset threshold and the second preset threshold, then it is a slug flow. S3. If the time-domain average energy is higher than the second preset threshold, it is determined to be a circular flow. Step 3: Fine Identification of Strong Slug Flow Based on Frequency Domain Features Based on the signal characteristics of slug flows, frequency domain features are introduced for secondary identification, and bimodal acoustic features are used to distinguish between ordinary slug flows and strong slug flows: S1. Signal framing and windowing: The continuous time-domain signal is divided into several short-time analysis frames of fixed length using the sliding window method, and a smoothing window function (such as Hamming window) is applied to each frame to reduce boundary effects. S2. Spectrum Transformation: Perform a Fast Fourier Transform on each frame of the time-domain signal after windowing to map the signal from the time domain to the frequency domain, thereby resolving the frequency components and their corresponding amplitude information contained in the signal within the time window. S3. Power spectral density calculation: The power spectral density of each frame is calculated based on the spectral amplitude, which is used to characterize the distribution characteristics of acoustic signal energy at different frequencies. S4. Effective frequency band selection and energy aggregation: Combining the effective response range of the sensor and the characteristics of fluid acoustics, select the effective frequency range, calculate the sound wave energy distribution in the range, and aggregate the energy of each frequency band for analysis. S5. Strong slug flow determination: Examine the spectral structure within the effective frequency band. If it exhibits a significant bimodal characteristic (i.e., simultaneously containing a low-frequency main peak corresponding to the slug motion and a high-frequency secondary peak corresponding to turbulent collapse), it is determined to be a strong slug flow; if it only exhibits a single main peak, it is determined to be a normal slug flow.

[0010] Step 4: Steady Flow Control Based on Flow Pattern Feedback Based on the identification results from steps two and three, the controller executes a hierarchical control strategy: S1. Steady-state maintenance: When the flow is identified as stratified flow, annular flow or ordinary slug flow, it is determined to be a non-water hammer risk condition, and the opening of the outlet regulating valve is kept unchanged. S2, Dynamic Throttling and Surge Suppression: When the identification result is strong slug flow, it is determined to be a risky operating condition. The controller immediately starts the "step-by-step throttling control" program: it sends a command to the electric regulating valve at the pipeline outlet to gradually reduce the valve opening by a preset step size (such as 5% of the current opening). After each adjustment, the system continuously monitors the flow pattern change (return to step one); S3. Termination condition: After each adjustment, continue monitoring until the flow pattern identification result changes to ordinary slug flow or stratified flow, then stop the adjustment and lock the current opening to achieve steady flow control.

[0011] Compared with existing technologies, the advantages of this invention are as follows: In terms of detection methods, either optical observation is relied upon, which is limited by the transparency of the pipe; or X-ray detection is used, which faces the problem of high cost; or invasive probes are used, which can obtain local data but disturb the flow field and are prone to wear. In terms of recognition logic, traditional methods mostly rely on single time-domain or frequency-domain features, or directly apply "black box" neural network models that lack physical constraints. The bottleneck of this approach is that single features are difficult to distinguish the transitional state in the evolution of flow patterns (such as ordinary slug flow and strong slug flow), and pure data-driven models often fail to make decisions in noisy industrial sites due to a lack of physical interpretability, and cannot provide a reliable basis for subsequent production control. This invention proposes a technical concept based on the deep integration of physical acoustic characteristics and multi-domain features: from "intrusion-restricted" to "full-frequency-domain non-intrusive sensing": This invention uses a non-intrusive acoustic sensor attached to the pipe wall to achieve broadband capture of acoustic emission signals of complex two-phase flow inside the pipe without destroying the flow field. This design not only avoids sensor wear and fluid interference issues, but also transforms complex non-stationary signals into quasi-stationary segments containing rich physical information through bandpass filtering and frame processing, laying a solid data foundation for subsequent refined identification. This invention innovatively constructs a two-level discrimination framework of "time-domain initial screening + frequency-domain fine measurement." Unlike traditional methods that blur the identification of slug flows, this invention successfully distinguishes between "ordinary slug flows" and "strong slug flows" by identifying the bimodal structure in the spectrum—the "low-frequency main peak" corresponding to the periodic motion of the liquid slug and the "high-frequency secondary peak" corresponding to the turbulent collapse of the gas-liquid interface. This feature extraction based on physical evolution mechanisms gives the model extremely high confidence and physical interpretability in capturing flow pattern changes.

[0012] Because the recognition algorithm deeply couples time-domain energy and frequency-domain structural features, its ability to resist background noise in actual working conditions is significantly enhanced. Even under strong noise interference, its recognition performance does not significantly decrease thanks to the "bimodal" physical criterion, demonstrating robustness that is difficult to achieve with traditional single-dimensional recognition methods.

[0013] This invention can be widely applied in fields such as petroleum engineering, long-distance multiphase flow transportation, and horizontal well mining. Attached Figure Description

[0014] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a horizontal well gas-liquid two-phase flow identification and steady flow control method based on acoustic feature fusion, as described in this invention.

[0015] Figure 2 This is a diagram of the testing device described in this invention.

[0016] Figure 3 The present invention refers to (a) the frequency domain characteristics of slug flow and (b) the frequency domain characteristics of strong slug flow. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other, and the described embodiments are only some embodiments of the present invention, not all embodiments.

[0018] Step 1: Non-invasive acoustic signal acquisition and multi-dimensional feature extraction S1. Sensor selection: Select a suitable sensor based on the pipeline material and actual working conditions to ensure that the effective response frequency range covers the entire multiphase flow characteristics, so as to ensure the complete acquisition of wideband signals. S2. Sensor Arrangement: A non-invasive piezoelectric acoustic sensor is installed on the outer wall of a horizontal well or horizontal gathering pipeline. The sensor is tightly attached to the pipe wall with a coupling agent and is used to receive acoustic emission signals generated by the gas-liquid two-phase flow inside the pipe. S3. Data preprocessing: The acquired raw acoustic signal is amplified by a preamplifier; the amplified signal is filtered by a bandpass filter, the passband frequency range is set to the effective frequency band (e.g., 15 kHz-70 kHz), and the non-stationary signal is divided into quasi-stationary segments using rolling time window technology. S4. Time / Frequency Feature Calculation: For each time window, calculate the time domain features including: average absolute energy and root mean square value; calculate the frequency domain features including: perform fast Fourier transform on the signal, calculate the power spectral density, and extract the feature height (the gain of the PSD peak relative to the floor noise) and the main peak frequency.

[0019] Step 2: Primary classification of manifolds based on time-domain energy A multi-level discrimination logic based on the time / frequency domain energy characteristics of acoustic signals is constructed. The flow pattern is divided into different energy level intervals by a preset energy threshold to achieve rapid initial screening.

[0020] Step 3: Fine Identification of Strong Slug Flow Based on Frequency Domain Features Based on the signal characteristics of the slug flow, frequency domain features are introduced for secondary decision-making, and bimodal acoustic features are used to distinguish between ordinary slug flow and strong slug flow.

[0021] Step 4: Steady Flow Control Based on Flow Pattern Feedback Based on the identification results of steps two and three, the controller executes a hierarchical control strategy.

[0022] Implementation methods, see Figure 1 This embodiment describes a Bayesian manifold identification method for microchannel boiling and fusion noise. The method includes: I. System Architecture Test equipment such as Figure 2 As shown, where: Non-invasive acoustic wave acquisition unit: Employs a wideband piezoelectric ceramic sensor (model G40), which is tightly attached to the outer wall of a horizontal pipe using a coupling agent. The sensor's effective response frequency covers 15 Hz - 70 kHz.

[0023] Signal processing unit: Includes a preamplifier (40 dB gain), data acquisition card, and computing terminal. This unit is responsible for performing bandpass filtering (15 kHz – 70 kHz), frame windowing, fast Fourier transform, and multi-dimensional feature extraction algorithms.

[0024] Feedback control execution unit: It consists of a programmable logic controller and an electric regulating valve installed at the outlet end of the pipeline. It is used to receive the identification results and perform valve opening adjustment.

[0025] II. Identification and Control Methods and Procedures Based on the above system, the method flow of this embodiment is as follows: Step S1: Wideband Signal Acquisition and Preprocessing The system acquires the acoustic signal of the pipeline in real time, and filters out mechanical vibration noise below 15 kHz and interference above 70 kHz using a Butterworth bandpass filter. Subsequently, a rolling time window of 5 ms is used to divide the continuous signal into several quasi-stationary analysis frames.

[0026] Step S2: Primary classification of manifolds based on time-domain energy Calculate the average absolute energy within each time window and perform three-level discrimination based on a preset energy threshold.

[0027] Stratified flow determination: When the energy is less than 0.01 J, it is determined to be stratified flow. At this time, the gas-liquid interface is stable and no control is required.

[0028] Circulating flow determination: When the energy is greater than 1.5 J, it is determined to be a circulating flow.

[0029] Slug flow determination: When the energy is between the two, it is determined to be a slug flow, and step S3 is automatically triggered.

[0030] Step S3: Fine-grained identification of strong slug flow based on frequency domain characteristics For the signal falling into the transition region in step S2, perform a fast Fourier transform to calculate the power spectral density and analyze its peak structure within the effective frequency band: Typical slug flow: If the spectrum shows a single main peak only around 2 kHz (e.g.) Figure 3 If the flow pattern is as shown in a), then it is determined to be a normal slug flow. This characteristic corresponds to the periodic motion of the liquid plug unit.

[0031] Strong slug flow determination: If the spectrum exhibits a significant bimodal characteristic, that is, simultaneously containing a low-frequency main peak near 2 kHz and a high-frequency secondary peak near 5 kHz (e.g., Figure 3 If the characteristic height is between 15 dB and 25 dB (as shown in b), it is considered a strong slug flow. This characteristic corresponds to the superposition effect of liquid slug motion and gas-liquid interface turbulent collapse.

[0032] Step S4: Closed-loop steady-flow control based on flow pattern feedback Steady-state maintenance: If the identification result is stratified flow, annular flow or ordinary slug flow, the pipeline is determined to be in a non-risk state, and the current opening of the outlet control valve is maintained unchanged.

[0033] Dynamic throttling and surge suppression: If the identification result is strong slug flow, it is judged as a risky operating condition. The controller immediately activates the step-by-step throttling strategy, sends a command to the outlet regulating valve, and gradually reduces the valve opening in steps of 5% of the current opening to increase the pipeline back pressure to suppress the acceleration of long liquid slugs.

[0034] Closed-loop feedback: After each adjustment, the system returns to step S1 to reacquire the signal. Adjustment stops and the current valve opening is locked when the flow pattern identification result changes to ordinary slug flow or stratified flow, thus achieving stable flow control.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit its protection scope. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading this disclosure, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the published pending claims.

Claims

1. A method for identifying and controlling the steady flow of gas-liquid two-phase flow in horizontal wells based on acoustic feature fusion, characterized in that, Includes the following steps: Step 1: Signal Acquisition and Multidimensional Feature Extraction: Acquire the acoustic signal of the pipeline using a non-invasive sensor, perform preprocessing such as filtering and framing on the signal, and extract the time-domain energy features and frequency-domain spectral density features within each time window. Step 2: Primary classification of flow patterns based on time-domain energy: Construct a multi-level discrimination logic based on time-domain energy characteristics, and divide the flow patterns into low-energy stratified flow regions, high-energy annular flow regions, and slug flow transition regions between the two by using preset energy thresholds; Step 3: Fine identification of strong slug flow based on frequency domain features: For signals falling into the slug flow transition region, the frequency domain spectral density features are introduced for secondary judgment, and the spectral peak structure features within the effective frequency band are used to distinguish between ordinary slug flow and strong slug flow. Step 4: Flow steady control based on flow pattern feedback: Perform graded control based on the final identification results of Step 2 and Step 3: When the identification result is strong slug flow, it is determined to be a risky condition, and the pipeline outlet is controlled to perform throttling operation to suppress fluid instability until the flow pattern characteristics change. When the identification result is a non-risk condition, maintain the current pipeline control status.

2. The method according to claim 1, characterized in that, The preprocessing and feature extraction in step one specifically include: setting the passband range of the bandpass filter to cover the core acoustic response frequency band of the gas-liquid two-phase flow and filtering out environmental background noise; calculating the time-domain energy characteristics, including the average absolute energy or root mean square value of the signal; performing spectral transformation on the signal, calculating the power spectral density, and extracting the frequency-domain spectral density characteristics, including feature height and main peak frequency distribution.

3. The method according to claim 1, characterized in that, The multi-level discrimination logic in step two is as follows: a first energy threshold and a second energy threshold are set, wherein the second energy threshold is greater than the first energy threshold; if the time-domain energy characteristic is lower than the first energy threshold, it is determined to be a stratified flow; if the time-domain energy characteristic is higher than the second energy threshold, it is determined to be a circular flow; if the time-domain energy characteristic is between the first energy threshold and the second energy threshold, it is determined to be a slug flow transition zone containing ordinary slug flow and strong slug flow, and step three is triggered.

4. The method according to claim 1, characterized in that, The step three, which distinguishes between ordinary slug flow and strong slug flow by utilizing the peak structure characteristics of the spectrum, specifically includes: analyzing the peak distribution pattern of the power spectral density within a preset effective frequency range; if the spectrum exhibits a single main peak structure, it is determined to be an ordinary slug flow; if the spectrum exhibits a double-peak structure, that is, simultaneously containing a low-frequency main peak and a high-frequency secondary peak, it is determined to be a strong slug flow.

5. The method according to claim 4, characterized in that, The low-frequency main peak in the bimodal structure corresponds to the pressure fluctuation characteristics generated by the periodic motion of the liquid plug unit, while the high-frequency secondary peak corresponds to the turbulent pulsation characteristics generated by the breakup and collapse of the gas-liquid interface at the front end of the liquid plug.

6. The method according to claim 1, characterized in that, The graded control in step four specifically includes: when the identification result is stratified flow, annular flow or ordinary slug flow, it is determined to be a steady state condition, and the opening of the pipeline outlet regulating valve remains unchanged; when the identification result is strong slug flow, the controller starts a step-by-step throttling strategy, sends a command to the pipeline outlet regulating valve, and gradually reduces the valve opening according to a preset step size.

7. The method according to claim 6, characterized in that, The step-by-step throttling strategy also includes a closed-loop feedback mechanism: after each operation to reduce the valve opening, the system returns to step one to reacquire the signal and perform flow pattern identification; this process is continuously repeated until the flow pattern identification result changes from strong slug flow to ordinary slug flow or stratified flow, at which point the adjustment is stopped and the current valve opening is locked.

8. A horizontal well gas-liquid two-phase flow identification and stabilization control system applying the method of any one of claims 1 to 7, characterized in that, include: The acoustic wave acquisition module is used to acquire broadband acoustic wave signals through a piezoelectric sensor attached to the pipe wall; The computational processing module is used to perform signal preprocessing, multidimensional feature extraction, and manifold discrimination algorithms based on time-frequency domain feature fusion; The feedback control module is used to receive the flow pattern identification results and control the operating status of the electric regulating valve at the pipeline outlet according to the preset flow stabilization strategy.

Images