Performance optimization method and system for underground gas-liquid separator based on artificial neural network

Abstract

The invention discloses an underground gas-liquid separator performance optimization method and system based on an artificial neural network, and relates to the technical field of underground gas-liquid separator performance optimizing.The underground gas-liquid separator performance optimization method comprises the following steps that when an underground gas-liquid mixture is introduced into a gas-liquid separator, flow information and pressure information are obtained at an inlet of the separator in real time; the flow information and the pressure information at an inlet when the gas enters the gas-liquid separator are preliminarily analyzed, and the fluctuation trend of the flow and the pressure is recognized. According to the method, the flow and pressure information is obtained in real time, the deep learning prediction model is combined, the abnormal fluctuation of underground gas-liquid flow is recognized, and the highly dynamic nonlinear change environment and the normal environment are distinguished through the gas-liquid flow environment change coefficient. In a normal environment, the artificial neural network optimizes energy consumption; and during violent fluctuation, fuzzy logic control adjusts parameters of the separator in real time, so that quick response is ensured, energy consumption and equipment loss are reduced, system stability is improved, and the service life of equipment is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of downhole gas-liquid separator performance optimization, and in particular to a downhole gas-liquid separator performance optimization method and system based on artificial neural network. Background Art

[0002] Optimizing the performance of downhole gas-liquid separators based on artificial neural networks refers to using artificial neural network (ANN) models to predict and optimize the working performance of downhole gas-liquid separators. This process usually involves collecting data from the gas-liquid separator under different operating conditions, and using these data to train a neural network model so that it can simulate the performance of the gas-liquid separator. Afterwards, the trained model is used to predict the separation effect under different parameter settings, and the optimal configuration is found by adjusting the input parameters (such as pressure, flow rate, etc.) to improve the gas-liquid separation efficiency, reduce energy consumption, and optimize the overall operating performance. This method can effectively deal with complex nonlinear relationships and multivariate influences, and improve the stability and adaptability of the gas-liquid separator.

[0003] The underground gas-liquid flow environment is complex and changeable, with significant dynamic nonlinear characteristics, which brings great challenges to the performance optimization of gas-liquid separators. These operating condition changes are often difficult to predict accurately. When the artificial neural network model faces drastic fluctuations, it may not be able to adjust the internal parameters in time, resulting in deviations between the predicted output and the actual operating conditions. Especially when there are sudden changes in the inlet flow and inlet pressure of the separator, the model may rely too much on historical data for prediction and fail to fully consider the significant changes in the current environment, which in turn causes a sharp drop in separator performance, not only reducing separation efficiency, but also causing increased energy consumption and even equipment failure. In this case, the real-time response capability of the optimization model is crucial.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network. By acquiring the flow and pressure information at the separator inlet in real time and combining it with the intelligent prediction of a deep learning model, it can effectively identify abnormal fluctuations in the downhole gas-liquid flow environment. Secondly, the gas-liquid flow environment change coefficient is used to distinguish between highly dynamic nonlinear change environments and normal change environments, and the operating performance of the separator is optimized through artificial neural networks and fuzzy logic control mechanisms. Under normal conditions, the artificial neural network optimizes energy consumption and efficiency; under severe fluctuations, the fuzzy logic control adjusts the separator parameters in real time to ensure rapid system response, reduce energy consumption and equipment loss, improve operating stability and extend equipment life, so as to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network, comprising the following steps:

[0007] When the underground gas-liquid mixture is passed into the gas-liquid separator, the flow rate information and pressure information are obtained in real time at the separator inlet;

[0008] Conduct a preliminary analysis of the flow rate and pressure information at the entrance of the gas-liquid separator to identify the fluctuation trend of flow rate and pressure. Input the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, and discover potential abnormal fluctuations in the underground gas-liquid flow environment in advance.

[0009] According to the prediction results of the deep learning model, the underground gas-liquid flow environment is divided into a highly dynamic nonlinear change environment and a normal change environment;

[0010] For the situation predicted as a normal changing environment, a pre-trained artificial neural network model is used to optimize the working performance of the gas-liquid separator;

[0011] In view of the highly dynamic nonlinear change environment, based on the highly dynamic nonlinear changes at the inlet of the gas-liquid separator, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator and reduce the deviation between the predicted output and the actual working conditions.

[0012] Preferably, the flow information of the gas-liquid flow is monitored by using one of a turbine flow sensor, an electromagnetic flow meter or an ultrasonic flow sensor, and the pressure information of the gas-liquid flow is monitored by using one of a piezoelectric pressure sensor, a strain gauge pressure sensor or an optical fiber pressure sensor.

[0013] Preferably, a preliminary analysis is performed on the flow information and pressure information at the entrance when entering the gas-liquid separator to identify the fluctuation trend of the flow and pressure, specifically: under the monitoring window, the flow information and pressure information at the entrance of the separator are acquired in real time, and after analyzing the flow information at the entrance when entering the gas-liquid separator, a flow fluctuation reference value is generated, and the flow fluctuation trend at the entrance when the gas-liquid mixture enters the gas-liquid separator is quantified by the flow fluctuation reference value; after analyzing the pressure information at the entrance when entering the gas-liquid separator, a pressure fluctuation reference value is generated, and the pressure fluctuation trend at the entrance when the gas-liquid mixture enters the gas-liquid separator is quantified by the pressure fluctuation reference value.

[0014] Preferably, the flow fluctuation reference value and pressure fluctuation reference value generated after preliminary analysis of the flow information and pressure information at the entrance of the gas-liquid separator are input into a pre-trained deep learning model to generate a gas-liquid flow environment variation coefficient. The gas-liquid flow environment variation coefficient is used to perform an intelligent prediction of the downhole gas-liquid flow environment, and potential abnormal fluctuations in the downhole gas-liquid flow environment are discovered in advance.

[0015] Preferably, in the monitoring window, the flow information at the inlet of the separator is obtained in real time, and the specific steps of analyzing the flow information at the inlet when entering the gas-liquid separator to generate the flow fluctuation reference value are as follows:

[0016] The gas-liquid flow data at the separator inlet is collected in real time within the monitoring window. The flow sampling is performed at a fixed time interval Δt within the monitoring window T, and the collected flow data is expressed as a discrete sequence Q(t i ), where t i =i·Δt and i=1, 2, 3, ..., n, n is the number of sampling points, Q(t i ) is the flow value of each sampling point, i represents the time point index;

[0017] After obtaining the discrete flow data, the instantaneous flow rate of change over time is calculated. The instantaneous flow rate of change is calculated by the difference in flow at adjacent time points. The calculation expression is:

[0018]

[0019] , where R(t i ) indicates that at time t i The instantaneous flow rate change rate at i ) and Q(t i-1 ) is the flow value at adjacent time points;

[0020] In order to more accurately reflect the violent fluctuation of flow, the instantaneous flow fluctuation acceleration is calculated. The instantaneous flow fluctuation acceleration represents the change of the flow change rate, that is, the violent degree of flow change. The calculation expression is:

[0021]

[0022] , where Q(t i ) indicates that at time t i The instantaneous flow fluctuation acceleration at i ) and R(t i-1 ) is the instantaneous flow rate change rate at adjacent time points;

[0023] Based on the cumulative effect of flow acceleration, a flow fluctuation reference value is generated to quantify the severity of flow fluctuation. By integrating and summing the absolute value of instantaneous flow fluctuation acceleration within the monitoring window T, a comprehensive index reflecting the severity of fluctuation within the entire monitoring window is obtained. The calculation expression is:

[0024]

[0025] , where Flow Wave It is the reference value of flow fluctuation.

[0026] Preferably, in the monitoring window, the pressure information at the inlet of the separator is obtained in real time, and the specific steps of analyzing the pressure information at the inlet when entering the gas-liquid separator to generate a pressure fluctuation reference value are as follows:

[0027] Under the monitoring window, the pressure information at the separator inlet is collected in real time, and the pressure collected at time t is recorded as P(t);

[0028] After obtaining the pressure data, the pressure change rate is calculated through a nonlinear function. The calculation expression is:

[0029]

[0030] , where PRC(t) represents the pressure change rate at time t, and P(t-Δt) represents the pressure value at time (t-Δt);

[0031] Based on the calculated pressure change rate, the instantaneous pressure fluctuation amplitude is calculated to measure the fluctuation intensity at the current moment. The calculation expression is:

[0032]

[0033] , where PFM(t) represents the instantaneous pressure fluctuation amplitude at time t, and τ is the integral variable, representing time;

[0034] In the monitoring window, the pressure fluctuation reference value is further calculated based on the instantaneous pressure fluctuation amplitude. The calculation expression is:

[0035]

[0036] , where Pressure Fluc Indicates the pressure fluctuation reference value, is a weight function used to nonlinearly enhance the high-frequency part of the fluctuation amplitude.

[0037] Preferably, the gas-liquid flow environment variation coefficient generated when predicting the downhole gas-liquid flow environment under the monitoring window using a pre-trained deep learning model is compared and analyzed with a pre-set gas-liquid flow environment variation coefficient reference threshold, and the downhole gas-liquid flow environment is divided. The specific division process is: if the gas-liquid flow environment variation coefficient is greater than or equal to the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a highly dynamic nonlinear variation environment; if the gas-liquid flow environment variation coefficient is less than the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a normal variation environment.

[0038] Preferably, for a highly dynamic nonlinear changing environment, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator. The specific steps are as follows:

[0039] The generated flow fluctuation reference value and pressure fluctuation reference value are passed as input to the fuzzy logic controller, which classifies the input data through fuzzification processing, maps the violent fluctuations of flow and pressure to the high fluctuation state in the fuzzy set, and maps the stable data to the low fluctuation state;

[0040] The fuzzy logic controller performs reasoning operations based on a pre-set fuzzy rule base, where the rules define the adjustment strategy of the system under different fluctuation conditions;

[0041] According to the inference results, the fuzzy logic controller defuzzifies the fuzzified output and converts it into clear control operation instructions, and adjusts the internal parameters of the separator through the system;

[0042] After the system completes the parameter adjustment, it continuously monitors the working status of the gas-liquid separator and evaluates the adjustment effect through a real-time feedback mechanism. If the fluctuation trend of the flow fluctuation reference value and the pressure fluctuation reference value changes, the fuzzy logic controller is immediately used to re-analyze and adjust the parameters again to ensure that the separator maintains the best working state in a dynamically changing environment. The closed-loop control mechanism enables the system to be continuously optimized under complex working conditions.

[0043] The downhole gas-liquid separator performance optimization system based on artificial neural network includes real-time data acquisition module, data analysis and intelligent prediction module, gas-liquid flow environment classification module, normal environment optimization module and dynamic nonlinear control module;

[0044] The real-time data acquisition module obtains the flow rate information and pressure information in real time at the inlet of the separator when the underground gas-liquid mixture is passed into the gas-liquid separator;

[0045] The data analysis and intelligent prediction module performs preliminary analysis on the flow rate and pressure information at the entrance of the gas-liquid separator, identifies the fluctuation trend of flow rate and pressure, and inputs the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, thus discovering potential abnormal fluctuations in the underground gas-liquid flow environment in advance;

[0046] The gas-liquid flow environment classification module divides the underground gas-liquid flow environment into a highly dynamic nonlinear change environment and a normal change environment based on the prediction results of the deep learning model;

[0047] Normal environment optimization module, which uses a pre-trained artificial neural network model to optimize the performance of the gas-liquid separator for conditions predicted to be normal changes;

[0048] The dynamic nonlinear control module, aimed at the highly dynamic nonlinear change environment, adjusts the internal parameters of the gas-liquid separator in real time through fuzzy logic control based on the highly dynamic nonlinear changes at the inlet of the gas-liquid separator, further optimizes the working performance of the downhole gas-liquid separator, and reduces the deviation between the predicted output and the actual working conditions.

[0049] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0050] The present invention effectively identifies potential abnormal fluctuations in the underground gas-liquid flow environment by acquiring the flow information and pressure information at the inlet of the separator in real time, combined with the intelligent prediction of the deep learning model. By calculating and dividing the coefficient of change of the gas-liquid flow environment, the system can accurately distinguish between highly dynamic nonlinear change environments and normal change environments. Under normal change environments, the artificial neural network can continuously optimize the working performance of the gas-liquid separator based on the existing historical data to ensure stable and efficient operation of the system. When facing a highly dynamic nonlinear environment with sudden fluctuations, the fuzzy logic control mechanism can dynamically adjust the internal parameters of the separator according to the characteristics of the real-time fluctuations, respond to complex changes in a timely manner, and ensure that the separator can respond quickly and maintain a high separation efficiency even under severe fluctuations. This ability to flexibly respond to complex changes greatly improves the adaptability of the gas-liquid separator in harsh underground environments and reduces the efficiency reduction and failure risks caused by severe environmental fluctuations.

[0051] The present invention can significantly reduce the deviation between the predicted output and the actual working condition through the intelligent prediction of the deep learning model and the real-time adjustment of the fuzzy logic control, avoid the increase in energy consumption and equipment loss caused by improper parameter settings, and enable the system to automatically adjust the operating parameters of the separator under different gas-liquid flow environments, and optimize the energy consumption distribution in the separation process. Under normal circumstances, the artificial neural network model can ensure that the system is in the best separation state and reduce unnecessary waste of resources; while in a highly dynamic nonlinear environment, the fuzzy logic control can quickly respond to fluctuations to avoid separator overload or parameter imbalance caused by delayed response, thereby reducing the possibility of equipment failure. In addition, the continuous monitoring and feedback control mechanism ensures that the separator can still operate safely and stably in a complex environment, which not only improves the working efficiency of the equipment, but also effectively extends the service life of the equipment and reduces the frequency of maintenance and replacement. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0053] Figure 1 The present invention is a method flow chart of the method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network.

[0054] Figure 2 It is a schematic diagram of the system modules of the downhole gas-liquid separator performance optimization method based on artificial neural network of the present invention. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0056] The present invention provides Figure 1 The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network includes the following steps:

[0057] When the underground gas-liquid mixture is passed into the gas-liquid separator, the flow rate information and pressure information are obtained in real time at the separator inlet;

[0058] The flow rate at the inlet refers to the velocity or flow rate of gas and liquid entering the gas-liquid separator. Changes in flow rate can affect the distribution of gas and liquid and the flow pattern within the separator, such as changes in flow pattern (e.g. from bubble flow to annular flow). A sudden increase or decrease in flow rate can lead to changes in separation efficiency and affect the gas-liquid separation effect.

[0059] The pressure at the inlet refers to the pressure of the gas-liquid mixture when it enters the separator. Sudden changes in pressure may affect the state of the gas and liquid. For example, the gas may expand due to reduced pressure, resulting in more bubbles, or the liquid may dissolve more gas due to increased pressure, which will interfere with the separation process. Fluctuations in inlet pressure may cause significant changes in the fluid dynamics within the device, affecting the separation efficiency and stability of the device.

[0060] These data are the basis for optimizing separator performance. By installing high-precision sensors at the separator inlet, key parameters of gas-liquid flow are monitored. The sensors need to be able to withstand high temperatures and high pressures to adapt to complex downhole conditions. The acquired data should be collected as frequently as possible to capture the dynamic changes in flow and pressure, which will provide rich feature inputs for subsequent deep learning models.

[0061] The flow information of gas-liquid flow can usually be monitored by turbine flow sensors, electromagnetic flow meters or ultrasonic flow sensors. The turbine flow sensor drives the turbine blades to rotate through the gas-liquid flow. The blade rotation speed is proportional to the flow rate, so the flow rate can be calculated by detecting the rotation frequency. Electromagnetic flow meters are suitable for measuring conductive fluids. By establishing a magnetic field in the pipeline, an induced electromotive force will be generated when the fluid flows through. The magnitude of the induced electromotive force is proportional to the flow rate, which is suitable for highly conductive liquids. Ultrasonic flow sensors measure flow by detecting the time difference of sound wave propagation in the fluid. Its advantages are non-contact and not limited by liquid properties. It is suitable for the measurement of complex mixed fluids such as gas-liquid two-phase flow.

[0062] Each flow sensor has its applicable scope. Turbine flow sensors are often used for clean liquids, but their accuracy may be affected for gas-liquid mixtures containing solid particles. Electromagnetic flow meters are more suitable for conductive liquids and are not suitable for gases and non-conductive fluids. Ultrasonic flow sensors are not limited by fluid types and are particularly suitable for gas-liquid two-phase flow measurements in downhole environments. Due to the high temperature and high pressure characteristics of the downhole environment, the high temperature and corrosion resistance of the sensor are crucial. Selecting a flow sensor suitable for the harsh downhole environment is critical to improving measurement accuracy and reliability.

[0063] Common sensors for monitoring gas-liquid flow pressure information include piezoelectric pressure sensors, strain gauge pressure sensors, and fiber optic pressure sensors. Piezoelectric pressure sensors work through the piezoelectric effect. When the pressure of the gas-liquid mixture changes, the piezoelectric material in the sensor generates an electric charge, and the amount of charge is proportional to the pressure, which is suitable for fast and dynamic pressure change detection. Strain gauge pressure sensors change the resistance value of the strain gauge through the mechanical deformation caused by fluid pressure. The resistance change can accurately measure the pressure, which is suitable for long-term monitoring of stable pressure occasions. Fiber optic pressure sensors can achieve high-precision and remote measurement by detecting the strain change of optical fiber under pressure, which is particularly suitable for high temperature and high pressure environments underground.

[0064] In the underground gas-liquid separation environment, due to high pressure and harsh conditions, the selection of sensors should pay special attention to high temperature resistance and corrosion resistance. Fiber optic pressure sensors are ideal for underground environments because they are not subject to electromagnetic interference and can adapt to high temperature and high pressure. Piezoelectric pressure sensors have fast response speeds and can capture transient changes in pressure, making them suitable for real-time monitoring of dynamic pressure changes. Strain gauge pressure sensors are suitable for long-term monitoring under stable pressure conditions. Therefore, combining the advantages of different types of sensors can more comprehensively and accurately capture the pressure changes of underground gas-liquid flow.

[0065] Conduct a preliminary analysis of the flow rate and pressure information at the entrance of the gas-liquid separator to identify the fluctuation trend of flow rate and pressure. Input the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, and discover potential abnormal fluctuations in the underground gas-liquid flow environment in advance.

[0066] A preliminary analysis is performed on the flow information and pressure information at the inlet when entering the gas-liquid separator to identify the fluctuation trend of the flow and pressure. Specifically, in the monitoring window, the flow information and pressure information at the inlet of the separator are obtained in real time. After analyzing the flow information at the inlet when entering the gas-liquid separator, a flow fluctuation reference value is generated. The flow fluctuation trend at the inlet when the gas-liquid mixture enters the gas-liquid separator is quantified by the flow fluctuation reference value. After analyzing the pressure information at the inlet when entering the gas-liquid separator, a pressure fluctuation reference value is generated. The pressure fluctuation trend at the inlet when the gas-liquid mixture enters the gas-liquid separator is quantified by the pressure fluctuation reference value.

[0067] The flow fluctuation reference value and pressure fluctuation reference value generated after preliminary analysis of the flow information and pressure information at the entrance of the gas-liquid separator are input into the pre-trained deep learning model to generate the gas-liquid flow environment variation coefficient. The gas-liquid flow environment variation coefficient is used to make an intelligent prediction of the downhole gas-liquid flow environment, and potential abnormal fluctuations in the downhole gas-liquid flow environment are discovered in advance.

[0068] Deep learning models (such as convolutional neural networks, long short-term memory networks, etc.) can capture the complex nonlinear characteristics of gas-liquid flow in the well, and predict future flow conditions by learning a large amount of historical data. The training data of the model should include flow information and pressure information under different working conditions and separator performance. Through large-scale data training, the model can capture the dynamic change pattern and improve the accuracy of future working conditions. The prediction results at this time provide a basis for subsequent operations.

[0069] Sudden changes in flow at the separator inlet usually indicate that the downhole gas-liquid flow environment is undergoing complex changes. This is because the flow of the gas-liquid mixture downhole is affected by many factors, such as fluctuations in formation pressure, adjustments in downhole equipment operations, or sudden changes in geological conditions. When the flow fluctuates violently, it means that the flow pattern of the gas-liquid mixture may have changed in a short period of time, such as from bubble flow to different flow types such as annular flow or stratified flow. Such mutations are often accompanied by nonlinear dynamic changes, making the entire gas-liquid separation process complex and unpredictable, which in turn affects the performance and efficiency of the separator. Therefore, sudden changes in flow are an important signal of changes in the complex downhole environment.

[0070] In the monitoring window, the flow information at the separator inlet is obtained in real time. The specific steps for analyzing the flow information at the inlet when entering the gas-liquid separator to generate a flow fluctuation reference value are as follows:

[0071] The gas-liquid flow data at the separator inlet is collected in real time within the monitoring window. The flow sampling is performed at a fixed time interval Δt within the monitoring window T, and the collected flow data is expressed as a discrete sequence Q(t i ), where t i =i·Δt and i=1, 2, 3, ..., n, n is the number of sampling points, Q(t i ) is the flow value of each sampling point, i represents the time point index;

[0072] After obtaining the discrete flow data, the instantaneous flow rate of change over time is calculated. The instantaneous flow rate of change is calculated by the difference in flow at adjacent time points. The calculation expression is:

[0073]

[0074] , where R(t i ) indicates that at time t i The instantaneous flow rate change rate at i ) and Q(t i-1 ) is the flow value at adjacent time points;

[0075] In order to more accurately reflect the violent fluctuation of flow, the instantaneous flow fluctuation acceleration is calculated. The instantaneous flow fluctuation acceleration represents the change of the flow change rate, that is, the violent degree of flow change. The calculation expression is:

[0076]

[0077] , where Q(t i ) indicates that at time t i The instantaneous flow fluctuation acceleration at i ) and R(t i-1 ) is the instantaneous flow rate change rate at adjacent time points;

[0078] Based on the cumulative effect of flow acceleration, a flow fluctuation reference value is generated to quantify the severity of flow fluctuation. By integrating and summing the absolute value of instantaneous flow fluctuation acceleration within the monitoring window T, a comprehensive index reflecting the severity of fluctuation within the entire monitoring window is obtained. The calculation expression is:

[0079]

[0080] , where Flow Wave It is the reference value of flow fluctuation.

[0081] It can be seen from the flow fluctuation reference value that the larger the performance value of the flow fluctuation reference value generated after analyzing the flow information at the inlet when entering the gas-liquid separator under the monitoring window, the more drastic the flow fluctuation and the more unpredictable the change. Specifically, the flow fluctuation reference value is a comprehensive indicator obtained by calculating and integrating the acceleration of the flow change rate. When the flow fluctuation reference value is large, it means that within the monitoring window, the flow changes are not only frequent and large in amplitude, but also accompanied by drastic fluctuations. Such drastic fluctuations usually reflect that the fluid flow state in the system is undergoing significant nonlinear changes, which may involve sudden flow pattern conversion or pressure changes. Such unpredictable fluctuations have a potential impact on the performance of the gas-liquid separator, so the flow fluctuation reference value can effectively identify and quantify these complex operating changes.

[0082] When the pressure at the separator inlet changes suddenly, it usually indicates that the gas-liquid flow environment in the well is undergoing complex changes. Sudden pressure changes may be caused by a variety of factors, such as changes in the formation conditions in the well, the release of formation pressure, and the sudden change of the flow pattern of the gas-liquid mixture. These changes are often accompanied by nonlinear dynamic fluctuations, which make the flow pattern in the gas-liquid separation process unpredictable and have a direct impact on the separation efficiency. Therefore, sudden pressure changes are usually a signal of complex changes in the gas-liquid flow environment in the well.

[0083] Under the monitoring window, the pressure information at the separator inlet is obtained in real time. The specific steps for analyzing the pressure information at the inlet when entering the gas-liquid separator to generate a pressure fluctuation reference value are as follows:

[0084] Under the monitoring window, the pressure information at the separator inlet is collected in real time, and the pressure collected at time t is recorded as P(t);

[0085] After obtaining the pressure data, the pressure change rate is calculated through a nonlinear function. The calculation expression is:

[0086]

[0087] , where PRC(t) represents the pressure change rate at time t, and P(t-Δt) represents the pressure value at time (t-Δt) (Δt is the time interval);

[0088] By performing nonlinear amplification processing (natural logarithm) on the pressure change rate, tiny changes can be captured and given greater weights, thereby accurately describing the transient changes in pressure.

[0089] Based on the calculated pressure change rate, the instantaneous pressure fluctuation amplitude is calculated to measure the fluctuation intensity at the current moment. The calculation expression is:

[0090]

[0091] , where PFM(t) represents the instantaneous pressure fluctuation amplitude at time t (T represents the window time length, which is used to calculate the historical time period of fluctuations, in seconds. By selecting a suitable T, it can reflect short-term fluctuations or long-term trends), τ is the integral variable, representing time;

[0092] This step provides a comprehensive quantification of the fluctuation amplitude through a nonlinear integration method combined with a square sum operation.

[0093] In the monitoring window, the pressure fluctuation reference value is further calculated based on the instantaneous pressure fluctuation amplitude. The calculation expression is:

[0094]

[0095] , where Pressure Fluc Indicates the pressure fluctuation reference value, is a weight function used to nonlinearly enhance the high-frequency part of the fluctuation amplitude;

[0096] This step assigns nonlinear weighting to the historical fluctuation amplitude through a weight function, which enhances the locality of pressure fluctuations and enables the fluctuation index to more sensitively reflect sudden changes.

[0097] It can be seen from the pressure fluctuation reference value that the larger the performance value of the pressure fluctuation reference value generated after analyzing the pressure information at the inlet when entering the gas-liquid separator under the monitoring window, the more drastic the pressure fluctuation and the more difficult to predict the change. This is because the pressure fluctuation reference value can quantify the instantaneous fluctuation intensity of the pressure by nonlinearly amplifying the rate of change of the real-time pressure information and combining it with a comprehensive analysis of the historical fluctuation amplitude. When the index value is large, it means that the pressure change amplitude is large, and the fluctuation is more frequent and complex, indicating that the pressure change in the gas-liquid flow environment has higher uncertainty and nonlinear characteristics, and therefore it is more difficult to accurately predict and control by conventional means. In this case, the system may need a more flexible real-time control mechanism to cope with sudden changes.

[0098] The deep learning model is not specifically limited here, and can realize the flow fluctuation reference value Flow Wave and pressure fluctuation reference value Pressure Fluc Any deep learning model that performs comprehensive analysis to generate the gas-liquid flow environment change coefficient Gas_Liquid can be used. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method; the calculation formula for generating the gas-liquid flow environment change coefficient Gas_Liquid is:

[0099]

[0100] , where r 1 、r 2 They are flow fluctuation reference values ​​Flow Wave and pressure fluctuation reference value Pressure Fluc The preset proportionality factor, and r 1 、r 2 Both are greater than 0.

[0101] It can be seen from the gas-liquid flow environment variation coefficient that, under the monitoring window, the larger the performance value of the flow fluctuation reference value generated after analyzing the flow information at the entrance when entering the gas-liquid separator, the larger the performance value of the pressure fluctuation reference value generated after analyzing the pressure information at the entrance when entering the gas-liquid separator. That is, the larger the performance value of the gas-liquid flow environment variation coefficient generated when predicting the downhole gas-liquid flow environment under the monitoring window using the pre-trained deep learning model, the more complex the dynamic nonlinear change of the downhole gas-liquid flow environment is, and vice versa.

[0102] According to the prediction results of the deep learning model, the underground gas-liquid flow environment is divided into a highly dynamic nonlinear change environment and a normal change environment;

[0103] The gas-liquid flow environment variation coefficient generated when the pre-trained deep learning model is used to predict the downhole gas-liquid flow environment under the monitoring window is compared and analyzed with the pre-set gas-liquid flow environment variation coefficient reference threshold, and the downhole gas-liquid flow environment is divided. The specific division process is: if the gas-liquid flow environment variation coefficient is greater than or equal to the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a highly dynamic nonlinear change environment. A highly dynamic nonlinear change environment refers to a rapid and drastic fluctuation of flow rate and pressure during the downhole gas-liquid flow process, and the characteristics of this fluctuation are complex and cannot be linearly predicted. In this environment, the flow state may quickly change from one flow type to another, such as from bubble flow to annular flow, or a sudden increase in pressure leads to liquid gasification. This mutation may be caused by a variety of factors, such as a sudden change in formation pressure, equipment abnormality, or a sudden change in geological conditions. Because the change speed and amplitude in this environment are fast. If the gas-liquid flow environment variation coefficient is less than the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a normal change environment. Normal change environment refers to the fluctuation of flow rate and pressure in underground gas-liquid flow within a certain range. These changes can be predicted by linear or nonlinear models of historical data, and these change trends are usually stable and controllable. In this case, the gas-liquid flow state is gradually adjusted under different working conditions, and there may be small fluctuations, but there is no sudden and drastic change in the overall flow characteristics. Common reasons for changes may be production process adjustments, slight fluctuations in daily underground conditions, etc.

[0104] For the situation predicted as a normal changing environment, a pre-trained artificial neural network model is used to optimize the working performance of the gas-liquid separator;

[0105] The artificial neural network model can predict the optimal operating parameters of the separator based on historical operating data, such as the optimal gas-liquid ratio, pressure drop, separation speed, etc. In this relatively stable environment, the artificial neural network model can provide accurate predictions to ensure that the separator maintains efficient operation under stable conditions. The artificial neural network model can continuously adjust the working state of the separator through iterative learning, optimize separation efficiency and reduce energy consumption.

[0106] Under normal changing conditions, the performance of the gas-liquid separator can be predicted and optimized through a pre-trained artificial neural network model to ensure that the system's separation efficiency and energy consumption remain at an optimal level. Since these changes are foreseeable, the system can adjust operating parameters based on historical data to ensure that the gas-liquid separator operates efficiently under stable conditions. Optimization operations under such conditions are mainly based on routine adjustments based on established models, and do not require rapid, real-time responses.

[0107] In view of the highly dynamic nonlinear change environment, based on the highly dynamic nonlinear change at the inlet of the gas-liquid separator, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator and reduce the deviation between the predicted output and the actual working conditions;

[0108] In view of the highly dynamic nonlinear changing environment, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator. The specific steps are as follows:

[0109] The generated flow fluctuation reference value and pressure fluctuation reference value are passed as input to the fuzzy logic controller, which classifies the input data through fuzzification processing, maps the violent fluctuations of flow and pressure to the high fluctuation state in the fuzzy set, and maps the stable data to the low fluctuation state;

[0110] This classification method allows the control system to perform qualitative analysis of fluctuation information in an uncertain environment and provide flexible control strategies. The uniqueness of the fuzzy logic controller is that it can handle complex input data with high uncertainty. The flow fluctuation reference value and the pressure fluctuation reference value are fuzzified into high, medium, and low fuzzy levels to provide decision support for subsequent adjustments. In a highly dynamic nonlinear environment, the system needs to handle uncertainty flexibly and quickly, and fuzzy logic is an effective way to solve this problem.

[0111] For example:

[0112] Assume that the flow fluctuation reference value collected at the inlet of the gas-liquid separator is "0.8" and the pressure fluctuation reference value is "0.9". These two values ​​are relatively high, indicating that the flow and pressure fluctuations in the current gas-liquid flow environment are very drastic. The fuzzy logic controller takes these two values ​​as input. In order to perform fuzzification, the flow fluctuation reference value and the pressure fluctuation reference value will be mapped to different fuzzy sets.

[0113] Traffic fluctuation reference value 0.8: can be classified as a high fluctuation state, indicating that the current traffic fluctuation is large.

[0114] Pressure fluctuation reference value 0.9: is classified as a high fluctuation state, indicating that the pressure fluctuation is also very severe.

[0115] In this step, the fuzzy logic controller defines “high” flow fluctuation and “high” pressure fluctuation as two fuzzified inputs, and then enters the reasoning stage.

[0116] The fuzzy logic controller performs reasoning operations based on a pre-set fuzzy rule base, where the rules define the adjustment strategy of the system under different fluctuation conditions;

[0117] For example, "when the flow fluctuation reference value is high and the pressure fluctuation reference value is high, reduce the flow at the separator inlet", "when the flow fluctuation reference value is low and the pressure fluctuation reference value is high, increase the opening degree of the separator pressure regulating valve". The flexibility of the fuzzy rule base allows the system to cope with a variety of complex working conditions.

[0118] The fuzzy rule base is the core part of the fuzzy logic controller. It reasonably adjusts the internal parameters of the gas-liquid separator according to different combinations of flow and pressure fluctuations. This reasoning method based on empirical rules can provide the system with real-time response strategies for specific environments and effectively optimize the operating performance of the gas-liquid separator.

[0119] For example:

[0120] In a fuzzy logic control system, a set of fuzzy rules is defined in advance to handle different combinations of flow and pressure fluctuations. For example, the rule base may contain the following rules:

[0121] Rule 1: If the flow fluctuation reference value is "high" and the pressure fluctuation reference value is "high", reduce the opening of the gas-liquid separator inlet valve and adjust the separation speed.

[0122] Rule 2: If the flow fluctuation reference value is "medium" and the pressure fluctuation reference value is "high", adjust the opening degree of the separator pressure regulating valve appropriately to keep the liquid level stable.

[0123] Rule 3: If the flow fluctuation reference value is "low" and the pressure fluctuation reference value is "medium", slightly increase the separation speed to improve efficiency.

[0124] In the example listed, both flow fluctuation and pressure fluctuation are classified as "high", so Rule 1 is met. According to Rule 1, the system will take measures to reduce the opening of the gas-liquid separator inlet valve and adjust the separation speed to cope with the current fluctuation.

[0125] According to the inference results, the fuzzy logic controller defuzzifies the fuzzified output and converts it into clear control operation instructions, and adjusts the internal parameters of the separator through the system;

[0126] For example, adjusting the opening of the inlet valve, controlling the liquid level, optimizing the separation speed, etc. The result of defuzzification makes the control operation specific and executable, thereby optimizing the overall performance of the separator. After defuzzification, the system generates clear operating instructions for adjusting the key parameters of the separator. In this way, the system can respond quickly according to the real-time flow and pressure fluctuations, ensuring that the separator can adapt to the current nonlinear dynamic change environment and optimize the working performance.

[0127] For example:

[0128] After rule reasoning, the fuzzy logic controller obtains preliminary fuzzy outputs, such as "slightly reduce the inlet valve opening" and "significantly reduce the separation speed." These fuzzy output instructions are converted into specific operation instructions through defuzzification, such as:

[0129] Reduce the inlet valve opening by 15%: The specific operation is to reduce the current valve opening by 15% to reduce the flow input to the gas-liquid separator and control fluctuations.

[0130] Reduce separation speed by 10%: The separation speed inside the separator will be reduced by 10% to ensure that there is enough time to separate gas and liquid in a violent fluctuation environment.

[0131] These specific instructions will be transmitted to the separator's actuator to adjust the working state of the gas-liquid separator in real time, thereby effectively responding to the current nonlinear fluctuation environment.

[0132] After the system has completed the parameter adjustment, it continuously monitors the working status of the gas-liquid separator and evaluates the adjustment effect through the real-time feedback mechanism. If the fluctuation trend of the flow fluctuation reference value and the pressure fluctuation reference value changes, the fuzzy logic controller will immediately re-analyze and adjust the parameters again to ensure that the separator maintains the best working state in a dynamically changing environment. The closed-loop control mechanism enables the system to be continuously optimized under complex working conditions.

[0133] The real-time feedback mechanism is the closed-loop part of the entire control process. Through continuous monitoring and feedback, the system can immediately evaluate the effect after parameter adjustment and further regulate unresolved fluctuations. Dynamic adjustment ensures that the separator can continue to maintain efficient and stable working conditions in highly nonlinear environments.

[0134] For example:

[0135] After the parameter adjustment is performed, the system will continue to monitor the flow fluctuation reference value and the pressure fluctuation reference value to observe the effect of the adjustment. For example, after reducing the inlet valve opening and separation speed, the system may find that:

[0136] The flow fluctuation reference value dropped from 0.8 to 0.5: the flow fluctuation was significantly reduced, indicating that the adjustment was effective.

[0137] The pressure fluctuation reference value dropped from 0.9 to 0.6: the pressure fluctuation was also alleviated and the system tended to be stable.

[0138] If the adjusted fluctuation index achieves the expected effect, the system will maintain the current parameter settings and continue to operate. If the fluctuation index is still high, the system will re-reason through the fuzzy logic controller and further adjust the internal parameters of the separator. For example, if the pressure fluctuation does not improve significantly, the system may further reduce the inlet valve opening or increase the liquid level control. The whole process is a real-time closed-loop feedback mechanism to ensure that the gas-liquid separator can continue to operate efficiently in a dynamically changing environment.

[0139] In the case of highly dynamic nonlinear changes, a more flexible fuzzy logic control system is used for real-time adjustment. In a highly dynamic nonlinear environment, the flow and pressure change rapidly and complexly, which is difficult to cope with by traditional fixed models. Fuzzy logic control can handle this complex nonlinear relationship. Through "fuzzification" processing, the precise input (flow, pressure change) is converted into a fuzzy set, and the adjustment parameters required by the separator are obtained through reasoning through the fuzzy rule set. In this way, the separator can be dynamically adjusted under severe fluctuations to ensure its stable operation. In a highly dynamic nonlinear environment, fuzzy logic control can adjust the internal parameters of the separator in real time. For example, the liquid interface height, pressure set point and separation speed of the separator can be dynamically adjusted to ensure that the gas-liquid mixture can be effectively separated. The core of fuzzy control is to convert complex nonlinear relationships into actual operating parameters through fuzzy reasoning rules. Its advantage is that it can respond to drastically changing working conditions in real time, especially when traditional control methods cannot adapt quickly, fuzzy control can quickly make reasonable adjustments.

[0140] The present invention effectively identifies potential abnormal fluctuations in the underground gas-liquid flow environment by acquiring the flow information and pressure information at the inlet of the separator in real time, combined with the intelligent prediction of the deep learning model. By calculating and dividing the coefficient of change of the gas-liquid flow environment, the system can accurately distinguish between highly dynamic nonlinear change environments and normal change environments. Under normal change environments, the artificial neural network can continuously optimize the working performance of the gas-liquid separator based on the existing historical data to ensure stable and efficient operation of the system. When facing a highly dynamic nonlinear environment with sudden fluctuations, the fuzzy logic control mechanism can dynamically adjust the internal parameters of the separator according to the characteristics of the real-time fluctuations, respond to complex changes in a timely manner, and ensure that the separator can respond quickly and maintain a high separation efficiency even under severe fluctuations. This ability to flexibly respond to complex changes greatly improves the adaptability of the gas-liquid separator in harsh underground environments and reduces the efficiency reduction and failure risks caused by severe environmental fluctuations.

[0141] The present invention can significantly reduce the deviation between the predicted output and the actual working condition through the intelligent prediction of the deep learning model and the real-time adjustment of the fuzzy logic control, avoid the increase in energy consumption and equipment loss caused by improper parameter settings, and enable the system to automatically adjust the operating parameters of the separator under different gas-liquid flow environments, and optimize the energy consumption distribution in the separation process. Under normal circumstances, the artificial neural network model can ensure that the system is in the best separation state and reduce unnecessary waste of resources; while in a highly dynamic nonlinear environment, the fuzzy logic control can quickly respond to fluctuations to avoid separator overload or parameter imbalance caused by delayed response, thereby reducing the possibility of equipment failure. In addition, the continuous monitoring and feedback control mechanism ensures that the separator can still operate safely and stably in a complex environment, which not only improves the working efficiency of the equipment, but also effectively extends the service life of the equipment and reduces the frequency of maintenance and replacement.

[0142] The present invention provides Figure 2 The downhole gas-liquid separator performance optimization system based on artificial neural network shown includes a real-time data acquisition module, a data analysis and intelligent prediction module, a gas-liquid flow environment classification module, a normal environment optimization module and a dynamic nonlinear control module;

[0143] The real-time data acquisition module obtains the flow rate information and pressure information in real time at the inlet of the separator when the underground gas-liquid mixture is passed into the gas-liquid separator;

[0144] The data analysis and intelligent prediction module performs preliminary analysis on the flow rate and pressure information at the entrance of the gas-liquid separator, identifies the fluctuation trend of flow rate and pressure, and inputs the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, thus discovering potential abnormal fluctuations in the underground gas-liquid flow environment in advance;

[0145] The gas-liquid flow environment classification module divides the underground gas-liquid flow environment into a highly dynamic nonlinear change environment and a normal change environment based on the prediction results of the deep learning model;

[0146] Normal environment optimization module, which uses a pre-trained artificial neural network model to optimize the performance of the gas-liquid separator for conditions predicted to be normal changes;

[0147] The dynamic nonlinear control module, for highly dynamic nonlinear change environments, adjusts the internal parameters of the gas-liquid separator in real time through fuzzy logic control based on the highly dynamic nonlinear changes at the gas-liquid separator inlet, further optimizes the working performance of the downhole gas-liquid separator, and reduces the deviation between the predicted output and the actual working conditions;

[0148] The downhole gas-liquid separator performance optimization method based on artificial neural network provided in an embodiment of the present invention is realized by the downhole gas-liquid separator performance optimization system based on artificial neural network. The specific method and process of the downhole gas-liquid separator performance optimization system based on artificial neural network are detailed in the embodiment of the downhole gas-liquid separator performance optimization method based on artificial neural network, which will not be repeated here.

[0149] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0150] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0151] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network, characterized in that: The following steps are involved: When the underground gas-liquid mixture is passed into the gas-liquid separator, the flow rate information and pressure information are obtained in real time at the separator inlet; Conduct a preliminary analysis of the flow rate and pressure information at the entrance of the gas-liquid separator to identify the fluctuation trend of flow rate and pressure. Input the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, and discover potential abnormal fluctuations in the underground gas-liquid flow environment in advance. According to the prediction results of the deep learning model, the underground gas-liquid flow environment is divided into a highly dynamic nonlinear change environment and a normal change environment; For the situation predicted as a normal changing environment, a pre-trained artificial neural network model is used to optimize the working performance of the gas-liquid separator; In view of the highly dynamic nonlinear change environment, based on the highly dynamic nonlinear changes at the inlet of the gas-liquid separator, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator and reduce the deviation between the predicted output and the actual working conditions.

2. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 1, characterized in that: The flow information of the gas-liquid flow is monitored by using one of a turbine flow sensor, an electromagnetic flow meter or an ultrasonic flow sensor, and the pressure information of the gas-liquid flow is monitored by using one of a piezoelectric pressure sensor, a strain gauge pressure sensor or an optical fiber pressure sensor.

3. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 1, characterized in that: A preliminary analysis is performed on the flow information and pressure information at the inlet when entering the gas-liquid separator to identify the fluctuation trend of the flow and pressure. Specifically, in the monitoring window, the flow information and pressure information at the inlet of the separator are obtained in real time. After analyzing the flow information at the inlet when entering the gas-liquid separator, a flow fluctuation reference value is generated. The flow fluctuation trend at the inlet when the gas-liquid mixture enters the gas-liquid separator is quantified by the flow fluctuation reference value. After analyzing the pressure information at the inlet when entering the gas-liquid separator, a pressure fluctuation reference value is generated. The pressure fluctuation trend at the inlet when the gas-liquid mixture enters the gas-liquid separator is quantified by the pressure fluctuation reference value.

4. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 3, characterized in that: The flow fluctuation reference value and pressure fluctuation reference value generated after preliminary analysis of the flow information and pressure information at the entrance of the gas-liquid separator are input into the pre-trained deep learning model to generate the gas-liquid flow environment variation coefficient. The gas-liquid flow environment variation coefficient is used to make an intelligent prediction of the downhole gas-liquid flow environment, and potential abnormal fluctuations in the downhole gas-liquid flow environment are discovered in advance.

5. The method for optimizing downhole gas-liquid separator performance based on artificial neural network according to claim 3, characterized in that: In the monitoring window, the flow information at the separator inlet is obtained in real time. The specific steps for analyzing the flow information at the inlet when entering the gas-liquid separator to generate a flow fluctuation reference value are as follows: The gas-liquid flow data at the separator inlet is collected in real time within the monitoring window. The flow sampling is performed at a fixed time interval Δt within the monitoring window T, and the collected flow data is expressed as a discrete sequence Q(t i ), where t i =i·Δt and i=1, 2, 3, ..., n, n is the number of sampling points, Q(t i ) is the flow value of each sampling point, i represents the time point index; After obtaining the discrete flow data, the instantaneous flow rate of change over time is calculated. The instantaneous flow rate of change is calculated by the difference in flow at adjacent time points. The calculation expression is: ， In the formula, R(t i ) indicates that at time t i The instantaneous flow rate change rate at i ) and Q(t i-1 ) is the flow value at adjacent time points; In order to more accurately reflect the violent fluctuation of flow, the instantaneous flow fluctuation acceleration is calculated. The instantaneous flow fluctuation acceleration represents the change of the flow change rate, that is, the violent degree of flow change. The calculation expression is: ， In the formula, A(t i ) indicates that at time t i The instantaneous flow fluctuation acceleration at i ) and R(t i-1 ) is the instantaneous flow rate change rate at adjacent time points; Based on the cumulative effect of flow acceleration, a flow fluctuation reference value is generated to quantify the severity of flow fluctuation. By integrating and summing the absolute value of instantaneous flow fluctuation acceleration within the monitoring window T, a comprehensive index reflecting the severity of fluctuation within the entire monitoring window is obtained. The calculation expression is: ， In the formula, Flow Wave It is the reference value of flow fluctuation.

6. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 5, characterized in that: Under the monitoring window, the pressure information at the separator inlet is obtained in real time. The specific steps for analyzing the pressure information at the inlet when entering the gas-liquid separator to generate a pressure fluctuation reference value are as follows: Under the monitoring window, the pressure information at the separator inlet is collected in real time, and the pressure collected at time t is recorded as P(t); After obtaining the pressure data, the pressure change rate is calculated through a nonlinear function. The calculation expression is: ， Where PRC(t) represents the pressure change rate at time t, and P(t-Δt) represents the pressure value at time (t-Δt); Based on the calculated pressure change rate, the instantaneous pressure fluctuation amplitude is calculated to measure the fluctuation intensity at the current moment. The calculation expression is: ， In the formula, PFM(t) represents the instantaneous pressure fluctuation amplitude at time t, and τ is the integral variable, representing time; In the monitoring window, the pressure fluctuation reference value is further calculated based on the instantaneous pressure fluctuation amplitude. The calculation expression is: ， Where, Pressure Fluc Indicates the pressure fluctuation reference value, is a weight function used to nonlinearly enhance the high-frequency part of the fluctuation amplitude.

7. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 4, characterized in that: The gas-liquid flow environment variation coefficient generated when predicting the downhole gas-liquid flow environment under the monitoring window using the pre-trained deep learning model is compared and analyzed with the pre-set gas-liquid flow environment variation coefficient reference threshold, and the downhole gas-liquid flow environment is divided. The specific division process is: if the gas-liquid flow environment variation coefficient is greater than or equal to the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a highly dynamic nonlinear change environment; if the gas-liquid flow environment variation coefficient is less than the gas-liquid flow environment variation coefficient reference threshold, the downhole gas-liquid flow environment under the monitoring window is divided into a normal change environment.

8. The method for optimizing the performance of a downhole gas-liquid separator based on an artificial neural network according to claim 1, characterized in that: In view of the highly dynamic nonlinear changing environment, the internal parameters of the gas-liquid separator are adjusted in real time through fuzzy logic control to further optimize the working performance of the downhole gas-liquid separator. The specific steps are as follows: The generated flow fluctuation reference value and pressure fluctuation reference value are passed as input to the fuzzy logic controller, which classifies the input data through fuzzification processing, maps the violent fluctuations of flow and pressure to the high fluctuation state in the fuzzy set, and maps the stable data to the low fluctuation state; The fuzzy logic controller performs reasoning operations based on a pre-set fuzzy rule base, where the rules define the adjustment strategy of the system under different fluctuation conditions; According to the inference results, the fuzzy logic controller defuzzifies the fuzzified output and converts it into clear control operation instructions, and adjusts the internal parameters of the separator through the system; After the system completes the parameter adjustment, it continuously monitors the working status of the gas-liquid separator and evaluates the adjustment effect through a real-time feedback mechanism. If the fluctuation trend of the flow fluctuation reference value and the pressure fluctuation reference value changes, the fuzzy logic controller is immediately used to re-analyze and adjust the parameters again to ensure that the separator maintains the best working state in a dynamically changing environment. The closed-loop control mechanism enables the system to be continuously optimized under complex working conditions.

9. A downhole gas-liquid separator performance optimization system based on artificial neural network, used to implement the downhole gas-liquid separator performance optimization method based on artificial neural network as described in any one of claims 1 to 8, characterized in that: It includes real-time data acquisition module, data analysis and intelligent prediction module, gas-liquid flow environment classification module, normal environment optimization module and dynamic nonlinear control module; The real-time data acquisition module obtains the flow rate information and pressure information in real time at the inlet of the separator when the underground gas-liquid mixture is passed into the gas-liquid separator; The data analysis and intelligent prediction module performs preliminary analysis on the flow rate and pressure information at the entrance of the gas-liquid separator, identifies the fluctuation trend of flow rate and pressure, and inputs the processed inlet flow rate and pressure data into the pre-trained deep learning model for intelligent prediction, thus discovering potential abnormal fluctuations in the underground gas-liquid flow environment in advance; The gas-liquid flow environment classification module divides the underground gas-liquid flow environment into a highly dynamic nonlinear change environment and a normal change environment based on the prediction results of the deep learning model; Normal environment optimization module, which uses a pre-trained artificial neural network model to optimize the performance of the gas-liquid separator for conditions predicted to be normal changes; The dynamic nonlinear control module, aimed at the highly dynamic nonlinear change environment, adjusts the internal parameters of the gas-liquid separator in real time through fuzzy logic control based on the highly dynamic nonlinear changes at the inlet of the gas-liquid separator, further optimizes the working performance of the downhole gas-liquid separator, and reduces the deviation between the predicted output and the actual working conditions.

Images