An experimental system for monitoring sand migration in wellbore based on distributed fiber optic acoustic sensing and a method for predicting migration velocity.

Abstract

This invention belongs to the field of sand production monitoring in oil and gas wellbore extraction, specifically involving an experimental system for monitoring sand migration in wellbore based on distributed fiber optic acoustic sensing and a method for predicting migration velocity. Based on the established experimental system for monitoring sand migration in wellbore using distributed fiber optic acoustic sensing, this invention obtains sand migration velocity data under different operating conditions through numerous indoor experiments. It can study the influence of fluid velocity, fluid medium, sand concentration in the slurry, and sand diameter in the slurry on sand migration velocity. Simultaneously, it provides data for training a subsequent symbolic regression model for sand migration velocity. This invention uses experimental data to train a symbolic regression model for sand migration velocity in the wellbore, which can generate a mathematical expression for predicting sand migration velocity, providing a new method for calculating sand migration velocity.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of oil and gas production wellbore sand production monitoring, and particularly relates to a wellbore sand transport monitoring experimental system and a transport speed prediction method based on distributed optical fiber acoustic wave sensing. BACKGROUND

[0002] For a long time, sand production has been a major problem in the oil and gas industry. Due to the influence of artificial factors and address factors in the production process, a large amount of sand will be produced in sandstone reservoirs, and sand particles will enter the production pipeline with the production fluid, affecting the normal production of oil and gas wells. Therefore, sand production is one of the reasons affecting the sustained and stable production of oil and gas, and has brought great challenges to the development of many domestic oilfields.

[0003] The traditional sand production monitoring device has significant limitations, and there is a time delay in responding to wellbore sand production events, and it cannot accurately locate the sand production position of the reservoir. Especially in low production flow conditions, sand particles are easy to deposit and accumulate at the bottom of the well, and ground sensors cannot monitor such abnormalities.

[0004] Chinese patent document CN120672508A (202410315041.3) discloses a method, device and equipment for predicting critical sand production pressure difference of gas storage and storage medium, the method obtains the injection-production type of each oil and gas well associated with the target gas storage; wherein the injection-production type includes reused injection-production well, newly drilled well and single production well; according to the injection-production type of each oil and gas well, the reservoir parameters matched with each oil and gas well are determined; according to the reservoir parameters matched with each oil and gas well, the critical sand production pressure difference of each oil and gas well is determined. Although this patent can determine the critical sand production pressure difference to determine whether sand production occurs, it cannot quantitatively analyze the sand particle transport speed.

[0005] Chinese patent document CN114091224A (202110965238.8) discloses a method for predicting sand production range and degree of natural gas hydrate reservoir, and provides a method for predicting sand production range and degree of reservoir, which mainly relies on the characteristics of weakly cemented or uncemented hydrate reservoir in sea area, and determines the sand particle starting critical flow velocity through reservoir fluid erosion criterion. This patent only determines the starting critical flow velocity of sand particles, and cannot determine the sand particle transport speed according to the parameter changes in the wellbore in the subsequent sand particle transport.

[0006] Therefore, a new method is needed to calculate the sand particle transport speed in the wellbore and provide support for quantifying the sand particle transport speed in the wellbore. SUMMARY

[0007] To solve the above technical problems, the application provides a wellbore sand migration monitoring experimental system and a migration speed prediction method based on distributed optical fiber acoustic sensing.

[0008] The technical problems solved by the application are achieved by the following technical solutions: a wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic sensing, comprising a full-size wellbore system, a fluid supply and recovery system and an optical fiber monitoring system.

[0009] The full-size wellbore system comprises a full-size wellbore arranged horizontally, and a sand particle injection port arranged above the left end of the full-size wellbore.

[0010] The fluid supply and recovery system comprises a liquid supply unit, a mortar supply unit and a waste liquid recovery tank.

[0011] The liquid supply unit is connected to the left end of the full-size wellbore, and is used to control the experimental fluid to enter the full-size wellbore at a set flow rate.

[0012] The mortar supply unit is connected to the sand particle injection port, and is used to control the mortar with a preset ratio to enter the full-size wellbore at a set flow rate.

[0013] The waste liquid recovery tank is connected to the right end of the full-size wellbore, and is used to recover the mixture of the experimental fluid and the mortar.

[0014] The optical fiber monitoring system comprises a DAS data acquisition system, an in-pipe optical fiber and an out-pipe optical fiber, the in-pipe optical fiber is arranged on the inner wall of the full-size wellbore in the axial direction of the full-size wellbore, the out-pipe optical fiber is arranged on the outer wall of the full-size wellbore in the axial direction of the full-size wellbore, the right end of the in-pipe optical fiber penetrates through the full-size wellbore and is connected to the out-pipe optical fiber, the left end of the in-pipe optical fiber penetrates through the full-size wellbore and is connected to the DAS data acquisition system, and the tail part of the out-pipe optical fiber is subjected to reflection elimination treatment.

[0015] Preferably, the liquid supply unit comprises a liquid supply tank, a water supply line, a centrifugal pump, a check valve and a ball valve.

[0016] The liquid supply tank is connected to the full-size wellbore through the water supply line, and the centrifugal pump, the check valve and the ball valve are arranged on the water supply line between the liquid supply tank and the full-size wellbore.

[0017] The liquid supply tank is connected with a centrifugal pump, a check valve and a ball valve in sequence.

[0018] Preferably, the mortar supply unit comprises a mortar mixing tank and a diaphragm pump, and the mortar mixing tank is connected with the sand injection inlet through the diaphragm pump. The diaphragm pump changes the volume of the working chamber by reciprocating deformation of the diaphragm to deliver the mortar in the mortar mixing tank to the full-size wellbore.

[0019] Preferably, the optical fiber tail outside the pipe is provided with an optical fiber tail end processing device to make the light "smoothly" transition into the air, thereby reducing the high reflection of the tail end of the optical fiber link.

[0020] The application further discloses a wellbore sand migration speed prediction method based on distributed optical fiber acoustic wave sensing.

[0021] S1, a wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic wave sensing is built;

[0022] S2, different working condition sand migration experiments are carried out by setting the fluid speed of the experimental fluid, the fluid medium of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter in the mortar;

[0023] S3, the original data acquired by the DAS data acquisition system are filtered and denoised, and the frequency band energy FBE is calculated;

[0024] S4, the sand particle migration speed is calculated by using the frequency band energy FBE.

[0025] The application can study the influence of the fluid speed of the experimental fluid, the fluid medium of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter in the mortar on the sand particle migration speed by adjusting the fluid speed of the experimental fluid, the fluid medium of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter in the mortar through a single test.

[0026] The application further discloses a wellbore sand migration speed prediction method based on distributed optical fiber acoustic wave sensing.

[0027] S1, a wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic wave sensing is built;

[0028] S2, different working condition sand migration experiments are carried out by setting the fluid speed of the experimental fluid, the fluid medium of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter in the mortar;

[0029] S3, filtering and denoising the raw data obtained by the DAS data acquisition system and calculating the frequency band energy FBE;

[0030] S4, calculating the sand particle migration speed using the frequency band energy FBE;

[0031] S5, preprocessing all experimental data in steps S2-S4 to meet the modeling requirements;

[0032] S7, after the experimental data preprocessing is completed, taking the sand particle migration speed as the target variable, using the random forest model to select the characteristic variables in the experimental data which are arranged in the preset ranking before the sand particle migration speed, and dividing the selected experimental data into a training set and a test set according to a certain proportion;

[0033] S8, using the selected characteristic variables and the corresponding sand particle migration speed to establish a symbolic regression model:

[0034] Configure the search space and evolution parameters of the symbolic regression model, allow to use addition, subtraction, multiplication, division and square root functions as basic operators, set the maximum length of the formula, the number of iterations of the evolution process, the number of candidate formulas generated each iteration, and the random seed;

[0035] In the training process, the candidate formulas are optimized using cross, mutation and selection three operation methods, and the obtained candidate formulas are used to calculate the sand particle migration speed with the selected characteristic variables;

[0036] S9, model evaluation and verification: using the test set to calculate the sand particle migration speed of the candidate formula obtained in step S8, obtaining the predicted value of the sand particle migration speed, calculating the difference between the predicted value and the actual value using mean square error and R² determination coefficient, and quantifying the model performance;

[0037] If the model performance does not meet the preset requirements, perform hyperparameter optimization to obtain the best equation of symbolic regression;

[0038] S10, after obtaining the best equation of symbolic regression, encapsulating it as a sand particle migration speed calculation function, and using the sand particle migration speed calculation function to read the related characteristic variables in the experimental data, the calculation of the sand particle migration speed can be realized.

[0039] Preferably, the specific steps of step S2 are:

[0040] S2.1, first keep the sand particle concentration of the mortar, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, and set different fluid velocities;

[0041] S2.2, during the experiment, first adjust the flow rate by the liquid supply unit, and after the flow is stable, the mortar is injected into the full-size wellbore at a preset speed by the mortar supply unit;

[0042] S2.3, the DAS data acquisition system acquires data of N minutes after the flow is stable;

[0043] S2.4, keeping the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, the fluid medium of the experimental fluid and the sand particle diameter in the mortar being the same as in step S2.1, the fluid velocity of the experimental fluid being the same as any of the velocities in step S2.1, and the mortar being provided with a sand particle concentration different from that in step S2.1, repeating steps S2.2 and S2.3;

[0044] S2.5, keeping the fluid velocity of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter unchanged, the fluid velocity of the experimental fluid being the same as any of the velocities in step S2.1, the sand particle concentration in the mortar and the sand particle diameter being the same as in step S2.1, and the fluid medium being provided with a fluid medium different from that in step S2.1, repeating steps S2.2 and S2.3;

[0045] S2.6, keeping the fluid velocity of the experimental fluid, the fluid medium and the sand particle concentration in the mortar unchanged, the fluid velocity of the experimental fluid being the same as any of the velocities in step S2.1, the fluid velocity of the experimental fluid and the sand particle concentration in the mortar being the same as in step S2.1, and the mortar being provided with a sand particle diameter different from that in step S2.1, repeating steps S2.2 and S2.3.

[0046] Preferably, the specific steps of step S3 are as follows:

[0047] The original DAS signal collected is subjected to filtering processing to remove noise influence, and the sand particle migration frequency range is determined on a frequency spectrum by recording the time when the pump and the valve are opened during the experiment;

[0048] The signal is converted from the time domain to the frequency domain through fast Fourier transform;

[0049] After the sand particle migration frequency range is determined, the frequency band energy FBE of the signal of the inner wall and the outer wall of the full-size wellbore is calculated, and the calculation method is as shown in formula (1):

[0050] (1)

[0051] In the formula, E(t) represents the frequency band characteristic energy of the time length t, the upper and lower limits a and b of the integral are the end point values of the determined sand particle migration frequency range, and s(f) is the frequency domain signal after Fourier transform.

[0052] Preferably, the specific steps of step S4 are as follows: based on the frequency band energy FBE data of the inner wall of the full-size wellbore, the trajectory slope is calculated through the FBE time-space waterfall diagram, and the sand particle migration velocity is derived;

[0053] The sand particle migration velocity is calculated by using the band energy FBE of the full-size wellbore outer wall in the same way, and the accuracy of the experimental data is verified.

[0054] Preferably, the specific method of step S5 is: all non-numeric columns in the experimental data in steps S2-S4 are forcibly converted into numeric types, and the failure is set as a null value, and the columns with all null values are deleted.

[0055] An IQR outlier processing method is used to calculate the first quartile Q1 and the third quartile Q3, obtain the interquartile range IQR=Q3-Q1, and set the outlier boundary, the lower boundary is Q1-1.5*IQR, and the upper boundary is Q3+1.5*IQR.

[0056] In step S8, two formulas are randomly selected and their subtree structures are exchanged.

[0057] The mutation indicates that the operator or constant in the formula is randomly modified.

[0058] The selection indicates that the comprehensive loss is calculated according to the error and complexity of the formula, the high-loss formula is eliminated by comparison with the loss threshold, the high-quality individual is reserved to enter the next generation, and finally the equation with the minimum comprehensive loss is extracted from all candidate formulas as the best model.

[0059] Compared with the prior art, the present application has the following beneficial effects:

[0060] 1. The wellbore sand particle migration monitoring experimental system based on distributed optical fiber acoustic sensing constructed by the present application can simulate sand particle migration under different working conditions, and provide data for subsequent training of the sand particle migration speed symbol regression model.

[0061] 2. The present application provides a new data analysis method, which trains a machine learning model based on DAS monitoring data, can effectively predict the sand particle migration speed, and forms a mathematical expression, realizes the intelligentization and quantization of sand particle migration speed prediction.

[0062] 3. The present application avoids the dependence of the traditional model on prior assumptions by introducing a symbol regression algorithm, and realizes the explicit expression of the sand particle migration speed prediction model.

[0063] 4. The method of the present application is convenient to operate, after the model training is completed, the trained model is deployed to the production environment, and only the DAS monitoring data needs to be processed by the computer to determine the sand particle migration speed, which is suitable for the prediction of sand particle migration speed under different working conditions.

[0064] The sand particle migration speed prediction method based on the symbolic regression algorithm can effectively mine the nonlinear relationship between data, generate a prediction model with a clear mathematical expression, not only reveal the internal relationship between variables, and the expression form is intuitive and convenient for engineering application. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 It is a structural schematic diagram of the wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic wave sensing.

[0066] Figure 2 It is a working process schematic diagram of the present application.

[0067] In the figure, the full-size wellbore 1, the sand particle injection port 11;

[0068] The liquid supply unit 2, the mortar supply unit 3, and the waste liquid recovery tank 4;

[0069] The DAS data acquisition system 5, the in-pipe optical fiber 6, the out-of-pipe optical fiber 7, and the optical fiber tail end processing device 8;

[0070] The liquid supply tank 201, the water supply line 202, the centrifugal pump 203, the check valve 204, and the ball valve 205;

[0071] The mortar mixing tank 301 and the diaphragm pump 302. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings.

[0073] In oil and gas exploitation, sand production from sandstone reservoirs caused by formation and human factors, sand particles entering the wellbore with production fluid can seriously affect normal production and damage equipment. Therefore, accurate monitoring and quantitative prediction of sand particle migration are crucial to ensure the safe production of oil and gas wells, prolong the service life, and optimize the production capacity. The present application uses a wellbore sand particle migration monitoring experimental system based on distributed optical fiber acoustic wave sensing to simulate the sand particle migration process under different working conditions and collect data. Based on the experimental data, a kind of interpretable wellbore sand particle migration speed prediction method is further constructed by using symbolic regression algorithm. The method can predict the sand particle migration speed in real time according to the monitoring data of distributed optical fiber.

[0074] The present application uses the experimental data obtained by the DAS experimental device to train the machine learning model, providing an efficient method for sand production monitoring and quantitative sand particle migration speed.

[0075] As Figure 1 shown, a wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic wave sensing, characterized by comprising a full-size wellbore system, a fluid supply and recovery system, and an optical fiber monitoring system.

[0076] The full-size wellbore system comprises a full-size wellbore 1 arranged horizontally for simulating a horizontal well, and a sand injection inlet 11 arranged above the left end of the full-size wellbore 1.

[0077] The fluid supply and recovery system comprises a liquid supply unit 2, a mortar supply unit 3, and a waste liquid recovery tank 4.

[0078] The liquid supply unit 2 is connected to the left end of the full-size wellbore 1, and is used to control the experimental fluid to enter the full-size wellbore 1 at a set flow rate.

[0079] The mortar supply unit 3 is connected to the sand injection inlet 11, and is used to control the mortar of a preset ratio to enter the full-size wellbore 1 at a set flow rate.

[0080] The waste liquid recovery tank 4 is connected to the right end of the full-size wellbore 1, and is used to recover the mixture of the experimental fluid and the mortar. The waste liquid recovery tank 4 is a plastic storage tank with a volume of 1 cubic meter, and is responsible for recovering the mixture of the experimental fluid and the sand particles.

[0081] The optical fiber monitoring system comprises a DAS data acquisition system 5, an in-pipe optical fiber 6, and an out-pipe optical fiber 7. The in-pipe optical fiber 6 is arranged axially on the inner wall of the full-size wellbore 1, and the out-pipe optical fiber 7 is arranged axially on the outer wall of the full-size wellbore 1. The right end of the in-pipe optical fiber 6 penetrates through the full-size wellbore 1 and is connected to the out-pipe optical fiber 7, and the left end of the in-pipe optical fiber 6 penetrates through the full-size wellbore 1 and is connected to the DAS data acquisition system 5. The tail of the out-pipe optical fiber 7 is subjected to reflection elimination treatment. Specifically, the optical fiber is installed closely to the inner and outer walls of the full-size wellbore 1. The optical fiber penetrates through the full-size wellbore 1 from the inside and is wound back from the outside, so as to simultaneously measure the sound signals of the inner and outer walls of the pipe. During the experiment, the sound signals of the inner wall are used to calculate the sand particle migration speed, and the signals of the outer wall are used to verify the correctness of the sand particle migration speed. One end of the optical fiber is connected to the DAS data acquisition system 5, and the other end is provided with an optical fiber tail processing device 8.

[0082] The liquid supply unit 2 comprises a liquid supply tank 201, a water supply line 202, a centrifugal pump 203, a check valve 204, and a ball valve 205.

[0083] The liquid supply tank 201 is connected to the full-size wellbore 1 through the water supply line 202. The centrifugal pump 203, the check valve 204, and the ball valve 205 are arranged on the water supply line 202 between the liquid supply tank 201 and the full-size wellbore 1.

[0084] The liquid supply tank 201 is connected to the centrifugal pump 203, the check valve 204, and the ball valve 205 in sequence.

[0085] The mortar supply unit 3 comprises a mortar mixing tank 301 and a diaphragm pump 302. The mortar mixing tank 301 is connected to the sand injection inlet 11 through the diaphragm pump 302.

[0086] The outer optical fiber 7 is equipped with an optical fiber tail processing device 8.

[0087] like Figure 2 As shown, a wellbore sand migration velocity prediction method based on distributed fiber optic acoustic sensing, using the aforementioned wellbore sand migration monitoring experimental system based on distributed fiber optic acoustic sensing, includes the following steps:

[0088] S1. An experimental system for monitoring well sand transport based on distributed fiber optic acoustic sensing was constructed.

[0089] S2. By setting the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid, the sand concentration in the mortar, and the sand diameter in the mortar, sand transport experiments under different working conditions are carried out.

[0090] The specific steps of step S2 are as follows:

[0091] S2.1 First, keep the sand concentration of the mortar, the fluid medium of the experimental fluid, and the diameter of the sand particles in the mortar constant, and set different fluid velocities.

[0092] S2.2 During the experiment, the flow rate is first adjusted by the liquid supply unit 2. After the flow stabilizes, the mortar supply unit 3 injects the mortar into the full-size well shaft 1 at a preset speed. In this embodiment, the mortar supply unit 3 controls the set flow rate of the mortar entering the full-size well shaft 1 to remain constant. The specific set value can be set by the user.

[0093] S2.3, DAS data acquisition system 5 collects data N minutes after the flow stabilizes. The value of N can be set by the user; in this embodiment, N is 2.

[0094] S2.4 Keep the fluid velocity, fluid medium, and sand particle diameter in the mortar constant. The fluid medium and sand particle diameter in the mortar are the same as in step S2.1. The fluid velocity of the experimental fluid is the same as any velocity in step S2.1. Set the sand particle concentration in the mortar to be different from that in step S2.1. Repeat steps S2.2 and S2.3.

[0095] S2.5. Keep the fluid velocity of the experimental fluid, the sand concentration and sand diameter in the mortar constant, the fluid velocity of the experimental fluid the same as any velocity in step S2.1, the sand concentration and sand diameter in the mortar the same as in step S2.1, set a fluid medium different from that in step S2.1, and repeat steps S2.2 and S2.3.

[0096] S2.6, keeping the fluid velocity of the experimental fluid, the fluid medium, and the sand concentration in the mortar unchanged, the fluid velocity of the experimental fluid being the same as any of the velocities in step S2.1, the fluid velocity of the experimental fluid and the sand concentration in the mortar being the same as in step S2.1, the sand grain diameter of the mortar being different from that in step S2.1, repeating steps S2.2 and S2.3.

[0097] S3, filtering and denoising the raw data acquired by the DAS data acquisition system 5 and calculating the band energy FBE.

[0098] The specific steps of step S3 are as follows:

[0099] The acquired raw DAS signals are filtered to remove noise effects. In the experiment, the time when the pump and the valve are opened is recorded, and the sand particle migration frequency range is determined on the frequency spectrum. In this embodiment, the pump is a centrifugal pump 203 and a diaphragm pump 302, and the valve is a check valve 204 and a ball valve 205.

[0100] The signal is converted from the time domain to the frequency domain by fast Fourier transform.

[0101] After the sand particle migration frequency range is determined, the band energy FBE of the signals on the inner wall and the outer wall of the full-size wellbore 1 is calculated, and the calculation method is as formula (1):

[0102] (1)

[0103] In the formula, E(t) represents the band characteristic energy with a time length of t, the upper and lower limits a and b of the integral are the end point values of the determined sand particle migration frequency range, and s(f) is the frequency domain signal after Fourier transform.

[0104] S4, calculating the sand particle migration velocity by using the band energy FBE.

[0105] The FBE reveals the frequency distribution characteristics of the signal and is used to identify the vibration type. Low-frequency energy usually corresponds to ambient noise and noise during pump operation; high-frequency energy usually corresponds to noise of sand migration in the wellbore.

[0106] The specific steps of step S4 are as follows: based on the band energy FBE data of the inner wall of the full-size wellbore 1, the trajectory slope (i.e., the ratio of distance change to time change) is calculated by FBE time-space waterfall diagram, so that the sand particle migration velocity can be directly derived. The band energy FBE of the outer wall is used to calculate the sand particle migration velocity by the same method, which is used to verify the accuracy of the experimental data.

[0107] As Figure 2As shown, a wellbore sand transport velocity prediction method based on distributed optical fiber acoustic sensing adopts the above-mentioned wellbore sand transport monitoring experimental system based on distributed optical fiber acoustic sensing. The above-mentioned method mainly includes two parts of wellbore sand particle transport monitoring experiment and construction of sand particle transport velocity prediction model. After obtaining the sand particle transport velocity data through the experiment, the construction of the sand particle transport velocity prediction model is carried out by using the experimental data, including the following steps:

[0108] S1, the wellbore sand transport monitoring experimental system based on distributed optical fiber acoustic sensing is built.

[0109] S2, by setting the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid, the sand particle concentration in the mortar and the sand particle diameter in the mortar, the sand transport experiment under different conditions is carried out.

[0110] The specific steps of step S2 are:

[0111] S2.1, first keep the sand particle concentration of the mortar, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, and set different fluid velocities.

[0112] S2.2, during the experiment, first adjust the flow rate by the liquid supply unit 2, and then inject the mortar into the full-size wellbore 1 at the preset speed by the mortar supply unit 3 after the flow is stable. In this embodiment, the mortar supply unit 3 controls the set flow rate of the mortar entering the full-size wellbore 1 to be constant, and the specific set value is set by itself.

[0113] S2.3, the DAS data acquisition system 5 collects data 2 minutes after the flow is stable.

[0114] S2.4, keep the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, the fluid medium of the experimental fluid and the sand particle diameter in the mortar are the same as step S2.1, the fluid velocity of the experimental fluid is the same as any velocity of step S2.1, the mortar is set to have a different sand particle concentration from step S2.1, and steps S2.2 and S2.3 are repeated.

[0115] S2.5, keep the fluid velocity of the experimental fluid, the sand particle concentration and the sand particle diameter in the mortar unchanged, the fluid velocity of the experimental fluid is the same as any velocity of step S2.1, the sand particle concentration and the sand particle diameter in the mortar are the same as step S2.1, and a fluid medium different from step S2.1 is set, and steps S2.2 and S2.3 are repeated.

[0116] S2.6, keeping the fluid velocity of the experimental fluid, the fluid medium, and the sand particle concentration in the mortar unchanged, the fluid velocity of the experimental fluid being the same as any of the velocities in step S2.1, the fluid velocity of the experimental fluid and the sand particle concentration in the mortar being the same as in step S2.1, the sand particles in the mortar being provided with a different sand particle diameter from that in step S2.1, repeating steps S2.2 and S2.3.

[0117] S3, filtering and denoising the raw data acquired by the DAS data acquisition system 5 and calculating the band energy FBE.

[0118] The specific steps of step S3 are as follows:

[0119] The acquired raw DAS signals are filtered to remove noise effects. During the experiment, the time when the pump and the valve are opened is recorded, and the sand particle migration frequency range is determined on the frequency spectrum.

[0120] The signal is converted from the time domain to the frequency domain through fast Fourier transform.

[0121] After determining the sand particle migration frequency range, the band energy FBE of the signals on the inner wall and the outer wall of the full-size wellbore 1 is calculated, and the calculation method is as shown in formula (1):

[0122] (1)

[0123] In the formula, E(t) represents the band characteristic energy with a time length of t, the upper and lower limits a and b of the integral are the end point values of the determined sand particle migration frequency range, and s(f) is the frequency domain signal after Fourier transform.

[0124] S4, calculating the sand particle migration velocity by using the band energy FBE.

[0125] S5, data preprocessing: all experimental data in steps S2-S4 are preprocessed to meet the modeling requirements.

[0126] Specifically, the specific method of step S5 is as follows: all non-numeric columns in all experimental data in steps S2-S4 are forcibly converted to numeric types, and if the conversion fails, the column is set to null. The columns with all null values are deleted.

[0127] The IQR outlier processing method is adopted to calculate the first quartile Q1 and the third quartile Q3, obtain the interquartile range IQR=Q3-Q1, and set the outlier boundary, with the lower limit being Q1-1.5×IQR and the upper limit being Q3+1.5×IQR.

[0128] S7, data division and feature analysis: after the experimental data preprocessing is completed, the sand particle migration speed is taken as the target variable, the random forest model is used to select the characteristic variables in the experimental data which are related to the sand particle migration speed and arranged before the preset name, and the selected experimental data is divided into a training set and a test set according to a certain proportion.

[0129] Specifically, after the data preprocessing is completed, the characteristic variables and the target variables required for modeling need to be determined first, for example, the characteristic variables are selected from the frequency band energy FBE, the fluid speed in the full-size wellbore pipeline, the sand particle concentration, the sand particle diameter, the pipeline diameter of the full-size wellbore, the Reynolds number, etc., and the target variable in this embodiment is the sand particle migration speed.

[0130] The selection of the characteristic variables uses the random forest model to analyze the correlation degree, calculates the splitting contribution of each feature in the decision tree, and outputs the importance ranking. According to the importance ranking, three parameters closely related to the target variable are selected, for example, the frequency band energy FBE, the fluid speed in the pipeline and the Reynolds number are selected, and the remaining characteristic variables are eliminated due to low contribution, and a model is established through these closely related characteristic variables to predict the sand particle migration speed.

[0131] Subsequently, the data set is divided into a training set and a test set according to a certain proportion (for example, 8:2), and a fixed random seed is set to ensure that the division result is consistent each time, facilitating the reproduction of the experiment.

[0132] S8, symbolic regression model establishment: the selected characteristic variables and the corresponding sand particle migration speed are used to establish a symbolic regression model:

[0133] The search space and evolution parameters of the symbolic regression model are configured, allowing the use of addition, subtraction, multiplication, division and square root functions as basic operators, setting the maximum length of the formula, the number of iterations of the evolution process, the number of candidate formulas generated each iteration, and the random seed.

[0134] Specifically, the core of symbolic regression is to search for mathematical expressions through evolution algorithm. First, the search space and evolution parameters of the model are configured, allowing the use of addition, subtraction, multiplication, division and square root functions as basic operators, limiting the maximum length of the formula to 20 nodes to prevent the generation of overly complex expressions. The evolution process is set to 50 iterations, 1000 candidate formulas are generated each generation, and the error and formula complexity are balanced through a simplicity penalty coefficient to avoid overfitting, and the simplicity penalty coefficient in this embodiment is the maximum length of the formula. To ensure the reproducibility of the experiment, the random seed is fixed, and the value of the random seed is randomly selected, which is not specifically limited here.

[0135] In the training process, the three operation methods of crossover, mutation and selection are used to optimize the candidate formula, and the obtained candidate formula is used to calculate the sand particle migration speed with the selected characteristic variable.

[0136] The symbolic regression model in the application is an existing model, the three operation methods of crossover, mutation and selection are all model built-in methods, and only the hyperparameters need to be adjusted in actual operation.

[0137] In step S8, crossover means randomly selecting two formulas and exchanging their subtree structures. For example, the addition node of the formula X+2 is exchanged with the multiplication node of Y*3 to generate a new formula X*3+Y+2.

[0138] Mutation means randomly modifying the operator or constant in the formula. For example, changing + to * or adjusting the constant 2 to 5, which introduces diversity to avoid falling into local optimum.

[0139] Selection means calculating the comprehensive loss according to the error and complexity of the formula, eliminating high-loss formulas by comparing with the loss threshold, and retaining high-quality individuals to enter the next generation. Finally, the equation with the minimum comprehensive loss is extracted from all candidate formulas as the best model, so that the equation has both prediction accuracy and physical interpretability.

[0140] S9, model evaluation and verification: the candidate formula obtained in step S8 is used to calculate the sand particle migration speed, and the difference between the predicted value and the actual value is calculated by using the mean square error and the R² determination coefficient to quantify the model performance.

[0141] If the model performance does not meet the preset requirements, the hyperparameter optimization is performed to obtain the best equation of symbolic regression. In the embodiment, the hyperparameter optimization specifically comprises: increasing the population size, adjusting the operator combination of the symbolic regression model, or increasing the iteration number to optimize the model.

[0142] S10, result saving: after obtaining the best equation of symbolic regression, it is encapsulated as a sand particle migration speed calculation function, and the related characteristic variables in the experimental data are read by using the sand particle migration speed calculation function, so that the calculation of the sand particle migration speed can be realized.

[0143] Specifically, all generated candidate formulas, best equations and performance indicators need to be saved to a text file for subsequent analysis and reporting. The file content includes: all candidate formula list, best equation and performance indicators (MSE and R²).

[0144] After the best equation of symbolic regression is obtained, it can be encapsulated as a function of sand particle migration velocity calculation. The experimental data is arranged into a csv format file, and the sand particle migration velocity calculation function is used to read the experimental data to calculate the sand particle migration velocity.

Claims

1. A method for predicting the velocity of sand migration in a wellbore based on distributed fiber optic acoustic sensing, using a distributed fiber optic acoustic sensing-based experimental system for monitoring sand migration in a wellbore, characterized by, The wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic sensing comprises a full-size wellbore system, a fluid supply and recovery system and an optical fiber monitoring system; The full-size wellbore system comprises a horizontally arranged full-size wellbore (1), and a sand injection inlet (11) is arranged above the left end of the full-size wellbore (1); The fluid supply and recovery system comprises a liquid supply unit (2), a mortar supply unit (3) and a waste liquid recovery tank (4); The liquid supply unit (2) is connected with the left end of the full-size wellbore (1) and is used for controlling the experimental fluid to enter the full-size wellbore (1) at a set flow rate; The mortar supply unit (3) is connected with the sand injection inlet (11) and is used for controlling the mortar with a preset ratio to enter the full-size wellbore (1) at a set flow rate; The waste liquid recovery tank (4) is connected with the right end of the full-size wellbore (1) and is used for recovering the mixture of the experimental fluid and the mortar; The optical fiber monitoring system comprises a DAS data acquisition system (5), an in-pipe optical fiber (6) and an out-pipe optical fiber (7), the in-pipe optical fiber (6) and the out-pipe optical fiber (7) are installed close to the inner and outer walls of the full-size wellbore (1), the in-pipe optical fiber (6) is arranged on the inner wall of the full-size wellbore (1) along the axis of the full-size wellbore (1) and is used for measuring the acoustic signal of sand migration, the out-pipe optical fiber (7) is arranged on the outer wall of the full-size wellbore (1) along the axis of the full-size wellbore (1) and is used for obtaining a verification signal for verifying the acoustic signal, the right end of the in-pipe optical fiber (6) penetrates through the full-size wellbore (1) and is connected with the out-pipe optical fiber (7), the left end of the in-pipe optical fiber (6) penetrates through the full-size wellbore (1) and is connected with the DAS data acquisition system (5), and the tail part of the out-pipe optical fiber (7) is provided with an optical fiber tail end processing device (8); The method comprises the following steps: S1, building the wellbore sand migration monitoring experimental system based on distributed optical fiber acoustic sensing; S2, setting the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid, the sand concentration in the mortar and the sand diameter in the mortar, carrying out sand migration experiments under different working conditions, and obtaining experimental data comprising a frequency band energy FBE, a fluid velocity in a pipeline, a sand concentration, a sand diameter, a pipeline diameter and a Reynolds number; S3, filtering and denoising the original data acquired by the DAS data acquisition system (5) and calculating the frequency band energy FBE; S4, based on the frequency band energy FBE data of the inner wall of the full-size wellbore (1), calculating the trajectory slope through a FBE time-space waterfall diagram and deducing the sand migration velocity; using the same method to calculate the sand migration velocity based on the frequency band energy FBE of the outer wall of the full-size wellbore (1) and verifying the accuracy of the experimental data; S5, pre-processing all the experimental data in steps S2-S4 to meet the modeling requirements, specifically, using an IQR outlier processing method to pre-process the experimental data; S7, after the experimental data pre-processing is completed, taking the sand migration velocity as a target variable, using a random forest model to select feature variables in the experimental data which are arranged in a preset ranking before a sand migration velocity correlation degree, dividing the selected experimental data into a training set and a test set according to a certain proportion; S8, the selected characteristic variable and the corresponding sand particle migration velocity are used to establish a symbolic regression model: Configure the search space and evolution parameters of the symbolic regression model, allow the use of addition, subtraction, multiplication, division and square root functions as basic operators, set the maximum length of the formula, the number of iterations of the evolution process, the number of candidate formulas generated each iteration, and the random seed; During the training process, the candidate formula is optimized using cross, mutation and selection operation methods, and the candidate formula is obtained by selecting the characteristic variable to calculate the sand particle migration velocity; S9, model evaluation and verification: using the test set to calculate the sand particle migration velocity of the candidate formula obtained in step S8, and obtaining the predicted value of the sand particle migration velocity, using the mean square error and R² determination coefficient to calculate the difference between the predicted value and the actual value, and quantifying the model performance; If the model performance does not meet the preset requirements, perform hyperparameter optimization to obtain the best equation of symbolic regression; S10, after obtaining the best equation of symbolic regression, it is packaged as a sand particle migration velocity calculation function, and the relevant characteristic variables in the experimental data are read using the sand particle migration velocity calculation function, so that the calculation of the sand particle migration velocity can be realized.

2. The distributed optical fiber acoustic sensing based wellbore sand migration velocity prediction method of claim 1, wherein, The specific steps of step S2 are: S2.1, keep the sand concentration of the mortar, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, and set different fluid velocities; S2.2, during the experiment, first adjust the flow rate by the liquid supply unit (2), and after the flow is stable, the mortar is injected into the full-size wellbore (1) at a predetermined speed by the mortar supply unit (3); S2.3, the DAS data acquisition system (5) collects data for N minutes after the flow is stable; S2.4, keep the fluid velocity of the experimental fluid, the fluid medium of the experimental fluid and the sand particle diameter in the mortar unchanged, the fluid medium of the experimental fluid and the sand particle diameter in the mortar are the same as in step S2.1, the fluid velocity of the experimental fluid is the same as any velocity in step S2.1, and the sand concentration of the mortar is different from that in step S2.1, repeat steps S2.2 and S2.3; S2.5, keep the fluid velocity of the experimental fluid, the sand concentration and the sand particle diameter in the mortar unchanged, the fluid velocity of the experimental fluid is the same as any velocity in step S2.1, the sand concentration and the sand particle diameter in the mortar are the same as in step S2.1, and the fluid medium is set to be different from that in step S2.1, repeat steps S2.2 and S2.3; S2.6, keep the fluid velocity of the experimental fluid, the fluid medium and the sand concentration in the mortar unchanged, the fluid velocity of the experimental fluid is the same as any velocity in step S2.1, the fluid velocity of the experimental fluid and the sand concentration in the mortar are the same as in step S2.1, and the sand particle diameter of the mortar is set to be different from that in step S2.1, repeat steps S2.2 and S2.

3.

3. The distributed optical fiber acoustic sensing based wellbore sand migration velocity prediction method of claim 1, wherein, The specific steps of step S3 are: The collected original DAS signal is filtered to remove noise, and the time when the pump and valve are opened is recorded during the experiment to determine the sand particle migration frequency range on the frequency spectrum; Through fast Fourier transform, the signal is converted from time domain to frequency domain; After determining the sand migration frequency range, the full-size wellbore (1) inner wall and outer wall signal frequency band energy FBE is calculated, and the calculation method is as shown in formula (1): （1） In the formula, E(t) represents the frequency band characteristic energy of the time length t, the upper and lower limits a and b of the integral are the end point values of the determined sand migration frequency range, and s(f) is the frequency domain signal after Fourier transform.

4. The distributed optical fiber acoustic sensing based wellbore sand migration velocity prediction method of claim 1, wherein, The specific method of step S5 is: forcibly converting the non-numeric columns in all experimental data in steps S2-S4 to a numeric type, setting to null if failed, and deleting columns with all null values; An IQR outlier processing method is used to calculate the first quartile Q1 and the third quartile Q3, obtain the interquartile range IQR=Q3-Q1, and set the outlier boundary, the lower boundary is Q1-1.5*IQR, and the upper boundary is Q3+1.5*IQR; In step S8, crossover means randomly selecting two formulas and exchanging their sub-tree structures; Mutation means randomly modifying the operators or constants in the formula; Selection means calculating the comprehensive loss according to the error and complexity of the formula, eliminating high-loss formulas by comparing with the loss threshold, keeping high-quality individuals to enter the next generation, and finally extracting the equation with the minimum comprehensive loss from all candidate formulas as the best model.

5. The distributed optical fiber acoustic sensing based wellbore sand migration velocity prediction method of claim 1, wherein, The liquid supply unit (2) comprises a liquid supply tank (201), a water supply line (202), a centrifugal pump (203), a check valve (204), and a ball valve (205); The liquid supply tank (201) is connected with the full-size wellbore (1) through the water supply line (202), and the centrifugal pump (203), the check valve (204), and the ball valve (205) are arranged on the water supply line (202) between the liquid supply tank (201) and the full-size wellbore (1); The liquid supply tank (201) is connected with the centrifugal pump (203), the check valve (204), and the ball valve (205) in sequence.

6. The distributed optical fiber acoustic sensing based wellbore sand migration velocity prediction method of claim 1, wherein: The mortar supply unit (3) comprises a mortar mixing tank (301) and a diaphragm pump (302), and the mortar mixing tank (301) is connected with the sand injection inlet (11) through the diaphragm pump (302).

Images