A prediction and evaluation method for wellbore natural gas hydrate freezing and plugging

Abstract

The present invention relates to the field of natural gas hydrate freeze-blocking prediction and provides a method for predicting and evaluating natural gas hydrate freeze-blocking in wellbores. The method comprises: obtaining real-time environmental data in the wellbore, substituting the real-time environmental data into a set of thermodynamic equations to determine the natural gas hydrate generation rate, and simulating the generation rate to obtain a simulated amount of natural gas hydrate generation; obtaining historical environmental data and historical natural gas hydrate generation data collected by sensors to construct sample vector data; training a neural network model with the sample vector data, and then substituting the real-time environmental data into the data to obtain a modeled amount of natural gas hydrate generation; and determining predicted amounts of natural gas hydrate generation and predicted blockage rates at multiple locations and times in the wellbore based on the simulated and predicted generation amounts. The method combines the physical constraints of the generation process of the thermodynamic equations with the predictive capabilities of the neural network model for unknown situations, making the predicted values ​​more accurate and reliable.

Description

Technical Field

[0001] The present invention relates to the field of natural gas hydrate freeze-blocking prediction, and in particular to a method for predicting and evaluating wellbore natural gas hydrate freeze-blocking. Background Art

[0002] Natural gas hydrates (combustible ice) are ice-like crystalline substances formed by natural gas and water under high pressure and low temperature conditions. Since Davy's initial discovery of natural gas hydrates in 1810, extensive research has been conducted on the thermodynamic phase equilibrium conditions for their formation. Currently, there are four main methods for predicting natural gas hydrate formation conditions: 1. Experimental models are used to derive the required model based on the thermodynamic phase equilibrium theory of natural gas hydrate formation; 2. Statistical thermodynamic models of natural gas hydrates are developed based on field experience; 3. Semi-empirical or empirical models are fitted based on outdoor sampling, indoor experiments, or field testing of the pressure and temperature at which natural gas hydrates form; and 4. Graphical methods. Theoretical models of natural gas hydrate phase equilibrium have a strong theoretical foundation and high computational accuracy, but they are computationally complex and difficult to master. Statistical thermodynamic models are the most widely used models, but they are less generalizable. Graphical methods and empirical and semi-empirical models offer advantages such as simplicity and computational convenience, but they are limited in scope and lack high computational accuracy.

[0003] In the actual natural gas extraction process, complex reservoir conditions often accompanied by high pressure and low temperatures, coupled with the presence of large amounts of natural gas within the wellbore, increase the probability of natural gas hydrate formation. The formation and accumulation of natural gas hydrates can cause freeze-blocking of the wellbore, hindering gas flow and increasing internal wellbore pressure, leading to wellbore ruptures, equipment damage, and even serious accidents such as blowouts. This can significantly reduce natural gas recovery efficiency and increase production costs. Even if subsequent freeze-blocking treatment is performed, the well can still be temporarily shut down, wasting production time and causing significant economic losses.

[0004] For example, patent application number 202110736243.1 discloses a method for predicting natural gas hydrate formation conditions based on a polynomial neural network. This method accurately predicts natural gas hydrate formation conditions by establishing a polynomial neural network algorithm. However, this method is limited to natural gas transmission pipelines and the single natural gas hydrate production conditions, making it unsuitable for the complex natural gas hydrate formation environment within wellbores.

[0005] Therefore, there is currently no accurate and applicable natural gas hydrate freezing and plugging prediction technology for gas well bores. Summary of the Invention

[0006] To this end, the present invention provides a method for predicting and evaluating wellbore natural gas hydrate freezing and blockage, which determines the predicted natural gas hydrate generation amount and predicted blockage rate through a set of thermodynamic equations and a neural network model. It combines the predictive ability of the neural network model for unknown situations with the physical constraints of the natural gas hydrate formation process of the thermodynamic equations, making the predicted values ​​more accurate and reliable and enhancing the robustness of the prediction system.

[0007] To achieve the above objectives, the present invention proposes a method for predicting and evaluating wellbore natural gas hydrate freezing and plugging, comprising:

[0008] Acquiring real-time environmental data from the wellbore collected by sensors, establishing a set of thermodynamic equations for representing natural gas hydrate formation conditions, substituting the real-time environmental data into the set of thermodynamic equations to obtain a natural gas hydrate formation rate, and performing a simulation based on the formation rate to obtain a simulated amount of natural gas hydrate formed within a set time period;

[0009] Acquiring historical data and historical generation amounts of natural gas hydrates in the wellbore collected by sensors, establishing a neural network model for determining a relationship between the historical data and the historical generation amounts, training the neural network model using the historical data and the historical generation amounts as sample vector data, and then substituting the real-time environmental data into the neural network model to obtain a model generation amount of natural gas hydrates;

[0010] The predicted generation amount and predicted blockage rate of natural gas hydrate at multiple locations in the wellbore are determined according to the simulated generation amount and the predicted generation amount.

[0011] Furthermore, the thermodynamic equations include the Clapeyron equation and a thermodynamic model, and the real-time environmental data include at least temperature and interface area;

[0012] Substituting the temperature into the Clapeyron equation to determine the equilibrium pressure for natural gas hydrate formation;

[0013] Substituting the temperature into the thermodynamic model to determine the driving force for natural gas hydrate formation;

[0014] The rate of natural gas hydrate formation is determined according to the equilibrium pressure, the driving force, and the interfacial area.

[0015] Furthermore, the thermodynamic model is a VWP model;

[0016] The occupancy rate, escape coefficient and adsorption constant of gas molecules in the natural gas hydrate cage are calculated and substituted into the VWP model to determine the difference between the liquid phase chemical potential of water and the natural gas hydrate phase chemical potential. The difference is the driving force.

[0017] In the above scheme, the Clapeyron equation can accurately describe the behavior of gas during phase change, while the VWP model can accurately simulate the formation process of natural gas hydrates in porous media. Combining the two can derive a more accurate natural gas hydrate generation rate, which is particularly suitable for the complex environment of wellbores.

[0018] Furthermore, the real-time environmental data also includes pressure;

[0019] determining a corresponding rate coefficient according to the driving force;

[0020] The generation rate is obtained by multiplying the rate coefficient, the interfacial area, and the difference between the pressure and the equilibrium pressure.

[0021] Furthermore, the neural network model is a multi-layer feedforward neural network model based on an error back propagation algorithm, and the historical data includes environmental parameters in the wellbore, natural gas composition parameters, and database parameters;

[0022] The input layer neurons of the neural network model are environmental data, and the output layer nodes are the production amount of natural gas hydrates, which adopts a loss function with a regularization term;

[0023] Training the neural network model using normalized historical data and the historical generation amount as sample vectors to generate an objective function;

[0024] Substituting the real-time environmental data into the objective function obtains the model generation amount.

[0025] Furthermore, the historical data includes the volume fraction of each gas source component of the natural gas, temperature, pressure, wellbore depth, interface area, rate constant and flow rate of natural gas hydrate formation;

[0026] generating a basis function matrix according to the historical data, wherein the basis function matrix includes each of the historical data, the square of each of the historical data, and the product of two different historical data;

[0027] The neural network model is trained by using the normalized basis function matrix and the historical generation amount as a sample vector.

[0028] Furthermore, the wellbore management cloud platform stores data based on the different locations of the wellbore in different production areas, the temperature collected by temperature sensors at different times, the pressure collected by pressure sensors, the volume fraction of each gas source component and the flow rate of natural gas collected by gas sensors, and the wellbore depth collected by well depth meters to generate a big data database, so as to use the big data to improve the accuracy of the neural network model;

[0029] The wellbore management cloud platform carries an interface area data table, and the temperature and the pressure are substituted into the interface area data table to determine the corresponding interface area, so as to determine the rate of natural gas hydrate formation;

[0030] The wellbore management cloud platform determines the rate constant based on the change in the volume fraction.

[0031] In the above scheme, the neural network model is used to achieve better prediction accuracy and promotion ability of model generation.

[0032] Furthermore, the wellbore management cloud platform simulates and derives a plurality of simulated generation amounts of different set durations according to the generation rate;

[0033] The wellbore management cloud platform determines the model generation amounts of multiple different set time lengths according to the initial model generation amount obtained by the objective function;

[0034] The wellbore management cloud platform generates a well depth curve of the predicted generation quantity based on the average value or weighted average value of the multiple simulated generation quantities and the multiple model generation quantities, determines the predicted blockage rate after different time lengths based on the time well depth curve, and generates a time curve for display.

[0035] Furthermore, the wellbore management cloud platform determines and displays a warning freezing and blocking level based on the predicted blocking rate to display the freezing and blocking situation intuitively.

[0036] Furthermore, when the wellbore management cloud platform determines that the duration of the warning freezing and blocking level being at the dangerous level reaches a first duration, the wellbore drainer is controlled to increase operating power to reduce the moisture content of the natural gas in the wellbore;

[0037] After the wellbore management cloud platform controls to increase the operating power of the drainer, when it is determined that the warning freezing level has been at the dangerous level for a second time period, the pipeline doser is controlled to inject antifreeze agent into the wellbore.

[0038] In the above scheme, it is achieved by accurately predicting the blockage rate and controlling the wellbore's drainer and dosing device, while automatically avoiding the complete freezing of natural gas hydrates in the wellbore, thereby minimizing the impact of freezing on wellbore production.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] 1. The predicted gas hydrate production and blockage rate are determined by using a set of thermodynamic equations and a neural network model. This combines the neural network model's ability to predict unknown situations with the physical constraints of the thermodynamic equations on the gas hydrate formation process, making the predicted values ​​more accurate and reliable and enhancing the robustness of the prediction system.

[0041] 2. The Clapeyron equation can accurately describe the behavior of gas during phase change, while the VWP model can accurately simulate the formation process of natural gas hydrates in porous media. Combining the two can derive a more accurate natural gas hydrate formation rate, which is particularly suitable for the complex environment of wellbores.

[0042] 3. By accurately predicting the blockage rate and controlling the wellbore's drainer and dosing device, the system can automatically avoid the complete freezing of natural gas hydrates in the wellbore, thus minimizing the impact of freezing on wellbore production. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of the system structure of an embodiment of the present invention;

[0044] Figure 2 Schematic diagram of the process of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of collected data curve of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of prediction results of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0048] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0049] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0050] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0051] like Figures 1 to 4 As shown, the present invention provides a method for predicting and evaluating wellbore natural gas hydrate freezing and plugging, which determines the predicted natural gas hydrate generation amount and predicted blockage rate through a thermodynamic equation group and a neural network model. It combines the predictive ability of the neural network model for unknown situations with the physical constraints of the natural gas hydrate generation process of the thermodynamic equation group, making the predicted values ​​more accurate and reliable and enhancing the robustness of the prediction system.

[0052] in, Figure 1 Schematic diagram of the system structure of an embodiment of the present invention; Figure 2 Schematic diagram of the process of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention; Figure 3 Schematic diagram of collected data curve of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention; Figure 4 Schematic diagram of prediction results of the natural gas hydrate freezing prediction and evaluation method according to an embodiment of the present invention.

[0053] Example 1:

[0054] like Figures 1 to 4 As shown, this embodiment proposes a method for predicting and evaluating natural gas hydrate freezing and blockage in a wellbore, including: obtaining real-time environmental data in the wellbore collected by sensors, establishing a set of thermodynamic equations for representing natural gas hydrate formation conditions, substituting the real-time environmental data into the set of thermodynamic equations to obtain a natural gas hydrate formation rate, and performing simulation based on the formation rate to obtain a simulated formation amount of natural gas hydrate within a set time period; obtaining historical data and historical formation amount of natural gas hydrate in the wellbore collected by sensors, establishing a neural network model for determining the relationship between the historical data and the historical formation amount, training the neural network model using the historical data and historical formation amount as sample vector data, and then substituting the real-time environmental data into the neural network model to obtain a model formation amount of natural gas hydrate; and determining a predicted formation amount and a predicted blockage rate of natural gas hydrate at multiple locations in the wellbore based on the simulated formation amount and the predicted formation amount, so as to improve the accuracy of the prediction and evaluation.

[0055] Robustness, as it's understood, refers to the ability of its output to remain relatively stable and reliable despite a certain range of perturbations in the input data. Since neural network models are inherently limited by the richness of sample vectors, and the complex environment within a wellbore can create extreme situations where these vectors can't accurately capture them, the physical constraints of the natural gas hydrate formation process within the thermodynamic equations can be used to correct the neural network model's predictions, thereby making them more accurate and robust within the complex wellbore environment.

[0056] Furthermore, the thermodynamic equations include the Clapeyron equation and a thermodynamic model, and the real-time environmental data includes at least temperature and interface area; the temperature is substituted into the Clapeyron equation to determine the equilibrium pressure of natural gas hydrate formation; the temperature is substituted into the thermodynamic model to determine the driving force of natural gas hydrate formation; and the rate of natural gas hydrate formation is determined based on the equilibrium pressure, the driving force and the interface area.

[0057] Furthermore, the thermodynamic model is a VWP model; by calculating the occupancy rate, escape coefficient and adsorption constant of gas molecules in the natural gas hydrate cage and substituting them into the VWP model, the difference between the liquid phase chemical potential of water and the natural gas hydrate phase chemical potential is determined, and the difference is the driving force.

[0058] Furthermore, the real-time environmental data also includes pressure; a corresponding rate coefficient is determined according to the driving force; and the generation rate is obtained by multiplying the rate coefficient with the interface area and the difference between the pressure and the equilibrium pressure.

[0059] Specifically, the real-time environmental data includes the temperature, pressure, gas constant, wellbore depth, methane volume fraction, empirical constant of Langmuir adsorption constant, gas molar volume, natural gas hydrate molar volume, wellbore temperature gradient, wellbore pressure gradient, etc. of the natural gas hydrate formation environment.

[0060] Specifically, the process of determining the generation rate is:

[0061] Step S1: using the Clapeyron equation to calculate the equilibrium pressure of natural gas hydrate formation at a specific temperature. The Clapeyron equation is:

[0062]

[0063] Where P is the equilibrium pressure, T is the temperature, L is the latent heat, Vg is the gas phase volume, and Vs is the solid phase volume. The temperature T, latent heat L, gas phase volume Vg, and solid phase volume Vs are acquired by sensors.

[0064] Step S2: Calculate the difference between the chemical potential of water in the natural gas hydrate phase and the chemical potential of water in the liquid phase using the VWP (VanderWaal's-Platteeuw) model. The VWP model is:

[0065]

[0066] Where, is the chemical potential of water in the gas hydrate phase, is the chemical potential of water in the liquid phase, v i is the occupancy rate of gas molecules in the natural gas hydrate cage, C ij is the Langmuir adsorption constant, f j is the escape coefficient of gas j in cage i.

[0067] Step S3: Calculate the natural gas hydrate formation rate using a rate equation. By dynamically simulating and calculating the formation rate, predict the amount and location of natural gas hydrate that may be formed in the wellbore within a given time. The rate equation is:

[0068]

[0069] Where, is the gas hydrate formation rate, k is the rate coefficient, A is the interfacial area, P eq is the equilibrium pressure for natural gas hydrate formation.

[0070] The above step S2 includes:

[0071] Step S21: Calculate the Langmuir adsorption constant of each gas molecule according to an empirical formula, wherein the empirical formula is:

[0072]

[0073] Where K ij is a constant, R is the gas constant, and T is the absolute temperature.

[0074] Step S22: Calculate the occupancy rate of gas molecules in different types of cages, specifically:

[0075]

[0076] Where θ i is the cage occupancy rate, C i and C j is the summed Langmuir adsorption constant, f i and f jis the escape coefficient of gas molecules. The escape rate of gas is indirectly assessed by continuously monitoring changes in gas concentration using an electrochemical gas sensor or a semiconductor gas sensor. The escape coefficient is determined based on the linear relationship between the escape rate and the escape coefficient in wellbore experiments.

[0077] Step S23: Substitute the adsorption constant and the occupancy rate into the VWP model to calculate the chemical potential difference. The chemical potential difference of water is used to represent the driving force of generation. The chemical potential difference is specifically:

[0078]

[0079] Where Δμ is the chemical potential difference, is the chemical potential of water in the liquid phase, is the chemical potential of water in the gas hydrate phase.

[0080] In the above scheme, the Clapeyron equation can accurately describe the behavior of gas during phase change, while the VWP model can accurately simulate the formation process of natural gas hydrates in porous media. Combining the two can derive a more accurate natural gas hydrate generation rate, which is particularly suitable for complex geological environments in wellbores.

[0081] Furthermore, the neural network model is a feedforward neural network model, and the historical environmental data include the temperature, pressure and depth of the wellbore, the volume fraction of each gas source component of the natural gas, the interface area, the rate constant and the natural gas flow rate; a polynomial basis function matrix is ​​constructed according to each of the historical environmental data, and the quadratic polynomial matrix includes the square of each of the historical environmental data and the product of two different historical environmental data; the basis function matrix and the historical generation amount are used to construct the sample vector data.

[0082] Furthermore, the neural network model is a multi-layer feedforward neural network model based on a back propagation error algorithm. The historical data includes wellbore environmental parameters, natural gas composition parameters, and database parameters. The input layer neurons of the neural network model are environmental data, and the output layer nodes are natural gas hydrate production. A loss function with a regularization term is used to improve the generalization ability of the neural network model. The neural network model is trained using normalized historical data and the historical production volume as sample vectors to generate an objective function. The real-time environmental data is substituted into the objective function to obtain the model production volume. The environmental data includes historical environmental data and real-time environmental data.

[0083] Furthermore, the historical data includes the volume fraction, temperature, pressure, wellbore depth, interface area, rate constant and flow rate of each gas source component of natural gas; a basis function matrix is ​​generated based on the historical data, and the basis function matrix includes each historical data, the square of each historical data and the product of two different historical data; the normalized basis function matrix and the historical generation amount are used as sample vectors to train the neural network model.

[0084] Furthermore, the wellbore management cloud platform stores and manages the temperatures collected by temperature sensors at different locations and at different times in the wellbore of different production areas, the pressures collected by pressure sensors, the volume fractions of various gas source components of natural gas collected by gas sensors and the flow rate of natural gas, and the wellbore depth collected by the well depth meter, and generates a big data database that stores at least 10,000 sets of data to improve the accuracy of the neural network model; the wellbore management cloud platform is equipped with an interface area data table, and the temperature and the pressure are substituted into the interface area data table to determine the corresponding interface area to determine the rate of natural gas hydrate formation; the wellbore management cloud platform determines the rate constant based on the change in the volume fraction.

[0085] Furthermore, the wellbore management cloud platform simulates the generation rate to obtain multiple simulated generation quantities with different set time lengths; the wellbore management cloud platform determines multiple model generation quantities with different set time lengths based on the initial model generation quantity obtained by the objective function; the wellbore management cloud platform generates a well depth curve of the predicted generation quantity based on the average or weighted average of the multiple simulated generation quantities and the multiple model generation quantities, determines the predicted blockage rate after different time lengths based on the time well depth curve and generates a time curve for displaying the predicted freezing and blockage situation.

[0086] In the above scheme, the neural network model is used to achieve better prediction accuracy and promotion ability of model generation.

[0087] Furthermore, the transfer function of the hidden layer neurons of the neural network model is a hyperbolic tangent S-type, and the transfer function of the output layer neurons is a linear transfer function.

[0088] In the above scheme, the neural network model is used to achieve better prediction accuracy and promotion ability of model generation.

[0089] Specifically, the process of determining the model generation amount is:

[0090] Step S4, obtaining parameters of the neural network model for wellbore natural gas hydrate formation, wherein environmental parameters include temperature (T), pressure (P), and wellbore depth (D); chemical parameters include natural gas composition (volume fraction of CH, C2H6, etc.); operating parameters include interfacial area (A), rate constant (k), and flow velocity (v); and environmental parameters include temperature gradient and pressure gradient.

[0091] Step S5: Use the above parameters as input variables to construct the polynomial basis function. Assuming we choose a quadratic polynomial basis function, the basis function can be defined as: φ0 = 1, φ1 = T, φ2 = P, φ3 = D, φ4 = xCH4, φ5 = xC2H6, φ6 = A, φ7 = k, φ8 = v, φ9 = T2, φ10 = P2, φ11 = D2, φ12 = x2CH4, φ13 = x2C2H6, φ14 = A2, φ15 = k2, φ16 = v2, φ17 = TP, φ18 = TD, φ19 = TxCH4, φ20 = TxC2H6, φ21 = TA, φ22 = Tk, φ23 = Tv, φ24 = PD, φ25 = PxCH4, φ26 = PxC2H6, φ27 = PA, φ28 = Pk, φ29 = Pv...

[0092] Step S6, construct a basis function matrix with the polynomial basis function, specifically: for each sample i, the i-th behavior of the design matrix is: Φ(xi)=[φ0(xi),φ1(xi),φ2(xi),…,φn(xi)], for example, assuming there is a sample data (T i,Pi,Di,xCH4,i,xC2H6,i,Ai,ki,vi)(Ti,Pi,D i,xCH4,i,xC2H6,i,Ai,ki,vi), the corresponding design matrix behavior is: Φ(xi)=[1,Ti,Pi,Di,xCH4,i,xC2H6,i,Ai,ki,vi,T i2,Pi2,Di2,xCH4,i2,xC2H6,i2,Ai2,ki2,vi2,TiPi,T iDi,…]Φ(xi)=[1,Ti,Pi,D i,xCH4,i,xC2H6,i,Ai,ki,vi i,xCH4,i,xC2H6,i,Ai,ki,vi,Ti2,Pi2,Di2,xCH4,i2,xC2H6,i2,Ai2,ki2,vi2,Ti].

[0093] Step S7: Use the constructed basis function matrix Φ and target variable y (such as natural gas hydrate formation rate, historical natural gas hydrate formation amount) to train the neural network model.

[0094] Step S8: After the neural network model is trained, the model can be used for prediction to obtain the model generation amount. new , first convert it into the design matrix form Φ(xnew ), and then calculate the predicted value: y = Φ(x new )w, where y is the predicted value and w is the weight obtained by the neural network.

[0095] Preferably, the collected natural gas hydrate formation environment data includes: temperature (T), the temperature at different locations in the wellbore, in Kelvin (K); pressure (P), the pressure at different locations in the wellbore, in megapascals (MPa); gas constant (R), 8.314 J / (mol·K); wellbore depth (D), in meters (m). More preferably, the imported collected data of different quantities are actual test pressure, temperature, gas source composition, and other related data under different operating conditions in different production areas.

[0096] Specifically, the wellbore management cloud platform Figure 1 The sensor shown acquires environmental data, real-time environmental data, and historical environmental data. Specifically, the synchronous X-ray imaging method of the X-ray imager is used to perform three-dimensional imaging of the natural gas hydrate, and the image is technically processed to obtain the numerical value of the interface area of ​​the natural gas hydrate.

[0097] Step S9: Determine the predicted generation amount and predicted blockage rate of natural gas hydrate at multiple locations and multiple times in the wellbore based on the average or weighted average of the simulated generation amount and the predicted generation amount, and draw the following diagram: Figure 3 and 4 The curve shown.

[0098] Preferably, the polynomial network basis function constructed based on the collected data is converted into input data volume and generated results by importing pressure data, temperature data, and gas source data into a computer to establish a generation environment model.

[0099] More specifically, the optimal generation conditions, generation quantity, and time model are compared with the optimal data volume in the comparison of the polynomial network basis function results, and more accurate wellbore natural gas hydrate generation data volume and generation time data are obtained.

[0100] More specifically, step S5 and step S6 include: collecting wellbore natural gas hydrate formation environmental data, the next step of collecting wellbore natural gas hydrate formation environmental data is to establish a wellbore formation prediction model, the next step of the computer establishing the wellbore formation prediction model is to construct a polynomial network basis function matrix based on the collected data, the next step of constructing the polynomial network basis function matrix based on the collected data is to import different amounts of collected data, the collected data is connected to the database through JDBC, the database related parameters are configured, and the index table space is created to define and collect initial parameters and data.

[0101] For example, the collected data in the following table:

[0102] T(K) P (MPa) D(m) <![CDATA[x_CH4]]> <![CDATA[x_C2H6]]> <![CDATA[A(m 2 )]]> k v(m / s) 275 10 1000 0.9 0.05 0.1 0.01 0.5 280 12 1200 0.85 0.1 0.12 0.02 0.6

[0103] For each sample i, construct the i-th row of the design matrix:

[0104] Sample 1:

[0105] Φ(x1)=[1,275,10,1000,0.9,0.05,0.1,0.01,0.5,275 2 ,10 2 ,1000 2 ,0.92,0.052,0.1 2 ,0.01 2 ,0.5 2 ,275·10,275·1000,275·0.9,…]

[0106] Sample 2:

[0107] Φ(x2)=[1,280,12,1200,0.85,0.1,0.12,0.02,0.6,280 2 ,12 2 ,1200 2 ,0.85 2 ,0.12,0.12 2 ,0.02 2 ,0.6 2 ,280·12,280·1200,280·0.85,…]

[0108] More specifically, step S8 includes: for the gas wells that need to be predicted, first collect pressure data, temperature data, and gas source data, and store the collected data in a temporary table of the database through a JDBC connection. The data cleaning program processes the format, anomalies, redundancy, etc. of the collected data and then organizes them. The relevant data is classified according to statistical conditions in the database to generate a table and save it. The collected data is imported into the optimized prediction model, that is, different amounts of collected data are imported.

[0109] More specifically, step S8 includes: analyzing and comparing the prediction results of the neural network model, that is, comparing the results of the polynomial network basis functions; by comparing the calculation results, determining the optimal pressure data, temperature data, gas source data, the amount of natural gas hydrate generated data and the generation time data, that is, the most accurate generation conditions and generation quantity and time model.

[0110] Because comprehensive assessment of the factors influencing natural gas hydrates in wellbores is quite challenging, the seven historical environmental factors mentioned above, which are easily and accurately measured, were selected as the inputs to a feedforward neural network model. This model is divided into 35 input layer neuron nodes, ranging from φ0(xi) to φ35(xi), where φ(xi) corresponds to data at multiple moments for the neuron, eliminating cases where the input neuron is zero. The output layer has one node, representing the model generation amount or model generation rate for natural gas hydrates at the corresponding composition and pressure. The choice of hidden layer depends on the system itself. For more complex systems, selecting one hidden layer results in a three-layer feedforward neural network (BP network). The number of hidden layer nodes is two to three times the number of input nodes, with 70 being the preferred value.

[0111] Common transfer functions used in neural networks include linear transfer function, hyperbolic tangent sigmoid transfer function, logarithmic sigmoid transfer function, etc. Preferably, the hyperbolic tangent sigmoid (tans ig) is used as the transfer function of the hidden layer neurons.

[0112] The learning method of feedforward neural networks, specifically the weight and threshold adjustment rules, utilize the backpropagation algorithm. By adjusting the thresholds and weights along the negative gradient of the network error, the network error is ultimately minimized, where the error gradient reaches zero. Due to inherent flaws in gradient descent algorithms, standard feedforward neural algorithms typically suffer from slow convergence and a tendency to get stuck in local minima. Therefore, regularization methods are used to improve the network's generalization capabilities.

[0113] In order to avoid the influence of the distance between the maximum and minimum values ​​in the network input vector on the network error propagation, the input data of each neuron in the input layer are uniformly normalized. The maximum number of training times set for the network is 20,000 times, and the network performance target is set to 1×10 —6 Since the model uses a fast and effective learning algorithm, after 936 training cycles, the network performance target is 8.82844×10 —7 The model achieved the required accuracy. The maximum absolute error between the historical and modeled generation amounts was calculated to be 0.41, and the minimum absolute error was 0.06. This demonstrates that the neural network model built on the selected samples accurately predicts the formation conditions of natural gas hydrates in this wellbore. It also demonstrates that the selection of the 35 input neurons required by the neural network model is appropriate, resulting in satisfactory prediction results.

[0114] Furthermore, the wellbore management cloud platform determines and displays a warning freezing blockage level based on the predicted blockage rate, providing a visual display of the freezing blockage situation. It is understood that by visually observing the freezing blockage situation of the wellbore, construction personnel can conveniently evaluate the safety of the wellbore and determine whether the wellbore construction process needs to be improved. This allows them to rationally determine the diameter of the freezing hole layout, correctly select pipe materials and connection methods, control the excavation section height and the exposure time of the wellbore wall, and thus reduce the occurrence of problems such as frozen pipe breakage and wellbore wall cracking, thereby improving construction quality, reducing construction risks, and laying the foundation for long-term and stable production of gas wells.

[0115] Furthermore, when the wellbore management cloud platform determines that the warning freezing and blocking level has been at a dangerous level for a first period of time, it controls the wellbore drainer to increase its operating power to reduce the moisture content of the natural gas in the wellbore; after the wellbore management cloud platform controls the increase of the operating power of the drainer, when it determines that the warning freezing and blocking level has been at a dangerous level for a second period of time, it controls the pipeline doser to inject antifreeze agent into the wellbore.

[0116] Specifically, see Figure 4 When the freezing rate is between 0.05 and 0.15, the warning freezing level is mild; when the freezing rate is between 0.15 and 0.25, the warning freezing level is moderate; when the freezing rate is between 0.25 and 0.4, the warning freezing level is severe, where the severe warning freezing level is the dangerous level. The first time duration is preferably 30 minutes, and the second time duration is preferably 20 minutes. That is, the freezing rate is between 0.25 and 0.4 for more than 30 minutes. Considering that the introduction of antifreeze agents will have a slight impact on the quality of natural gas extraction, at this time the wellbore management cloud platform only controls the wellbore drainer to increase the operating power and extract the moisture from the natural gas in the wellbore. If the complex geological environment of the wellbore generates natural gas hydrates at a rate greater than the rate at which the drainer extracts water, the drainer alone may not be able to effectively prevent the formation of natural gas hydrates. As a result, after increasing the operating power of the drainer, the freezing rate is between 0.25 and 0.4 for more than 20 minutes. In this case, the wellbore management cloud platform controls the drainer to maintain high-power operation while controlling the pipeline doser to introduce antifreeze agents into the wellbore to completely prevent the formation of natural gas hydrates.

[0117] It is understandable that increasing the operating power of the drainage machine can increase the drainage volume, thereby improving the liquid carrying capacity of the wellbore, so that the produced water of the gas in the formation can be carried out of the wellbore more quickly, reducing the formation of liquid accumulation in the wellbore, and then reducing the pressure and temperature gradient of the wellbore, thereby reducing the amount of natural gas hydrate formed.

[0118] Specifically, organic chemical inhibitors such as methanol, ethylene glycol, and diethylene glycol are used. These inhibitors form solutions with very low freezing points with the water condensed during the cooling process. The high-concentration inhibitor solution can absorb water, resulting in a significant drop in the natural gas hydrate formation temperature, thereby preventing the formation of natural gas hydrates.

[0119] In the above scheme, it is achieved by accurately predicting the blockage rate and controlling the wellbore's drainer and dosing device, while automatically avoiding the complete freezing of natural gas hydrates in the wellbore, thereby minimizing the impact of freezing on wellbore production.

[0120] See also Figure 3 and Figure 4 The prediction results of this embodiment can be used to intuitively study the factors affecting the formation of natural gas hydrates through the curve, and timely intervention can be carried out to ensure that the drainage effect and antifreeze blocking agent have sufficient time to reach the bottom of the wellbore where the blockage is most serious.

[0121] Furthermore, in this embodiment, the data acquisition process uses a two-dimensional Gaussian function and an inverse tangent function to determine a duty cycle perturbation step function, and the perturbation step function is:

[0122]

[0123] Where G(x,y) is the function value at point (x,y), A is the amplitude, i.e. the maximum value, μ x and μ y are the means in the x and y directions respectively; σ x and σ y are the standard deviations in the x and y directions, respectively. arctan²(y,x), with a result range of -π to π, calculates the angle of a point (x,y) relative to the origin. The Gaussian function was chosen because it affects the amplitude of the perturbation step size, which can be used to prevent excessive duty cycle changes when the external environment undergoes drastic changes. The inverse tangent function was chosen because it can increase or decrease the perturbation step size, thereby improving the MPPT's dynamic response speed. It can also limit oscillations near the MPP, thereby improving the system's steady-state performance.

[0124] In this embodiment, the predicted gas hydrate production and blockage rate are determined using a set of thermodynamic equations and a neural network model. This combines the neural network model's predictive capabilities for unknown situations with the physical constraints of the thermodynamic equations for the gas hydrate formation process, making the predicted values ​​more accurate and reliable and enhancing the robustness of the prediction system. The Clapeyron equation accurately describes the behavior of gas during phase transitions, while the VWP model accurately simulates the formation of gas hydrates in porous media. Combining the two yields a more accurate gas hydrate generation rate, which is particularly applicable to complex wellbore geological environments. By accurately predicting the blockage rate, the wellbore's drainer and dosing device are controlled, minimizing the impact of freezing on wellbore production while automatically preventing complete blockage of the wellbore by gas hydrates.

[0125] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0126] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for predicting and evaluating wellbore natural gas hydrate freezing and plugging, characterized in that: include: Acquiring real-time environmental data from the wellbore collected by sensors, establishing a set of thermodynamic equations for representing natural gas hydrate formation conditions, substituting the real-time environmental data into the set of thermodynamic equations to obtain a natural gas hydrate formation rate, and performing a simulation based on the formation rate to obtain a simulated amount of natural gas hydrate formed within a set time period; Acquiring historical data and historical generation amounts of natural gas hydrates in the wellbore collected by sensors, establishing a neural network model for determining a relationship between the historical data and the historical generation amounts, training the neural network model using the historical data and the historical generation amounts as sample vector data, and then substituting the real-time environmental data into the neural network model to obtain a model generation amount of natural gas hydrates; determining predicted generation amounts and predicted blockage rates of natural gas hydrates at multiple locations in the wellbore based on the simulated generation amounts and the model generation amounts; The thermodynamic equations include the Clapeyron equation and a thermodynamic model, the real-time environmental data include at least the temperature and interface area of ​​the natural gas hydrate formation environment, and the historical data include wellbore environmental parameters, natural gas composition parameters, and database parameters. Substituting the temperature into the Clapeyron equation to determine the equilibrium pressure for natural gas hydrate formation; Substituting the temperature into the thermodynamic model to determine the driving force for natural gas hydrate formation; determining a rate of natural gas hydrate formation based on the equilibrium pressure, the driving force, and the interfacial area; The thermodynamic model is a VWP model; The occupancy rate, escape coefficient and adsorption constant of gas molecules in the natural gas hydrate cage are calculated and substituted into the VWP model to determine the difference between the liquid phase chemical potential of water and the natural gas hydrate phase chemical potential. The difference is the driving force.

2. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 1, characterized in that: The real-time environmental data also includes pressure; determining a corresponding rate coefficient according to the driving force; The generation rate is obtained by multiplying the rate coefficient, the interfacial area, and the difference between the pressure and the equilibrium pressure.

3. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 1 or 2, characterized in that: The neural network model is a multi-layer feedforward neural network model based on the error back propagation algorithm; The input layer neurons of the neural network model are environmental data, and the output layer nodes are the production amount of natural gas hydrates, which adopts a loss function with a regularization term; Training the neural network model using normalized historical data and the historical generation amount as sample vectors to generate an objective function; Substituting the real-time environmental data into the objective function obtains the model generation amount.

4. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 3, characterized in that: The historical data includes the volume fraction of each gas source component of the natural gas, temperature, pressure, wellbore depth, interface area, rate constant and flow rate of natural gas hydrate formation; generating a basis function matrix according to the historical data, wherein the basis function matrix includes each of the historical data, the square of each of the historical data, and the product of two different historical data; The neural network model is trained by using the normalized basis function matrix and the historical generation amount as a sample vector.

5. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 4, characterized in that: The wellbore management cloud platform stores data based on the different locations of wellbores in different production areas, the temperatures collected by temperature sensors at different times, the pressures collected by pressure sensors, the volume fractions of various gas source components and the flow rates of natural gas collected by gas sensors, and the wellbore depths collected by well depth meters to generate a big data database, so as to improve the accuracy of the neural network model using big data. The wellbore management cloud platform carries an interface area data table, and the temperature and the pressure are substituted into the interface area data table to determine the corresponding interface area, so as to determine the rate of natural gas hydrate formation; The wellbore management cloud platform determines the rate constant based on the change in the volume fraction.

6. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 5, characterized in that: The wellbore management cloud platform simulates and derives a plurality of simulated generation amounts with different set time lengths according to the generation rate; The wellbore management cloud platform determines the model generation amounts of multiple different set time lengths according to the initial model generation amount obtained by the objective function; The wellbore management cloud platform generates a well depth curve of the predicted generation quantity based on the average value or weighted average value of the multiple simulated generation quantities and the multiple model generation quantities, determines the predicted blockage rate after different time lengths based on the well depth curve, and generates a time curve for display.

7. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 6, characterized in that: The wellbore management cloud platform determines and displays the early warning freezing and blocking level according to the predicted blocking rate.

8. The method for predicting and evaluating wellbore natural gas hydrate freezing and plugging according to claim 7, characterized in that: When the wellbore management cloud platform determines that the warning freezing and blocking level has been at a dangerous level for a first duration, the wellbore drainer is controlled to increase operating power to reduce the moisture content of the natural gas in the wellbore; After the wellbore management cloud platform controls to increase the operating power of the drainer, when it is determined that the warning freezing level has been at the dangerous level for a second time period, the pipeline doser is controlled to inject antifreeze agent into the wellbore.

Images