System and method for improving yield of horizontal well of low-permeability oil field

Abstract

The invention relates to the technical field of petroleum engineering, in particular to a system and method for improving the yield of a horizontal well of a low-permeability oilfield, and the system comprises a permeability distribution characteristic data acquisition unit, an oil well test data acquisition unit, a numerical simulation unit, a calculation unit, an optimization unit and a control unit. Permeability data, pressure measurement data and production dynamic data are collected, numerical modeling is conducted in a non-equilibrium state seepage simulation mode, and pressure distribution in the horizontal shaft streamline direction and in the horizontal shaft streamline direction is calculated. And the optimization unit performs coupling calculation according to the flow characteristic data and the single well characteristic data of the horizontal well to obtain an optimization scheme. And the control unit feeds the optimization scheme back to the underground data acquisition unit so as to adjust liquid injection measures in the horizontal shaft, optimize flow characteristics and improve the yield of the horizontal well. The method breaks through the limitation of a traditional equilibrium state model, the complex flow characteristics of the fluid in the low-permeability oil field can be described more accurately, and the simulation precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of petroleum engineering, and in particular to a system and method for improving the production of horizontal wells in low-permeability oil fields. Background Art

[0002] Low-permeability oilfields, defined as those with permeabilities below 50 mD, present poor reservoir properties, high fluid flow resistance, and high development challenges. Horizontal well technology, a primary method for developing low-permeability oilfields, effectively increases reservoir contact area and improves single-well productivity. However, due to the high heterogeneity of low-permeability reservoirs and the significant differences in physical properties between reservoir sections within horizontal wells, horizontal well production often results in uneven flow within the wellbore, localized high pressure drops, and "huff-and-puff" effects within the wellbore, severely impacting well productivity.

[0003] Existing technologies primarily rely on empirical experience or relatively simple mathematical models to optimize horizontal well productivity, which struggle to accurately describe the complex fluid flow characteristics of low-permeability oilfields. Traditional methods typically employ uniform injection strategies, ignoring reservoir heterogeneity and resulting in inefficient resource utilization. Furthermore, existing technologies generally lack mechanisms for refined management and real-time optimization of each reservoir segment in a horizontal well, making it impossible to adjust production strategies in response to dynamic changes during the production process. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for improving the production of horizontal wells in low permeability oil fields, aiming to solve technical problems in the prior art such as inaccurate optimization of production capacity of horizontal wells in low permeability oil fields and unreasonable injection strategies.

[0005] The present invention proposes a system for improving the production of horizontal wells in low permeability oil fields, comprising:

[0006] a permeability distribution characteristic data acquisition unit, for acquiring permeability data of each oil layer section of a horizontal well in the low permeability oil field;

[0007] An oil well test data acquisition unit, communicatively connected to the permeability distribution characteristic data acquisition unit, for acquiring pressure measurement data and production dynamic data of the horizontal well;

[0008] a numerical simulation unit, communicatively connected to the permeability distribution characteristic data acquisition unit and the oil well test data acquisition unit, for performing numerical modeling based on the permeability distribution characteristic data and a non-equilibrium seepage simulation mode, and establishing a simulation model based on fitting and correction of the pressure measurement data and the production performance data;

[0009] a calculation unit, in communication with the numerical simulation unit, for calculating the direction of horizontal wellbore streamlines and the pressure distribution thereon according to the simulation model to form flow characteristic data;

[0010] an optimization unit, communicatively connected to the calculation unit and the oil well test data acquisition unit, and configured to couple the flow characteristic data with the horizontal well characteristic data to obtain an optimization solution based on the horizontal well characteristic data of the low permeability oil field, wherein the horizontal well characteristic data includes production dynamic pressure data and water absorption profile data of the horizontal well in the block where the horizontal well is located;

[0011] A control unit is communicatively connected to the optimization unit and is used to control the optimization unit to feed back the optimization scheme to a downhole data acquisition unit in the horizontal well oil well. The downhole data acquisition unit adjusts the injection measures in the horizontal wellbore according to the optimization scheme, optimizes the flow characteristics in the horizontal wellbore, and thus improves the production of the horizontal well.

[0012] 2. The system according to claim 1, wherein the optimization unit comprises:

[0013] A test model building subunit is used to receive the water absorption profile data in the horizontal well single well characteristic data, apply the wavelet transform algorithm to perform multi-scale decomposition on the water absorption profile data, identify the abnormal water absorption area of ​​the oil layer through adaptive threshold division, establish a multi-dimensional profile model considering geological heterogeneity, divide the injection and production mode of the single well oil layer, and establish a test model;

[0014] a numerical simulation algorithm optimization subunit, which is in communication with the test model building subunit and is used to receive the test model and the flow characteristic data, perform local optimization on the abnormal area using an adaptive grid refinement technique, introduce a dynamic boundary condition self-adjustment mechanism, adjust model parameters using a gradient descent optimization algorithm, and optimize the numerical simulation algorithm to obtain an optimized model;

[0015] an oil layer parameter division subunit, communicatively connected to the numerical simulation algorithm optimization subunit, for receiving the optimization model and the production dynamic pressure data, dividing the water absorption profile data of the oil layer into injection unit parameters using a spatiotemporal response prediction model based on an LSTM neural network, and calculating injection water flow distribution characteristic data for each unit layer, the injection water flow distribution characteristic data including flow characteristics and / or distribution characteristics of the injected water;

[0016] The data coupling subunit is respectively communicated with the oil layer parameter division subunit and the calculation unit, and is used to establish a nonlinear mapping relationship between the injection water flow characteristics and the water saturation, receive the flow characteristic data, couple the injection water flow distribution characteristic data of each unit layer with its injection water saturation, and use the Bayesian optimization algorithm to search the parameter space to form an optimization scheme for the optimization unit.

[0017] Preferably, the optimization unit includes:

[0018] An oil layer parameter division subunit is used to receive the flow characteristic data, divide the oil layer of the single well into injection unit parameters, and calculate the distribution characteristic data of the injected water in each unit layer;

[0019] The data coupling subunit is in communication connection with the oil layer parameter division subunit and is used to couple the distribution characteristic data of the injected water of each unit layer with the injection and production unit parameters of each unit layer to form an optimization scheme for the optimization unit.

[0020] Preferably, the optimization unit includes:

[0021] The solution output subunit is communicatively connected to the data coupling subunit, and is used to receive the optimization solution, adopt a real-time adjustment algorithm of injection volume based on fuzzy control, introduce a multi-objective optimization decision-making mechanism of production-injection volume, realize equipment health status monitoring, and adjust the injection measures of each unit layer in the single well oil layer according to the optimization solution.

[0022] Preferably, the system further comprises:

[0023] The verification unit is respectively connected to the control unit and the optimization unit for testing the effect of the optimization scheme on the downhole data acquisition unit of another well in the low permeability oil field according to the optimization scheme, and feeding back the test results to the control unit.

[0024] Preferably, the horizontal wellbore streamline optimization method is verified, and the area with the smallest change in streamline diversion ratio in the horizontal wellbore streamline optimization method is selected to carry out downhole dynamic monitoring and obtain dynamic monitoring results; through downhole dynamic monitoring, it is judged whether the dynamic monitoring results have achieved the set effect; when it is determined that the dynamic monitoring results have achieved the set effect, it is proved that the optimization scheme is effective; when it is determined that the dynamic monitoring results have not achieved the set effect, it is proved that the optimization scheme is not effective.

[0025] Preferably, the pressure measurement data includes: pressure gradient data, pressure recovery data, and pressure dynamic data; the production dynamic data includes: production dynamic pressure data and production dynamic water absorption profile data.

[0026] Preferably, the permeability distribution characteristic data acquisition unit includes at least one well logging device and / or coring device; the downhole data acquisition unit includes at least a water absorption profile tester, a pressure gauge, a pressure sensor and a flow meter.

[0027] Preferably, the system further comprises:

[0028] The logging unit is respectively communicated with the permeability distribution characteristic data acquisition unit and the optimization unit, and is used to design a logging plan based on the actual oil layer conditions of the block where the horizontal well is located, and to collect downhole dynamic test data according to the logging plan; the downhole dynamic test data is combined with the optimization plan to form a final optimization plan.

[0029] A method for increasing the production of horizontal wells in low-permeability oil fields, the method being applied to the system, the method comprising:

[0030] Step S1, using a permeability distribution characteristic data acquisition unit to acquire permeability data of each oil layer of a horizontal well in the low permeability oil field;

[0031] Step S2: using an oil well test data acquisition unit to acquire pressure measurement data and production performance data of the horizontal well;

[0032] Step S3, using a numerical simulation unit to perform numerical modeling based on the permeability distribution characteristic data and a non-equilibrium seepage simulation mode, and fitting and correcting based on the pressure measurement data and the production dynamic data to establish a simulation model;

[0033] Step S4: using a calculation unit to calculate the direction of the horizontal wellbore streamline and the pressure distribution thereon according to the simulation model to form flow characteristic data;

[0034] Step S5: using an optimization unit to couple the flow characteristic data with the horizontal well characteristic data of the low permeability oil field to obtain an optimization solution;

[0035] Step S6: Using the control unit to control the optimization unit to feed back the optimization plan to the downhole data acquisition unit in the horizontal well, and the downhole data acquisition unit adjusts the injection measures in the horizontal wellbore according to the optimization plan to optimize the flow characteristics in the horizontal wellbore, thereby increasing the production of the horizontal well;

[0036] Wherein, in step S5:

[0037] A test model building subunit is used to receive water absorption profile data in the horizontal well single well characteristic data, a wavelet transform algorithm is applied to perform multi-scale decomposition on the water absorption profile data, an abnormal water absorption area of ​​the oil layer is identified through adaptive threshold division, a multi-dimensional profile model considering geological heterogeneity is established, the injection and production mode of the single well oil layer is divided, and a test model is established;

[0038] A numerical simulation algorithm optimization subunit is used to receive the test model and the flow characteristic data, an adaptive grid refinement technology is used to perform local optimization on the abnormal area, a dynamic boundary condition self-adjustment mechanism is introduced, model parameters are adjusted by a gradient descent optimization algorithm, and the numerical simulation algorithm is optimized to obtain an optimized model;

[0039] An oil layer parameter division subunit is used to receive the optimization model and the production dynamic pressure data, and a spatiotemporal response prediction model based on an LSTM neural network is used to divide the water absorption profile data of the oil layer into injection unit parameters, and the injected water flow distribution characteristic data of each unit layer is calculated;

[0040] A data coupling subunit is used to establish a nonlinear mapping relationship between the flow characteristics of the injected water and the water saturation, the flow characteristic data is received, the injected water flow distribution characteristic data of each unit layer is coupled with its injected water saturation, and a Bayesian optimization algorithm is used to search the parameter space to form an optimization solution for the optimization unit;

[0041] A scheme output subunit is used to receive the optimization scheme, a real-time adjustment algorithm for injection volume based on fuzzy control is adopted, a multi-objective optimization decision-making mechanism for production volume and injection volume is introduced to realize equipment health status monitoring, and injection measures for each unit layer in the single well oil layer are adjusted according to the optimization scheme.

[0042] The beneficial effects of the present invention are:

[0043] 1. Through non-equilibrium seepage simulation modeling, the limitations of traditional equilibrium models have been broken through, and the complex flow characteristics of fluids in low permeability oil fields can be more accurately described, thereby improving simulation accuracy;

[0044] 2. Adaptive multi-scale oil layer water absorption profile identification and zoning technology is used to achieve refined management of oil layers and formulate differentiated injection strategies for different permeability areas;

[0045] 3. The automatic reservoir parameter segmentation system based on deep reinforcement learning breaks through the limitations of traditional linear segmentation methods and realizes intelligent nonlinear segmentation of reservoir parameters;

[0046] 4. The dynamic closed-loop optimization mechanism with multi-parameter coupling can timely adjust the optimization strategy according to changes in the production process to achieve continuous optimization;

[0047] 5. Intelligent adaptive injection control technology reduces operational complexity and professional skill requirements, making it easier for front-line engineers to implement optimization plans.

[0048] Based on the above technical effects, the present invention can significantly increase the production of horizontal wells in low-permeability oil fields, extend the effective production cycle, improve resource utilization, and reduce development costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a schematic diagram of the overall structure of the system for improving the production of horizontal wells in low permeability oil fields according to the present invention;

[0050] Figure 2 Schematic diagram of the internal structure of the optimization unit of the present invention;

[0051] Figure 3 A workflow diagram for building subunits for the test model of the present invention;

[0052] Figure 4 This is a workflow diagram of the numerical simulation algorithm optimization subunit of the present invention;

[0053] Figure 5 The flowchart of the oil reservoir parameter division subunit of the present invention;

[0054] Figure 6 This is a workflow diagram of the data coupling subunit of the present invention;

[0055] Figure 7 Output the workflow diagram of the subunits of the solution of the present invention;

[0056] Figure 8 This is a workflow diagram of the verification unit of the present invention;

[0057] Figure 9 The present invention is a flow chart of a method for increasing the production of horizontal wells in low permeability oil fields. DETAILED DESCRIPTION

[0058] Please refer to the attached Figure 1-9 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] like Figure 1 As shown, the present invention provides a system for improving the production of horizontal wells in low-permeability oil fields, comprising a permeability distribution characteristic data acquisition unit 1, an oil well test data acquisition unit 2, a numerical simulation unit 3, a calculation unit 4, an optimization unit 5, a control unit 6, a verification unit 7, and a logging unit 8. Each unit is connected via a communication interface, forming a complete data acquisition, analysis, optimization, and control closed loop.

[0060] The permeability distribution characteristic data acquisition unit 1 is used to collect permeability data for various oil layer sections in horizontal wells in low-permeability oil fields. Preferably, this unit includes multiple logging devices and / or coring equipment, and can utilize various logging methods, such as acoustic logging, resistivity logging, and nuclear magnetic resonance logging, to obtain the permeability distribution characteristics of the oil layer. In one embodiment of the present invention, the logging equipment can collect permeability data every meter to ensure spatial resolution, while the coring equipment can sample key locations for indoor analysis to obtain more accurate permeability data.

[0061] The oil well test data acquisition unit 2 is communicatively connected to the permeability distribution characteristic data acquisition unit 1 and is used to collect pressure measurement data and production performance data from horizontal wells. Pressure measurement data includes pressure gradient data, pressure buildup data, and pressure performance data; production performance data includes production dynamic pressure data and production dynamic water absorption profile data. In a preferred embodiment of the present invention, this unit can be configured with multiple distributed pressure sensors to collect pressure data every four hours. During critical production periods (e.g., when pressure fluctuates significantly), the sampling frequency can be increased to every 30 minutes to ensure that critical pressure changes are captured.

[0062] The numerical simulation unit 3 is in communication with the permeability distribution characteristic data acquisition unit 1 and the oil well test data acquisition unit 2, respectively, and is used to perform numerical modeling based on the permeability distribution characteristic data and the non-equilibrium seepage simulation mode, and to establish a simulation model based on the fitting and correction of the pressure measurement data and production performance data. Preferably, the non-equilibrium seepage simulation mode is described by the following formula:

[0063] ,

[0064] in: is the reservoir pressure, in MPa; 、 、 They are 、 、 Permeability in the direction, in mD; is the porosity, dimensionless; is the fluid viscosity, in mPa·s; is the comprehensive compression coefficient, in units of ; is the time in hours; is the source-sink term, which indicates the injection and production volume per unit volume, and its unit is m h); 、 、 and Respectively express 、 、 and Partial derivative operations.

[0065] In one embodiment of the present invention, model fitting and correction employ a history matching algorithm to minimize the error between simulation results and actual production data by adjusting key parameters (such as permeability distribution and oil-water relative permeability curve). Preferably, the following objective function is used for history matching:

[0066] ,

[0067] in: is the parameter vector to be adjusted, and its dimension is the number of adjustment parameters; and Respectively The simulated pressure value and the observed pressure value at each time point are in MPa; and Respectively The simulated and observed yield values ​​at each time point are in units of and are the number of observation points for pressure and production, respectively; and is the weight coefficient, dimensionless; and is the characteristic scale factor of the physical quantity, dimensionless; Indicates that all The sum of the pressure observation points is calculated; Indicates that all The sum is calculated for each production observation point.

[0068] The improvement of the present invention is that the physical quantity characteristic scale factor is introduced and ,These factors take into account the differences in sensitivity of different physical quantities in actual production. For example, for low permeability oil fields, a small change in pressure may cause safety problems for production equipment or significantly affect flow characteristics, while the same relative change in production may have a smaller impact. The determination of characteristic scale factors can be based on engineering experience or obtained through sensitivity analysis. For example, assuming that under specific oil field conditions, a 10% relative change in pressure is equivalent to a 30% relative change in production in terms of production impact, you can set 3.0. By introducing these scaling factors, the normalized pressure error and yield error have considerable practical importance in the objective function.

[0069] On this basis, the weight coefficient and Can be set to be equal (such as At the same time, the actual influence weights of different physical quantities have been adjusted through characteristic scale factors. This approach not only maintains the simplicity of the objective function structure, but also takes into account the differences in the importance of different physical quantities in actual production, effectively solving the problem that simple normalization cannot reflect the actual importance of physical quantities.

[0070] In practice, the optimal characteristic scaling factor value can be determined through iterative optimization according to the specific oilfield characteristics and production targets, so that the history matching process can accurately fit the pressure and production curves at the same time, achieving more accurate model calibration.

[0071] The calculation unit 4 is in communication with the numerical simulation unit 3 and is used to calculate the direction of horizontal wellbore streamlines and the pressure distribution thereon based on the simulation model to generate flow characteristic data. In a preferred embodiment of the present invention, the calculation unit 4 uses a particle tracking method to calculate the streamline distribution. The basic principle of this method is to release tracer particles in the flow field and obtain the particle trajectory, i.e., the streamline distribution, by solving the particle motion equation. The particle motion equation can be expressed as:

[0072] ,

[0073] in: is the particle position vector, with unit m and dimension 3 (representing coordinates in three-dimensional space); is the time in hours; is the velocity vector of the flow field, in m / h, with the same dimension as same; Represents the particle position vector About time The total derivative of represents the rate of change of the particle position with time. This velocity vector can be calculated using Darcy's law:

[0074] ,

[0075] in: is the permeability tensor, with the unit of mD and the dimension of 3×3, indicating the permeability in different directions; is the porosity, dimensionless; is the fluid viscosity, in mPa·s; is the pressure gradient vector, with the unit of MPa / m and dimension of 3; Represents the dot product operation between a tensor and a vector.

[0076] Optimization unit 5 is communicatively connected to calculation unit 4 and oil well test data acquisition unit 2, and is configured to couple flow characteristic data with horizontal well characteristic data based on the low permeability oilfield to generate an optimization solution. The horizontal well characteristic data includes production dynamic pressure data and water absorption profile data for individual wells in the block where the horizontal wells are located in the low permeability oilfield.

[0077] like Figure 2 As shown, the optimization unit 5 further includes: a test model building subunit 51, a numerical simulation algorithm optimization subunit 52, an oil layer parameter division subunit 53, a data coupling subunit 54 and a solution output subunit 55.

[0078] The test model building subunit 51 is used to receive water absorption profile data from the horizontal well single well characteristic data, apply the wavelet transform algorithm to perform multi-scale decomposition on the water absorption profile data, identify abnormal water absorption areas in the oil layer through adaptive threshold division, establish a multi-dimensional profile model that takes into account geological heterogeneity, divide the injection and production methods of the single well oil layer, and establish a test model. Preferably, the wavelet transform algorithm uses the following formula:

[0079] ,

[0080] in: For function The continuous wavelet transform result is about the scale parameter and translation parameters Two-dimensional function of ; The water absorption profile data to be analyzed is about the spatial position function; is the mother wavelet function; express The complex conjugate of is the scale parameter, which controls the expansion and contraction of the wavelet. ; is the translation parameter, which controls the translation of the wavelet; is the spatial position, the unit is m; Represents the integral operation from negative infinity to positive infinity; is the normalization factor, ensuring that wavelets at different scales have the same energy.

[0081] In practical applications, discrete wavelet transform is preferably used for calculation, which can be expressed as:

[0082] ,

[0083] in: is the result of discrete wavelet transform, which is about the discrete scale parameter and discrete translation parameters Two-dimensional function of ; and are the discrete parameters of scale and translation, both are integers; is the discrete spatial position index, corresponding to the sampling point; is the water absorption profile data at discrete sampling points; is the discrete mother wavelet function; represents the summation over all discrete spatial positions; is the normalization factor.

[0084] In practical applications, Daubechies wavelet (such as DB4) can be selected as the mother wavelet and decomposed at levels 3 to 5 to capture features at different scales. For the determination of the adaptive threshold, the following strategy is adopted:

[0085] ,

[0086] in: For the The threshold of the level decomposition is used to determine whether the wavelet coefficients of this level represent meaningful features; For the The standard deviation of the wavelet coefficients can be obtained by statistically analyzing the All wavelet coefficients of level are calculated; For the The number of level wavelet coefficients; Indicates the natural logarithm operation. It can be robustly estimated by dividing the absolute median of the wavelet coefficients by 0.6745, as follows: median , where median represents the median operation, Indicates taking the absolute value.

[0087] The numerical simulation algorithm optimization subunit 52 is in communication with the test model construction subunit 51, and is used to receive the test model and flow characteristic data, use adaptive grid refinement technology to perform local optimization on abnormal areas, introduce a dynamic boundary condition self-adjustment mechanism, adjust model parameters through the gradient descent optimization algorithm, optimize the numerical simulation algorithm, and obtain an optimized model.

[0088] The adaptive mesh refinement technique performs local mesh refinement based on an error indicator, which is defined as:

[0089] ,

[0090] in: For unit The error indicator on the ,is used to measure the size of the calculation error of the unit; For unit The size is usually taken as the maximum side length of the unit, in meters; For unit Error estimate of upper pressure gradient, in MPa / m; Indicates in unit on norm, defined as ,in Indicates in unit The integral on Represents the volume element. Greater than the preset threshold (usually taken as 1.5 times the average error indicator), the unit is encrypted.

[0091] The gradient descent optimization algorithm is used to adjust the model parameters, and the iterative formula is:

[0092] ,

[0093] in: For the The parameter vector of the iteration, whose dimension is equal to the number of parameters to be optimized; For the The parameter vector of the iteration; For the The learning rate of the iteration is a positive scalar; is the objective function About parameters At the point The gradient vector at , whose dimension is Same. Gradient vector No. The component is , represents the objective function For the first parameters The partial derivative of .

[0094] Learning rate Adopting an adaptive adjustment strategy:

[0095] ,

[0096] in: is the initial learning rate, which is a positive scalar and preferably takes a value of 0.1; is the attenuation coefficient, which is a positive scalar and preferably takes a value of 0.01; is the number of iterations. This strategy makes the learning rate gradually decrease as the number of iterations increases, which helps the algorithm converge.

[0097] The oil layer parameter division subunit 53 is communicated with the numerical simulation algorithm optimization subunit 52, and is used to receive the optimization model and production dynamic pressure data. The water absorption profile data of the oil layer is divided into injection unit parameters using the spatiotemporal response prediction model based on the LSTM neural network, and the injection water flow distribution characteristic data of each unit layer is calculated. The injection water flow distribution characteristic data includes the flow characteristics of the injected water and / or the distribution characteristics of the injected water.

[0098] The core structure of the LSTM neural network includes the forget gate, input gate, output gate, and unit state. Its operation process is as follows:

[0099] ,

[0100] ,

[0101] ,

[0102] ,

[0103] ,

[0104] ,

[0105] in: is the output of the forget gate, the range is [0,1], and the dimension is same; is the output of the input gate, the range is [0,1], and the dimension is same; is the output of the output gate, the range is [0,1], and the dimension is same; is the cell state at the current time step, with the same dimensions as same; is the candidate unit state, dimension and same; is the hidden state output of the current time step, and its dimension is the number of hidden layer nodes is the hidden state output of the previous time step; is the input vector of the current time step, which contains the water absorption profile data and pressure data of the current time step, and its dimension is the number of input features; 、 、 、 They are the weight matrices of the forget gate, input gate, candidate unit state calculation and output gate, respectively, with dimensions of ,in is the number of hidden layer nodes, is the input vector dimension; 、 、 、 They are the bias vectors of the forget gate, input gate, candidate unit state calculation and output gate, respectively, and their dimensions are is the sigmoid activation function, defined as ,map the input to the interval [0,1]; is the hyperbolic tangent activation function, defined as tanh , maps the input to the interval [-1,1]; © represents the Hadamard product (element-wise multiplication), that is, multiplying the corresponding elements of two vectors or matrices of the same dimension; Represents a vector concatenation operation.

[0106] In a preferred embodiment of the present invention, the LSTM network is configured as a three-layer network with 128, 64, and 32 hidden layer nodes, respectively. The learning rate is set to 0.001, the batch size is 64, and the number of training rounds is 1000. By continuously optimizing the network parameters, the error between the prediction results and the actual production data is minimized.

[0107] The data coupling subunit 54 is respectively connected to the oil layer parameter division subunit 53 and the calculation unit 4 for establishing a nonlinear mapping relationship between the flow characteristics of the injected water and the water saturation, receiving the flow characteristic data, coupling the injected water flow distribution characteristic data of each unit layer with its injected water saturation, and using the Bayesian optimization algorithm to search the parameter space to form an optimization scheme for the optimization unit.

[0108] The Bayesian optimization algorithm builds a proxy model of the objective function based on the Gaussian process regression model and guides the parameter space search through the acquisition function. Its core process is as follows:

[0109] (1) Constructing a Gaussian process regression model:

[0110] ,

[0111] in: is the objective function, which is about the parameter vector function; is the parameter vector, and the dimension is the dimension of the parameter space; represents a Gaussian process; is the mean function, which means The prior mean of is usually taken as 0; is the kernel function, describing two points in space and relevance; It means "obey" or "distributed as". It is preferred to use the radial basis function (RBF) kernel:

[0112] ,

[0113] in: is the signal variance, which is a positive scalar that controls the amplitude of the function value; is the length scale, a positive scalar that controls the smoothness of the function change; Represents a vector and The Euclidean distance between ,in is the vector dimension. , .

[0114] (2) Collection function selection:

[0115] It is preferred to use Expected Improvement (EI) as the acquisition function:

[0116] ,

[0117] in: For the parameter point Expected improvement at is the currently known optimal function value; represents the expected value operation; Indicates taking and The larger value in .

[0118] The analytical expression of EI is:

[0119] EI(x) = (μ(x) - f(x+))Φ(Z) + σ(x)φ(Z),

[0120] where μ(x) and σ(x) are the mean and standard deviation of the Gaussian process prediction, respectively, and are both scalars; Z = μ(x) - f(x+) / σ(x) is a scalar; Φ(·) and φ(·) are the cumulative distribution function and probability density function of the standard normal distribution, respectively, defined as and .

[0121] (3) Iterative optimization process:

[0122] a. Initialize the parameter space sampling point set ,in For the The parameter vector of sampling points, is the corresponding function value, is the number of initial sampling points;

[0123] b. Based on Construct a Gaussian process regression model;

[0124] c. Find the next sampling point by maximizing the acquisition function , where argmax represents the independent variable that makes the function reach its maximum value;

[0125] d. Evaluate the objective function value of the new sampling point , update the sampling point set , where U represents the union operation of sets;

[0126] e. Repeat steps bd until the preset number of iterations or convergence condition is reached.

[0127] In a preferred embodiment of the present invention, the number of iterations is set to 50, and the convergence condition is that the improvement of the optimal value of 5 consecutive iterations is less than 0.1%.

[0128] The solution output subunit 55 is in communication with the data coupling subunit 54 and is used to receive the optimization solution, adopt a real-time adjustment algorithm of injection volume based on fuzzy control, introduce a multi-objective optimization decision-making mechanism of production-injection volume, realize equipment health status monitoring, and adjust the injection measures of each unit layer in the single well oil layer according to the optimization solution.

[0129] The fuzzy control algorithm includes three steps: fuzzification, fuzzy reasoning and defuzzification. In the present invention, the injection pressure and the water absorption ratio are used as input variables, and the injection volume adjustment is used as the output variable.

[0130] (1) Fuzzification: Convert the exact value into a fuzzy set. The input variable "injection pressure" is set to five fuzzy sets: {very low, low, medium, high, very high}; the input variable "water absorption ratio" is set to five fuzzy sets: {very poor, poor, medium, good, very good}; the output variable "injection volume adjustment" is set to seven fuzzy sets: {substantially reduce, moderately reduce, slightly reduce, maintain unchanged, slightly increase, moderately increase, substantially increase}.

[0131] The membership function preferably adopts a triangular or trapezoidal function. For example, the triangular membership function can be expressed as:

[0132] ,

[0133] in: For elements Fuzzy Set The membership degree of is in the range of [0,1]; is the exact value of the input variable; 、 、 are the three characteristic points of the triangle membership function, satisfying ,in is the point where the membership reaches the maximum value of 1, and are two endpoints with membership degree 0.

[0134] (2) Fuzzy reasoning: Establish a fuzzy rule base, such as:

[0135] If the injection pressure is high and the water absorption ratio is poor, the injection volume is greatly reduced;

[0136] If the injection pressure is moderate and the water absorption ratio is moderate, the injection volume remains unchanged;

[0137] If the injection pressure is low and the water absorption ratio is good, the injection volume can be greatly increased.

[0138] The fuzzy output is calculated by the Mamdani inference method, using the minimum-maximum synthesis operation:

[0139] ,

[0140] in: Output variables in the inference results The degree of membership; For input variables The degree of membership; For fuzzy relationship Input value and output value The corresponding membership degree; Represents a set All elements in Take the maximum value; express and The smaller value in .

[0141] (3) Defuzzification: Convert the fuzzy output to an exact value, preferably using the centroid method:

[0142] ,

[0143] in: is the exact output value after defuzzification; is the domain of the output variable, that is, the set of all possible output values; Output value in the fuzzy output set The degree of membership; Indicates that in the domain Integral operations on ; express and The product of .

[0144] The production-water injection rate multi-objective optimization decision-making mechanism adopts the linear weighting method, and the objective function is:

[0145] ,

[0146] in: It is a comprehensive evaluation indicator used to measure the overall effect of the optimization plan; is the normalized value of the yield index, indicating the yield improvement effect; is the normalized value of the water injection efficiency index, indicating the water injection utilization efficiency; is the normalized value of the input cost indicator, indicating the control effect of implementation cost; 、 、 is the weight coefficient, which indicates the relative importance of each indicator. and . Optimal settings 、 、 , to balance production increase, water injection efficiency and cost control.

[0147] Control unit 6 is in communication with optimization unit 5 and is configured to control optimization unit 5 to feed back the optimization solution to a downhole data acquisition unit in the horizontal well. The downhole data acquisition unit then adjusts injection measures within the horizontal wellbore based on the optimization solution, optimizing flow characteristics within the horizontal wellbore and thereby increasing the horizontal well's production. In a preferred embodiment of the present invention, the downhole data acquisition unit includes at least a water absorption profile tester, a pressure gauge, a pressure sensor, and a flow meter, configured to collect downhole data in real time and execute injection adjustments.

[0148] The verification unit 7 is communicated with the control unit 6 and the optimization unit 5 respectively, and is used to test the effect of the optimization scheme on the downhole data acquisition unit of another well in the low permeability oil field according to the optimization scheme, and feed back the test results to the control unit 6.

[0149] To verify the horizontal wellbore streamline optimization method, the area with the smallest change in streamline diversion ratio in the horizontal wellbore streamline optimization method is selected to carry out downhole dynamic monitoring and obtain dynamic monitoring results. Through downhole dynamic monitoring, it is judged whether the dynamic monitoring results have achieved the set effect. If it is determined that the dynamic monitoring results have achieved the set effect, it is proved that the optimization scheme is effective. If it is determined that the dynamic monitoring results have not achieved the set effect, it is proved that the optimization scheme is not effective.

[0150] Preferably, the effects set include: a yield increase of at least 15%, a moisture content reduction of at least 5 percentage points, or a comprehensive benefit index increase of at least 20%. The dynamic monitoring period is preferably set to 30 days to fully evaluate the long-term effects of the optimization plan.

[0151] The logging unit 8 is respectively connected to the permeability distribution characteristic data acquisition unit 1 and the optimization unit 5 for designing a logging plan based on the actual oil layer conditions in the block where the horizontal well is located, and collecting downhole dynamic test data according to the logging plan; the downhole dynamic test data is combined with the optimization plan to form a final optimization plan.

[0152] like Figure 9As shown, the method for improving the production of horizontal wells in low permeability oil fields based on the above system includes the following steps:

[0153] Step S1, using the permeability distribution characteristic data acquisition unit 1 to acquire permeability data of each oil layer of a horizontal well in a low permeability oil field;

[0154] Step S2: using the oil well test data acquisition unit 2 to collect pressure measurement data and production performance data of the horizontal well;

[0155] Step S3, using the numerical simulation unit 3 to perform numerical modeling based on the permeability distribution characteristic data and the non-equilibrium seepage simulation mode, and fitting and correcting based on the pressure measurement data and production dynamic data to establish a simulation model;

[0156] Step S4, using the calculation unit 4 to calculate the direction of the horizontal wellbore streamline and the pressure distribution thereon according to the simulation model to form flow characteristic data;

[0157] Step S5, using the optimization unit 5 to couple the flow characteristic data with the horizontal well characteristic data based on the low permeability oil field single well characteristic data to obtain an optimization solution;

[0158] Step S6: The control unit 6 controls the optimization unit 5 to feed back the optimization plan to the downhole data acquisition unit in the horizontal well oil well. The downhole data acquisition unit adjusts the injection measures in the horizontal wellbore according to the optimization plan, optimizes the flow characteristics in the horizontal wellbore, and thus improves the production of the horizontal well.

[0159] Wherein, in step S5:

[0160] The test model building subunit 51 receives water absorption profile data from the horizontal well single well characteristic data, applies a wavelet transform algorithm to perform multi-scale decomposition on the water absorption profile data, identifies abnormal water absorption areas of the oil layer through adaptive threshold division, establishes a multi-dimensional profile model considering geological heterogeneity, divides the injection and production methods of the single well oil layer, and establishes a test model;

[0161] The numerical simulation algorithm optimization subunit 52 receives the test model and flow characteristic data, uses the adaptive grid refinement technology to perform local optimization on the abnormal area, introduces a dynamic boundary condition self-adjustment mechanism, adjusts the model parameters through the gradient descent optimization algorithm, optimizes the numerical simulation algorithm, and obtains the optimized model;

[0162] The reservoir parameter division subunit 53 receives the optimization model and production dynamic pressure data, and uses the spatiotemporal response prediction model based on the LSTM neural network to divide the water absorption profile data of the reservoir into injection unit parameters, and calculates the injection water flow distribution characteristic data of each unit layer;

[0163] A data coupling subunit 54 is used to establish a nonlinear mapping relationship between the flow characteristics of the injected water and the water saturation, receives the flow characteristic data, couples the injected water flow distribution characteristic data of each unit layer with its injected water saturation, and uses a Bayesian optimization algorithm to search the parameter space to form an optimization solution for the optimization unit;

[0164] The scheme output subunit 55 receives the optimization scheme, adopts a real-time adjustment algorithm of injection volume based on fuzzy control, introduces a multi-objective optimization decision-making mechanism of production-injection volume, realizes equipment health status monitoring, and adjusts the injection measures of each unit layer in the single well oil layer according to the optimization scheme.

[0165] Taking the horizontal well A in the M block of a low permeability oil field as an example, the specific implementation process and effect of the present invention are described.

[0166] The well's horizontal section is 1,200 meters long, and the reservoir in which it is located has an average permeability of 8.5 mD, a typical low-permeability oilfield. Initially, the well produced 35 tons of oil per day. Six months into production, this production dropped to 12 tons per day, while the water cut increased to 65%, indicating significant capacity decline.

[0167] The system and method of the present invention are used for optimization, and the specific steps are as follows:

[0168] (1) The permeability distribution characteristic data acquisition unit 1 was used to log the horizontal section of Well A and obtain the permeability distribution data of each oil layer. The logging results showed that the permeability distribution of the horizontal section was uneven and divided into three distinct sections: the permeability of the 0-400 m section was 5-7 mD, the permeability of the 400-800 m section was 12-15 mD, and the permeability of the 800-1200 m section was 3-5 mD.

[0169] (2) The oil well test data acquisition unit 2 was used to collect the pressure measurement data and production performance data of Well A. The pressure measurement data showed that the bottomhole pressure was 15.2 MPa, and the production performance data showed that there were large differences in the liquid production and water content of each section.

[0170] (3) A non-equilibrium seepage simulation model was established using numerical simulation unit 3, and key parameters were adjusted through historical fitting. The fitted model accurately reflects the production dynamic characteristics of Well A, with a historical fitting error of less than 5%.

[0171] (4) Calculation unit 4 is used to calculate the streamline direction and pressure distribution in the horizontal wellbore based on the simulation model. The calculation results show that the high-permeability section between 400 and 800 meters has a dominant flow channel, resulting in severe channeling of injected water in this section, while the low-permeability section has the problem of insufficient utilization of injected water.

[0172] (5) Optimization unit 5 is used to perform coupled calculation of flow characteristic data and single well characteristic data:

[0173] a. The test model building subunit 51 performs wavelet transform decomposition on the water absorption profile data and identifies three characteristic sections, which are basically consistent with the logging results;

[0174] b. Numerical simulation algorithm optimization subunit 52 performs local grid encryption on the high permeability section between 400 and 800 meters to improve simulation accuracy;

[0175] c. The reservoir parameter division subunit 53 builds a spatiotemporal response prediction model based on the LSTM neural network, divides the reservoir into five injection units, and calculates the injection water flow characteristics of each unit;

[0176] d. The data coupling subunit 54 uses a Bayesian optimization algorithm to find the optimal parameter combination and form a differentiated injection strategy;

[0177] e. The program output subunit 55 generates a specific injection adjustment program: increase the injection volume by 25% in the 0-400 meter segment, reduce the injection volume by 40% in the 400-800 meter segment, and increase the injection volume by 35% in the 800-1200 meter segment.

[0178] (6) The control unit 6 sends the optimization plan to the downhole data acquisition unit to implement the injection measures adjustment.

[0179] (7) 60 days after the implementation of the optimization plan, the optimization effect was evaluated using verification unit 7. The results showed that the daily oil production of Well A increased to 28 tons, an increase of 133%; the water cut decreased to 48%, a decrease of 17 percentage points; and the comprehensive benefit index increased by 185%, proving that the optimization plan had a significant effect.

[0180] In addition, applying the same optimization scheme to Wells B and C in the same block also achieved similar good results, with production increases of 89% and 112%, respectively, proving that the method has good scalability.

[0181] In summary, the system and method for improving the production of horizontal wells in low-permeability oil fields provided by the present invention achieve a significant increase in horizontal well production capacity by collecting multi-dimensional data, performing precise simulation based on a non-equilibrium seepage model, and combining advanced technologies such as adaptive multi-scale recognition, deep reinforcement learning, and multi-parameter coupling optimization, providing an effective technical means for the efficient development of low-permeability oil fields.

[0182] 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 various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A system for increasing the production of horizontal wells in low permeability oil fields, characterized in that: The system comprises: a permeability distribution characteristic data acquisition unit, for acquiring permeability data of each oil layer section of a horizontal well in the low permeability oil field; An oil well test data acquisition unit, communicatively connected to the permeability distribution characteristic data acquisition unit, for acquiring pressure measurement data and production dynamic data of the horizontal well; a numerical simulation unit, communicatively connected to the permeability distribution characteristic data acquisition unit and the oil well test data acquisition unit, for performing numerical modeling based on the permeability distribution characteristic data and a non-equilibrium seepage simulation mode, and establishing a simulation model based on fitting and correction of the pressure measurement data and the production performance data; a calculation unit, in communication with the numerical simulation unit, for calculating the direction of horizontal wellbore streamlines and the pressure distribution thereon according to the simulation model to form flow characteristic data; an optimization unit, communicatively connected to the calculation unit and the oil well test data acquisition unit, and configured to couple the flow characteristic data with the horizontal well characteristic data to obtain an optimization solution based on the horizontal well characteristic data of the low permeability oil field, wherein the horizontal well characteristic data includes production dynamic pressure data and water absorption profile data of the horizontal well in the block where the horizontal well is located; A control unit is communicatively connected to the optimization unit and is used to control the optimization unit to feed back the optimization scheme to a downhole data acquisition unit in the horizontal well oil well. The downhole data acquisition unit adjusts the injection measures in the horizontal wellbore according to the optimization scheme, optimizes the flow characteristics in the horizontal wellbore, and thus improves the production of the horizontal well.

2. The system according to claim 1, wherein: The optimization unit includes: A test model building subunit is used to receive the water absorption profile data in the horizontal well single well characteristic data, apply the wavelet transform algorithm to perform multi-scale decomposition on the water absorption profile data, identify the abnormal water absorption area of ​​the oil layer through adaptive threshold division, establish a multi-dimensional profile model considering geological heterogeneity, divide the injection and production mode of the single well oil layer, and establish a test model; a numerical simulation algorithm optimization subunit, which is in communication with the test model building subunit and is used to receive the test model and the flow characteristic data, perform local optimization on the abnormal area using an adaptive grid refinement technique, introduce a dynamic boundary condition self-adjustment mechanism, adjust model parameters using a gradient descent optimization algorithm, and optimize the numerical simulation algorithm to obtain an optimized model; an oil layer parameter division subunit, communicatively connected to the numerical simulation algorithm optimization subunit, for receiving the optimization model and the production dynamic pressure data, dividing the water absorption profile data of the oil layer into injection unit parameters using a spatiotemporal response prediction model based on an LSTM neural network, and calculating injection water flow distribution characteristic data for each unit layer, the injection water flow distribution characteristic data including flow characteristics and / or distribution characteristics of the injected water; The data coupling subunit is respectively communicated with the oil layer parameter division subunit and the calculation unit, and is used to establish a nonlinear mapping relationship between the injection water flow characteristics and the water saturation, receive the flow characteristic data, couple the injection water flow distribution characteristic data of each unit layer with its injection water saturation, and use the Bayesian optimization algorithm to search the parameter space to form an optimization scheme for the optimization unit.

3. The system according to claim 1, wherein: The optimization unit includes: An oil layer parameter division subunit is used to receive the flow characteristic data, divide the oil layer of the single well into injection unit parameters, and calculate the distribution characteristic data of the injected water in each unit layer; The data coupling subunit is in communication connection with the oil layer parameter division subunit and is used to couple the distribution characteristic data of the injected water of each unit layer with the injection and production unit parameters of each unit layer to form an optimization scheme for the optimization unit.

4. The system according to claim 2 or 3, characterized in that The optimization unit includes: The solution output subunit is communicatively connected to the data coupling subunit, and is used to receive the optimization solution, adopt a real-time adjustment algorithm of injection volume based on fuzzy control, introduce a multi-objective optimization decision-making mechanism of production-injection volume, realize equipment health status monitoring, and adjust the injection measures of each unit layer in the single well oil layer according to the optimization solution.

5. The system according to claim 4, characterized in that The system further comprises: The verification unit is respectively connected to the control unit and the optimization unit for testing the effect of the optimization scheme on the downhole data acquisition unit of another well in the low permeability oil field according to the optimization scheme, and feeding back the test results to the control unit.

6. The system according to claim 5, characterized in that The horizontal wellbore streamline optimization method is verified, and the area with the smallest change in streamline diversion ratio in the horizontal wellbore streamline optimization method is selected to carry out downhole dynamic monitoring and obtain dynamic monitoring results; through downhole dynamic monitoring, it is judged whether the dynamic monitoring results achieve the set effect; when it is determined that the dynamic monitoring results achieve the set effect, it is proved that the optimization scheme is effective; when it is determined that the dynamic monitoring results do not achieve the set effect, it is proved that the optimization scheme is not effective.

7. The system according to claim 6, characterized in that The pressure measurement data includes: pressure gradient data, pressure recovery data, and pressure dynamic data; the production dynamic data includes: production dynamic pressure data and production dynamic water absorption profile data.

8. The system according to claim 7, characterized in that The permeability distribution characteristic data acquisition unit includes at least one well logging device and / or coring device; the downhole data acquisition unit includes at least a water absorption profile tester, a pressure gauge, a pressure sensor and a flow meter.

9. The system according to claim 8, characterized in that The system further comprises: The logging unit is respectively communicated with the permeability distribution characteristic data acquisition unit and the optimization unit, and is used to design a logging plan based on the actual oil layer conditions of the block where the horizontal well is located, and to collect downhole dynamic test data according to the logging plan; the downhole dynamic test data is combined with the optimization plan to form a final optimization plan.

10. A method for increasing the production of horizontal wells in low permeability oil fields, characterized in that: The method is applied to the system according to any one of claims 1 to 9, and the method includes: Step S1, using a permeability distribution characteristic data acquisition unit to acquire permeability data of each oil layer of a horizontal well in the low permeability oil field; Step S2: using an oil well test data acquisition unit to acquire pressure measurement data and production performance data of the horizontal well; Step S3, using a numerical simulation unit to perform numerical modeling based on the permeability distribution characteristic data and a non-equilibrium seepage simulation mode, and fitting and correcting based on the pressure measurement data and the production dynamic data to establish a simulation model; Step S4: using a calculation unit to calculate the direction of the horizontal wellbore streamline and the pressure distribution thereon according to the simulation model to form flow characteristic data; Step S5: using an optimization unit to couple the flow characteristic data with the horizontal well characteristic data of the low permeability oil field to obtain an optimization solution; Step S6: Using the control unit to control the optimization unit to feed back the optimization plan to the downhole data acquisition unit in the horizontal well, and the downhole data acquisition unit adjusts the injection measures in the horizontal wellbore according to the optimization plan to optimize the flow characteristics in the horizontal wellbore, thereby increasing the production of the horizontal well; Wherein, in step S5: A test model building subunit is used to receive water absorption profile data in the horizontal well single well characteristic data, a wavelet transform algorithm is applied to perform multi-scale decomposition on the water absorption profile data, an abnormal water absorption area of ​​the oil layer is identified through adaptive threshold division, a multi-dimensional profile model considering geological heterogeneity is established, the injection and production mode of the single well oil layer is divided, and a test model is established; A numerical simulation algorithm optimization subunit is used to receive the test model and the flow characteristic data, an adaptive grid refinement technology is used to perform local optimization on the abnormal area, a dynamic boundary condition self-adjustment mechanism is introduced, model parameters are adjusted by a gradient descent optimization algorithm, and the numerical simulation algorithm is optimized to obtain an optimized model; An oil layer parameter division subunit is used to receive the optimization model and the production dynamic pressure data, and a spatiotemporal response prediction model based on an LSTM neural network is used to divide the water absorption profile data of the oil layer into injection unit parameters, and the injected water flow distribution characteristic data of each unit layer is calculated; A data coupling subunit is used to establish a nonlinear mapping relationship between the flow characteristics of the injected water and the water saturation, the flow characteristic data is received, the injected water flow distribution characteristic data of each unit layer is coupled with its injected water saturation, and a Bayesian optimization algorithm is used to search the parameter space to form an optimization solution for the optimization unit; A scheme output subunit is used to receive the optimization scheme, a real-time adjustment algorithm for injection volume based on fuzzy control is adopted, a multi-objective optimization decision-making mechanism for production volume and injection volume is introduced to realize equipment health status monitoring, and injection measures for each unit layer in the single well oil layer are adjusted according to the optimization scheme.

Images