Method for predicting productivity of tight oil reservoir, electronic device and storage medium

Abstract

This invention discloses a method, electronic device, and storage medium for predicting the production capacity of tight oil reservoirs. The method includes: establishing a mathematical model of seepage flow in horizontal wells of tight oil reservoirs based on the characteristics of multi-stage fracturing development; establishing a multi-stage fracturing production capacity model for horizontal wells of tight oil reservoirs based on the mathematical model of seepage flow in horizontal wells of tight oil reservoirs; calculating tight oil production capacity data under different fracturing completion parameters and formation parameters through orthogonal experiments based on the multi-stage fracturing production capacity model; normalizing the fracturing completion parameters, formation parameters, and their corresponding tight oil production capacity data, and dividing the normalized dataset into a training set and a test set; constructing a tight oil production capacity prediction model based on machine learning and setting hyperparameters for the tight oil production capacity prediction model; training the tight oil production capacity prediction model using the training set; and using the trained tight oil production capacity prediction model to predict tight oil production capacity. This method achieves rapid and effective prediction of tight oil production.

Description

Technical Field

[0001] This invention relates to the field of petroleum exploration and development, and more specifically, to a method for predicting the production capacity of tight oil reservoirs, an electronic device, and a storage medium. Background Technology

[0002] Currently, tight oil reservoirs are developing rapidly and have become a key area of ​​industrial investment. China's tight oil reservoirs have complex geological conditions, and reservoir development is still in its early stages. Currently, methods such as horizontal well multi-stage fracturing and volumetric fracturing are commonly used for oil and gas extraction from tight reservoirs. Accurately predicting production capacity has a significant impact on oil and gas development design.

[0003] There are two main methods for establishing capacity forecasting models: one is to use analytical, analytical-semi-analytical, and numerical simulation methods, which predict capacity using capacity formulas. The process is as follows:

[0004] Step 1: Obtain the reservoir static parameters or production parameters of tight oil and gas reservoirs and perform relevant processing;

[0005] Step 2: Identify the factors affecting the productivity of horizontal wells in tight oil and gas reservoirs;

[0006] Step 3: Establish a comprehensive evaluation system or production prediction formula for fractured horizontal wells;

[0007] Step 4: Predict tight oil production capacity based on the production capacity prediction model.

[0008] However, the production capacity of tight oil and gas reservoirs is affected by many factors, including reservoir modification methods, geological conditions, and production systems. Moreover, the effectiveness of on-site fracturing development, such as fracture distribution and fracture aperture, is difficult to describe accurately. Therefore, production capacity prediction methods based on empirical formulas have low calculation accuracy, complex numerical simulation modeling, and long time requirements. They also require certain background experience and have significant limitations.

[0009] Another approach is to use machine learning methods, such as LSTM (Long Short-Term Memory) neural network models, Bayesian classification models, and support vector machine models, for prediction. The process is as follows:

[0010] Step 1: Obtain production parameters and reservoir static parameters;

[0011] Step 2: Build a machine learning model;

[0012] Step 3: Conduct capacity forecasting.

[0013] Machine learning methods are convenient and quick, but it is difficult to obtain field data after fracturing, the data volume is small, and there is no seepage mechanism of tight oil and gas reservoirs as physical support, resulting in poor accuracy and reliability. Summary of the Invention

[0014] The purpose of this invention is to propose a method, electronic device and storage medium for predicting the production capacity of tight oil reservoirs, so as to achieve rapid and effective prediction of tight oil production.

[0015] In a first aspect, the present invention proposes a method for predicting the productivity of tight oil reservoirs, comprising:

[0016] Based on the characteristics of multi-stage fracturing development in horizontal wells of tight oil reservoirs, a mathematical model for seepage during multi-stage fracturing development in horizontal wells of tight oil reservoirs is established.

[0017] Based on the mathematical model of seepage development in multi-stage fracturing, a multi-stage fracturing production capacity model for horizontal wells in tight oil reservoirs is established.

[0018] Based on the multi-stage fracturing production capacity model, tight oil production capacity data under different fracturing completion parameters and formation parameters were calculated through orthogonal experiments.

[0019] Normalize the fracturing completion parameters, formation parameters and their corresponding tight oil production data, and divide the normalized dataset into training set and test set;

[0020] A tight oil production capacity prediction model based on machine learning is constructed and hyperparameters are set for the tight oil production capacity prediction model.

[0021] The tight oil production capacity prediction model is trained using the training set.

[0022] Predict tight oil production capacity using the trained tight oil production capacity prediction model.

[0023] Optionally, establishing a mathematical model for the seepage flow in a horizontal well of a tight oil reservoir through multi-stage fracturing includes:

[0024] Based on the characteristics of multi-stage fracturing development of horizontal wells in tight oil reservoirs, the flow field structure of horizontal wells in tight oil reservoirs is divided into horizontal wellbore zone, strongly stimulated zone, weakly stimulated zone, matrix activated zone, and unactivated zone.

[0025] The seepage resistance of the heavily modified zone is calculated using the following formula:

[0026]

[0027] In the formula: R fi To enhance the seepage resistance of the modified area, k fi Where μ is the permeability, h is the fluid viscosity, and ω is the reservoir thickness. fi x is the crack width. f For half the length of the crack, r w Where is the wellbore radius;

[0028] The seepage resistance of the weakly modified zone is calculated using the following formula:

[0029]

[0030] In the formula: R ni The seepage resistance of the weakly modified zone is given by μ, where μ is the fluid viscosity and k is the viscosity. ni r represents the permeability of the weakly modified zone, h represents the reservoir thickness, and r represents the permeability of the weakly modified zone. n r is the seepage radius at the boundary between the weakly modified zone and the matrix mobilization zone. f The seepage radius at the boundary between the heavily modified zone and the lightly modified zone;

[0031] The seepage resistance of the matrix mobilization zone is calculated using the following formula:

[0032]

[0033] In the formula: R mi k is the seepage resistance in the matrix mobilization zone. mi r is the permeability of the matrix activation zone, h is the reservoir thickness, and r is the permeability of the reservoir. e r is the seepage radius at the boundary between the unused matrix zone and the used matrix zone. n The seepage radius is located at the boundary between the matrix zone and the weakly modified zone.

[0034] Optionally, the calculation formula for the multi-stage fracturing productivity model of horizontal wells in tight oil reservoirs is as follows:

[0035]

[0036] In the formula, Q represents production capacity, and p e To alleviate supply pressure, p wi For the bottom hole pressure, G ni ′ represents the equivalent starting pressure gradient in the weakly modified region, G mi ′ represents the equivalent initiation pressure gradient of the matrix mobilization zone, S i Let be the area of ​​the overlapping region between adjacent crack segments, 'a' be the length of the major axis of the elliptical flow field, 'b' be the length of the minor axis of the elliptical flow field, and 'r' be the area of ​​the overlapping region between adjacent crack segments. e r is the seepage radius at the boundary between the unused matrix zone and the used matrix zone. n Let r be the seepage radius at the boundary between the matrix region and the weakly modified region. f R is the seepage radius at the boundary between the heavily modified zone and the lightly modified zone. fi To enhance the seepage resistance of the modified area, R ni R represents the seepage resistance in the weakly modified zone. mi The seepage resistance in the matrix mobilization zone.

[0037] Optionally, the fracturing completion parameters and formation parameters include: horizontal well length, number of single-end fracturing clusters, fracture width, matrix permeability, original formation pressure, bottom hole flowing pressure, extent of the stimulated zone and matrix zone, and rock compressibility coefficient.

[0038] Optionally, normalized fracturing completion parameters, formation parameters, and their corresponding tight oil production data include:

[0039] The fracturing completion parameters, formation parameters, and their corresponding tight oil production data are normalized using the following formula:

[0040]

[0041] Where X represents the raw data of fracturing completion parameters, formation parameters, and tight oil production capacity data. min X max X represents the minimum and maximum values ​​of the original dataset, respectively. norm This is the normalized data.

[0042] Optionally, the step of constructing a tight oil production capacity prediction model based on machine learning and setting hyperparameters for the tight oil production capacity prediction model includes:

[0043] A deep neural network model is constructed as the tight oil production capacity prediction model. The hyperparameters of the deep neural network model include: the number of neural network layers, the number of hidden layer neurons, the activation function type, the weight initialization method, and the optimizer settings.

[0044] The deep neural network model is configured with one input layer, four hidden layers, and one output layer, with 50 neurons in each layer. The activation function is set to ReLU, and the network propagation method is as follows:

[0045]

[0046]

[0047] Where f() represents the activation function, i is the index of the hidden layer of the network, j is the index of the hidden layer neuron, z represents the input of the i-th layer, y represents the output of the i-th layer, and w and b represent the weights and biases of the i-th layer, respectively.

[0048] Weights are initialized to the Xavier method:

[0049]

[0050] In the formula, To represent any hidden layer, w i Let Var[w] be the weight of the i-th layer. i ] indicates w i The variance, n i Enter the number;

[0051] Wherein, the weight w of the i-th layer i Initialize using a Gaussian distribution:

[0052]

[0053] The Adam optimizer is used, and the specific solution is as follows:

[0054]

[0055] m t ←β1m t-1 +(1-β1)g t

[0056] v t ←β2v t-1 +(1-β2)g t 2

[0057]

[0058]

[0059]

[0060] Where t is the time step, g t Let g be the gradient at time step t. t 2 =g t ⊙g t ;θ t-1 Let be the parameter vector at time step t-1. The gradient is represented by J(θ), which is the stochastic objective function of parameter θ. α is the step size, defaulting to 0.001. β1, β2 (∈ 0, 1) represent the exponential decay rate of the moment estimation, with β1 defaulting to 0.9 and β2 defaulting to 0.999. t Let t be the first moment vector at time step t. For the correction of the first-order moment vector estimate at time step t, v t Let t be the second moment vector at time step t. For the correction of the second-order moment vector estimate at time step t, ε is a parameter to prevent the denominator from being zero, and is set to 10. -8 .

[0061] Optionally, training the tight oil production capacity prediction model using the training set includes:

[0062] The training set is input into the tight oil production capacity prediction model for training. During the training process, the mean square error value is used as the error standard, and the tight oil production capacity prediction model is physically constrained by the multi-stage fracturing production capacity model.

[0063] The MSE value is calculated using the following formula:

[0064]

[0065] In the formula, MSE is the mean square error value, and y i Represents the true value. This represents the predicted value, where N is the sample size.

[0066] The physical constraints applied to the tight oil production prediction model using a multi-stage fracturing production model include:

[0067] The obtained parameter values ​​are substituted into the multi-stage fracturing capacity model for calculation, and the error function is designed as follows:

[0068]

[0069] Among them, phy_MSE design error value, Q represents the data predicted by the multi-stage fracturing capacity model. i This represents real data, where N is the number of samples;

[0070] The loss function is then:

[0071] loss = MSE + phy_MSE

[0072] In the formula, loss is the loss value;

[0073] If the loss value is less than 10 after training is complete. -2 If the loss value is greater than or equal to 10, then the model has completed training. -2 Then, the tight oil production capacity prediction model will continue to be trained.

[0074] Optionally, after training the tight oil production capacity prediction model using the training set, the method further includes:

[0075] The trained tight oil production capacity prediction model is tested using the test set. The production capacity prediction results of the tight oil production capacity prediction model are compared with the actual values ​​to determine whether the prediction accuracy meets the requirements. If it does not meet the requirements, the hyperparameters are modified until the requirements are met.

[0076] The prediction accuracy of the tight oil production capacity prediction model is determined by the following formula:

[0077]

[0078] Among them, R 2 As the coefficient of determination, y i Represents the true value. Indicates the predicted value. This represents the average value.

[0079] Secondly, the present invention provides an electronic device, the electronic device comprising:

[0080] At least one processor; and,

[0081] A memory communicatively connected to the at least one processor; wherein,

[0082] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the tight reservoir productivity prediction method described in the first aspect.

[0083] Thirdly, the present invention provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the tight oil reservoir production capacity prediction method described in the first aspect.

[0084] The beneficial effects of this invention are as follows:

[0085] This invention first establishes a mathematical model for the seepage development of multi-stage fracturing in horizontal wells of tight oil reservoirs and a multi-stage fracturing production capacity model for horizontal wells of tight oil reservoirs. Then, based on the multi-stage fracturing production capacity model, orthogonal experiments are used to calculate tight oil production capacity data under different fracturing completion parameters and formation parameters. Next, the fracturing completion parameters, formation parameters, and their corresponding tight oil production capacity data are normalized, and the normalized dataset is divided into training and testing sets. Then, a machine learning-based tight oil production capacity prediction model is constructed, and hyperparameters are set for the tight oil production capacity prediction model. The training set is used to train the tight oil production capacity prediction model, and finally, the trained tight oil production capacity prediction model is used to predict tight oil production capacity. This invention combines seepage laws with machine learning to establish a tight oil production capacity prediction method. This method can accurately predict tight oil production capacity under different fracturing conditions, with a prediction accuracy of over 90%. Compared with numerical simulation methods, it has higher prediction accuracy and faster speed.

[0086] The apparatus of the present invention has other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description

[0087] The above and other objects, features and advantages of the present invention will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0088] Figure 1 A flowchart illustrating the steps of a tight oil reservoir productivity prediction method according to Example 1 is shown.

[0089] Figure 2 The diagram shows a schematic of the flow field structure zoning in a horizontal well of a tight oil reservoir in an example of a tight oil reservoir productivity prediction method.

[0090] Figure 3 The regression analysis diagram of the tight oil reservoir production capacity prediction model in Example 1 is shown. Detailed Implementation

[0091] In existing technologies, tight oil production capacity prediction mainly relies on numerical models. However, predictions based on production data cannot accurately reflect the complex mechanisms of tight oil seepage, and numerical calculations are slow and time-consuming. Therefore, this invention provides a tight oil production capacity prediction method based on seepage patterns and machine learning, enabling rapid and effective prediction of tight oil reservoir production.

[0092] The invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0093] Example 1

[0094] Figure 1 A flowchart illustrating the steps of a tight oil reservoir productivity prediction method according to Example 1 is shown.

[0095] like Figure 1 As shown, a method for predicting the productivity of tight oil reservoirs includes:

[0096] Step S1: Based on the characteristics of multi-stage fracturing development of horizontal wells in tight oil reservoirs, establish a mathematical model for seepage flow in multi-stage fracturing development of horizontal wells in tight oil reservoirs;

[0097] Specifically, in the multi-stage fracturing development process of horizontal wells in tight oil reservoirs, tight oil flows through the reservoir matrix (nanoscale), microfractures (micrometer scale), and fracturing fractures (millimeter scale), ultimately flowing into the horizontal wellbore (centimeter scale) and being lifted to the surface. The flow field structure is divided into three major zones and five sub-zones: I. Horizontal Wellbore Zone, II. Stimulated Zone (strong and weakly stimulated zones), and III. Unstimulated Zone (matrix activated and unactivated zones), such as... Figure 2 As shown.

[0098] In this embodiment, the following assumptions are made: ① The reservoir is an infinitely large, closed formation; ② Single-phase fluid seepage occurs, and the rock and fluid are slightly compressible, ignoring gravity; ③ n fractures are performed along the horizontal wellbore direction, assuming that the main fracture has limited conductivity and is uniformly distributed along the horizontal wellbore, with the fractures penetrating the entire oil layer longitudinally; ④ The effects of different starting pressure gradients are considered in the weakly stimulated zone and the matrix activation zone; ⑤ The influence of temperature on seepage is ignored.

[0099] (1) Horizontal wellbore area

[0100] The momentum balance equations for formation fluid flow inside a horizontal wellbore are as follows:

[0101] Δp·A-τ·πd·Δx=ρA(v1 2 -v2 2 )

[0102] In the formula: Δp is the pressure difference between the two ends of the horizontal wellbore; A is the cross-sectional area of ​​the wellbore, in meters. 2 τ is the pipe friction resistance; Δx is the wellbore length; ρ is the fluid density; v1 and v2 are the flow velocities; d is the wellbore diameter.

[0103] (2) Area undergoing major renovation

[0104] The strongly modified zone is the main fracture formed by hydraulic fracturing. The fracture aperture is typically on the order of millimeters, and its flow characteristics are characterized by linear flow within the fracture. Referring to the parallel plate model, its equation of motion is:

[0105]

[0106] In the formula: k fi q represents the permeability of the heavily modified zone; μ represents the fluid viscosity; q represents the permeability of the modified zone. i ω is the volumetric flow rate; h is the reservoir thickness; fi For crack opening, This represents the pressure gradient.

[0107] Boundary conditions: ① Inner boundary: at the junction of the heavily modified zone and the horizontal wellbore, i.e., when x = r w When, p = p wi p wi ① Pressure at the junction of the strong-modification zone and the horizontal wellbore; ② Outer boundary: at the tip of the main fracture, i.e., when x = x f When, p = p nfi p nfi The pressure at the tip of the main crack.

[0108] The seepage resistance R in the heavily modified zone can be obtained. fi for:

[0109]

[0110] In the formula: R fi To enhance the seepage resistance of the modified area, k fi Where μ is the permeability, h is the fluid viscosity, and ω is the reservoir thickness. fi x is the crack width. f For half the length of the crack, r w Where is the wellbore radius;

[0111] (3) Areas with weak transformation

[0112] The weakly modified zone is a network of derivative fractures formed by hydraulic fracturing. Its flow characteristics are characterized by elliptical radial flow of formation fluids within the derivative microfractures towards the main fracture. A radial steady-state seepage model is established:

[0113]

[0114] In the formula: r is the seepage radius, ρ is the fluid density, and v is the flow velocity.

[0115] The equation of motion is given by considering the starting pressure gradient G in the weakly modified region. ni For low-speed non-Darcy flow, we have:

[0116]

[0117] Boundary conditions: ① Inner boundary: at the boundary between the strongly modified region and the weakly modified region, i.e., r = r f When, p = p nfi p nfi ① Pressure at the boundary between the strongly modified zone and the weakly modified zone; ② Outer boundary: at the boundary between the weakly modified zone and the matrix mobilization zone, i.e., when r = r n When, p = p mni p mni Pressure at the boundary between the weakly modified zone and the matrix activation zone.

[0118] The seepage resistance R in the weakly modified zone ni for

[0119]

[0120] In the formula: R ni The seepage resistance of the weakly modified zone is given by μ, where μ is the fluid viscosity and k is the viscosity. ni r represents the permeability of the weakly modified zone, h represents the reservoir thickness, and r represents the permeability of the weakly modified zone. n r is the seepage radius at the boundary between the weakly modified zone and the matrix mobilization zone. f The seepage radius at the boundary between the heavily modified zone and the lightly modified zone;

[0121] (4) Matrix activation zone

[0122] The equation of motion is similar to that of the weakly modified zone, characterized by considering the initiation pressure gradient G of the tight reservoir matrix. miFor low-speed non-Darcy flow, we have:

[0123]

[0124] Boundary conditions: ① Inner boundary: at the junction of the matrix region and the weakly modified region, i.e., r = r n When, p = p mni ②Outer boundary: at the boundary between the unused area and the used area of ​​the matrix (supply radius), i.e., when r = r e When, p = p e p e The pressure at the boundary between the unused area and the used area of ​​the matrix.

[0125] seepage resistance in the matrix mobilization zone

[0126]

[0127] (5) Untouched substrate area

[0128] Unused areas of the substrate do not contribute to production, but as the pressure-affected area expands, unused areas of the substrate are gradually incorporated into the utilized areas of the substrate, thus achieving energy replenishment.

[0129] Step S2: Based on the multi-stage fracturing development seepage mathematical model, establish a multi-stage fracturing production capacity model for horizontal wells in tight oil reservoirs;

[0130] Specifically, based on the equivalent seepage resistance method, the flow field of multiple zones is connected in series for oil supply. By solving the seepage resistance of each zone simultaneously, the production of the i-th segment can be obtained. Then, considering the interference between multi-stage fractures, the production of n fractures is summed to obtain the production capacity calculation formula for the development of n segments of horizontal well fracturing in tight oil reservoirs:

[0131]

[0132] In the formula, Q represents production capacity, and p e To alleviate supply pressure, p wi For the bottom hole pressure, G ni ′ represents the equivalent starting pressure gradient in the weakly modified region, G mi ′ represents the equivalent initiation pressure gradient of the matrix mobilization zone, S i Let be the area of ​​the overlapping region between adjacent crack segments, 'a' be the length of the major axis of the elliptical flow field, 'b' be the length of the minor axis of the elliptical flow field, and 'r' be the area of ​​the overlapping region between adjacent crack segments. e r is the seepage radius at the boundary between the unused matrix zone and the used matrix zone. n Let r be the seepage radius at the boundary between the matrix region and the weakly modified region. f R is the seepage radius at the boundary between the heavily modified zone and the lightly modified zone. fi To enhance the seepage resistance of the modified area, R ni R represents the seepage resistance in the weakly modified zone. miThe seepage resistance in the matrix mobilization zone.

[0133] Step S3: Based on the multi-stage fracturing production capacity model, calculate tight oil production capacity data under different fracturing completion parameters and formation parameters through orthogonal experiments;

[0134] Specifically, fracturing completion parameters and formation parameters include horizontal well length, number of fracture clusters at one end, fracture width, matrix permeability, original formation pressure, bottom hole flowing pressure, extent of the stimulated zone and matrix zone, and rock compressibility. These fracturing completion parameters and formation parameters together constitute the dataset for the tight oil production prediction model.

[0135] Step S4: Normalize the fracturing completion parameters, formation parameters and their corresponding tight oil production data, and divide the normalized dataset into training set and test set;

[0136] Specifically, normalized fracturing and completion parameters, formation parameters, and their corresponding tight oil production data were used, and these were divided into training and testing sets. The first 70% of the dataset was used as the training set, and the last 30% as the testing set.

[0137] Significant differences in the magnitude of data can impact the results, leading to poor predictions. To address this issue, the dataset is normalized. The normalization method is as follows:

[0138]

[0139] Where X is the original data, X min X max X represents the minimum and maximum values ​​of the original dataset, respectively. norm The data is normalized, so that the normalized values ​​can be mapped to the interval [0,1].

[0140] Step S5: Construct a tight oil production capacity prediction model based on machine learning and set the hyperparameters of the tight oil production capacity prediction model;

[0141] Specifically, a deep neural network model is constructed as the tight oil production capacity prediction model, and the hyperparameters of the deep neural network model include: the number of neural network layers, the number of hidden layer neurons, the activation function type, the weight initialization method, and the optimizer settings;

[0142] The deep neural network model is configured with one input layer, four hidden layers, and one output layer, with 50 neurons in each layer. The activation function is set to ReLU, and the network propagation method is as follows:

[0143]

[0144]

[0145] Where f() represents the activation function, i is the index of the hidden layer of the network, j is the index of the hidden layer neuron, z represents the input of the i-th layer, y represents the output of the i-th layer, and w and b represent the weights and biases of the i-th layer, respectively.

[0146] Weights are initialized to the Xavier method:

[0147]

[0148] In the formula, Let w represent any i-th hidden layer. i Let Var[w] be the weight of the i-th layer. i ] indicates w i The variance, n i Enter the number;

[0149] Wherein, the weight w of the i-th layer i Initialize using a Gaussian distribution:

[0150]

[0151] The Adam optimizer is used, and the specific solution is as follows:

[0152]

[0153] m t ←β1m t-1 +(1-β1)g t

[0154] v t ←β2v t-1 +(1-β2)g t 2

[0155]

[0156]

[0157]

[0158] Where t is the time step, g t Let g be the gradient at time step t. t 2 =g t ⊙g t ;θ t-1 Let be the parameter vector at time step t-1. The gradient is represented by (θ), which is the stochastic objective function of parameter θ; α is the step size, which defaults to 0.001; β1, β2 ∈ 0, 1), which represent the exponential decay rate of the moment estimation, with β1 defaulting to 0.9 and β2 defaulting to 0.999. t Let t be the first moment vector at time step t. For the correction of the first-order moment vector estimate at time step t, v t Let t be the second moment vector at time step t. For the correction of the second-order moment vector estimate at time step t, ε is a parameter to prevent the denominator from being zero, and is set to 10. -8 .

[0159] Step S6: Train the tight oil production capacity prediction model using the training set;

[0160] Specifically, the training set is input into the tight oil production capacity prediction model for training. During the training process, the mean square error value is used as the error standard, and the tight oil production capacity prediction model is physically constrained by the multi-stage fracturing production capacity model.

[0161] The MSE value is calculated using the following formula:

[0162]

[0163] In the formula, MSE is the mean square error value, and y i Represents the true value. This represents the predicted value, where N is the sample size.

[0164] The physical constraints applied to the tight oil production prediction model using a multi-stage fracturing production model include:

[0165] The obtained parameter values ​​are substituted into the multi-stage fracturing capacity model for calculation, and the error function is designed as follows:

[0166]

[0167] Among them, phy_MSE design error value, Q represents the data predicted by the multi-stage fracturing capacity model. i This represents real data, where N is the number of samples;

[0168] The loss function is then:

[0169] loss = MSE + phy_MSE

[0170] In the formula, loss is the loss value;

[0171] If the loss value is less than 10 after training is complete. -2 If the loss value is greater than or equal to 10, then the model has completed training.-2 Then, the tight oil production capacity prediction model will continue to be trained.

[0172] Step S7: Predict tight oil production capacity using the trained tight oil production capacity prediction model.

[0173] In this embodiment, after training the tight oil production capacity prediction model using the training set, the method further includes:

[0174] Step S8: The trained tight oil production capacity prediction model is tested using the test set. The production capacity prediction result of the tight oil production capacity prediction model is compared with the actual value to determine whether the prediction accuracy meets the requirements. If it does not meet the requirements, the hyperparameters are modified until the requirements are met.

[0175] Among them, the coefficient of determination R 2 To evaluate the accuracy of the tight oil production capacity prediction model, the calculation method is as follows:

[0176]

[0177] Among them, R 2 As the coefficient of determination, y i Represents the true value. Indicates the predicted value. This represents the average value.

[0178] Preferably, if R 2 If the value reaches 90% or higher, the model is considered to have high prediction accuracy and can effectively predict tight oil production capacity.

[0179] In a specific application scenario of the method in this embodiment, taking a horizontal well in a certain block of Changqing as an example, an orthogonal experiment was designed, and the parameter design is shown in Table 1:

[0180] Table 1. List of parameters for orthogonal experiments

[0181]

[0182] The normalized fracturing and completion parameters, formation parameters and their corresponding tight oil production data are used as production prediction indicators based on the average daily production of the first three months of production. The first 70% of the dataset is divided into the training set and the last 30% into the test set.

[0183] A deep neural network model is constructed for tight oil production capacity prediction. The neural network model is set up with 1 input layer, 4 hidden layers, and 1 output layer. Each hidden layer has 50 neurons, the activation function is set to ReLU, the optimizer is Adam optimizer, and the weights are initialized using the Xavier method. At this point, the tight oil production capacity prediction model has been established.

[0184] The normalized training set is input into the tight oil production capacity prediction model for training. The loss function is set as the sum of MSE and phy_MSE. Training is repeated until the loss function value is less than 10. -2 Ultimately, the model predicts the actual production capacity of the blocks with an accuracy R0. 2 It is 92%, such as Figure 3 As shown.

[0185] In summary, the tight oil reservoir production prediction method of the present invention is based on seepage law and machine learning, which can accurately predict tight oil production capacity. It can accurately predict tight oil production capacity under different fracturing conditions, improve prediction accuracy, and provide a reference for tight oil production capacity prediction.

[0186] Example 2

[0187] An electronic device, the electronic device comprising:

[0188] At least one processor; and,

[0189] A memory communicatively connected to the at least one processor; wherein,

[0190] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the tight reservoir productivity prediction method described in Example 1.

[0191] An electronic device according to an embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions.

[0192] Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0193] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.

[0194] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.

[0195] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0196] Example 3

[0197] Thirdly, the present invention proposes a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the tight oil reservoir production capacity prediction method described in Example 1.

[0198] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.

[0199] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).

[0200] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for predicting the productivity of tight oil reservoirs, characterized in that, include: Based on the characteristics of multi-stage fracturing development in horizontal wells of tight oil reservoirs, a mathematical model for seepage during multi-stage fracturing development in horizontal wells of tight oil reservoirs is established. Based on the mathematical model of seepage development in multi-stage fracturing, a multi-stage fracturing production capacity model for horizontal wells in tight oil reservoirs is established. Based on the multi-stage fracturing production capacity model, tight oil production capacity data under different fracturing completion parameters and formation parameters were calculated through orthogonal experiments. Normalize the fracturing completion parameters, formation parameters and their corresponding tight oil production data, and divide the normalized dataset into training set and test set; A tight oil production capacity prediction model based on machine learning is constructed and hyperparameters are set for the tight oil production capacity prediction model. The tight oil production capacity prediction model is trained using the training set. Predict tight oil production capacity using the trained tight oil production capacity prediction model; Among them, the mathematical model for the seepage development of multi-stage fracturing in horizontal wells of tight oil reservoirs includes: Based on the characteristics of multi-stage fracturing development of horizontal wells in tight oil reservoirs, the flow field structure of horizontal wells in tight oil reservoirs is divided into horizontal wellbore zone, strongly stimulated zone, weakly stimulated zone, matrix activated zone, and unactivated zone. The seepage resistance of the heavily modified zone is calculated using the following formula: In the formula: To enhance the seepage resistance of the modified area, For penetration rate, For fluid viscosity, For reservoir thickness, The width of the crack. For half the length of the crack, Where is the wellbore radius; The seepage resistance of the weakly modified zone is calculated using the following formula: In the formula: For the seepage resistance of the weakly modified area, For fluid viscosity, For the penetration rate in areas with weak transformation, For reservoir thickness, The seepage radius at the boundary between the weakly modified zone and the matrix mobilization zone. The seepage radius at the boundary between the heavily modified zone and the lightly modified zone; The seepage resistance of the matrix mobilization zone is calculated using the following formula: In the formula: For the seepage resistance of the matrix mobilization zone, The permeability of the matrix mobilization zone, For reservoir thickness, The seepage radius at the boundary between the unused and activated areas of the matrix. The seepage radius at the boundary between the matrix zone and the weakly modified zone; The calculation formula for the multi-stage fracturing productivity model of horizontal wells in tight oil reservoirs is as follows: In the formula, For production capacity, To alleviate supply pressure, For the bottom hole pressure, For the equivalent starting pressure gradient of the weakly modified area, The equivalent starting pressure gradient of the matrix mobilization zone. This represents the area of ​​the overlapping zone between adjacent crack segments. Let be the length of the major axis of the elliptical flow field. Let be the length of the minor axis of the elliptical flow field. The seepage radius at the boundary between the unused and activated areas of the matrix. The seepage radius at the boundary between the matrix zone and the weakly modified zone. The seepage radius at the boundary between the heavily modified zone and the lightly modified zone. To enhance the seepage resistance of the modified area, For the seepage resistance of the weakly modified area, The seepage resistance in the matrix mobilization zone.

2. The method for predicting the productivity of tight oil reservoirs according to claim 1, characterized in that, Fracturing completion parameters and formation parameters include: horizontal well length, number of fracture clusters at one end, fracture width, matrix permeability, original formation pressure, bottom hole flowing pressure, extent of the stimulated zone and matrix zone, and rock compressibility coefficient.

3. The method for predicting the productivity of tight oil reservoirs according to claim 1, characterized in that, Normalized fracturing completion parameters, formation parameters, and their corresponding tight oil production data include: The fracturing completion parameters, formation parameters, and their corresponding tight oil production data are normalized using the following formula: in, This is the raw data for fracturing completion parameters, formation parameters, and tight oil production capacity data. , These are the minimum and maximum values ​​of the original dataset, respectively. This is the normalized data.

4. The method for predicting the productivity of tight oil reservoirs according to claim 1, characterized in that, The construction of a machine learning-based tight oil production capacity prediction model and the setting of hyperparameters for the tight oil production capacity prediction model include: A deep neural network model is constructed as the tight oil production capacity prediction model. The hyperparameters of the deep neural network model include: the number of neural network layers, the number of hidden layer neurons, the activation function type, the weight initialization method, and the optimizer settings. The deep neural network model is configured with one input layer, four hidden layers, and one output layer, with 50 neurons in each layer. The activation function is set to ReLU, and the network propagation method is as follows: in, Let represent the activation function, i be the index of the hidden layer, j be the index of the hidden layer neuron, z be the input of the i-th layer, y be the output of the i-th layer, and w and b be the values ​​of the neurons in the i-th layer and the y-th layer, respectively. Layer weights and biases; Weights are initialized to the Xavier method: In the formula, Represents any hidden layer. Let be the weight of the i-th layer. express variance Enter the number; Wherein, the weights of the i-th layer Initialize using a Gaussian distribution: The Adam optimizer is used, and the specific solution is as follows: Where t is the time step, Let be the gradient at time step t. ; for The parameter vector of the time step, Represents the gradient. For parameters random objective function This is the step size, which defaults to 0.

001. The meaning is the exponential decay rate of the moment estimate. The default value is 0.

9. The default setting is 0.

999. Let t be the first moment vector at time step t. Correction for the first-order moment vector estimate at time step t. Let t be the second moment vector at time step t. Correction for the second-order moment vector estimate at time step t. This is a parameter to prevent the denominator from being 0; it is set to 10. -8 .

5. The method for predicting the productivity of tight oil reservoirs according to claim 1, characterized in that, The step of training the tight oil production capacity prediction model using the training set includes: The training set is input into the tight oil production capacity prediction model for training. During the training process, the mean square error value is used as the error standard, and the tight oil production capacity prediction model is physically constrained by the multi-stage fracturing production capacity model. The MSE value is calculated using the following formula: In the formula, This is the mean square error value. Represents the true value. This represents the predicted value, where N is the sample size. The physical constraints applied to the tight oil production prediction model using a multi-stage fracturing production model include: The obtained parameter values ​​are substituted into the multi-stage fracturing capacity model for calculation, and the error function is designed as follows: in, Design error value, This represents data predicted using a multi-stage fracturing capacity model. This represents real data, where N is the number of samples; The loss function is then: In the formula, This is the loss value; If after training, if Value less than 10 -2 If the model completes training, then the training is finished. Value greater than or equal to 10 -2 Then, the tight oil production capacity prediction model will continue to be trained.

6. The method for predicting the productivity of tight oil reservoirs according to claim 4, characterized in that, After training the tight oil production capacity prediction model using the training set, the method further includes: The trained tight oil production capacity prediction model is tested using the test set. The production capacity prediction results of the tight oil production capacity prediction model are compared with the actual values ​​to determine whether the prediction accuracy meets the requirements. If it does not meet the requirements, the hyperparameters are modified until the requirements are met. The prediction accuracy of the tight oil production capacity prediction model is determined by the following formula: in, As the coefficient of determination, Represents the true value. Indicates the predicted value. This represents the average value.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the tight reservoir productivity prediction method according to any one of claims 1-6.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to execute the tight oil reservoir production prediction method according to any one of claims 1-6.

Images