Shale oil and gas production dynamic prediction method based on GCN-GRU

Abstract

The application discloses a shale oil and gas production dynamic prediction method based on GCN-GRU, and relates to the technical field of oil and gas well production prediction.According to the related parameters of each fractured horizontal well in a shale oil and gas well station, the volume reconstruction region spatial relationship and the artificial fracture distribution characteristics are combined, the interwell interference coefficient is calculated, the multi-well adjacency matrix is constructed, the production dynamic data is obtained according to the production of each fractured horizontal well in the shale oil and gas well station, the production dynamic data is distributed to a training set and a test set, the GCN-GRU model is constructed, the production of the fractured horizontal well is predicted by using the production dynamic data in the training set and the multi-well adjacency matrix to train the model, and the accuracy of the model prediction after training is verified by using the production dynamic data in the test set.The interwell interference is quantified, and the high-precision prediction of the shale oil and gas production under the conditions of three-dimensional development and large-scale hydraulic fracturing is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oil and gas well production prediction, in particular to a shale oil and gas production dynamic prediction method based on GCN-GRU, and particularly to a shale oil and gas production dynamic prediction method that fuses the spatial relationship of stimulated reservoir volume (SRV) and the fracture distribution characteristics of a graph convolutional neural network (GCN) and a gate recurrent unit (GRU). BACKGROUND

[0002] With the continuous increase of the development of unconventional oil and gas resources, especially shale oil and gas, reservoir three-dimensional development, horizontal well staged fracturing and volume fracturing engineering technologies have been widely used in low porosity and ultra-low permeability shale reservoirs, which has greatly improved the oil and gas production degree. However, due to the complex multi-scale fracture network developed in the stimulated reservoir volume formed by three-dimensional development and large-scale hydraulic fracturing, the interwell pressure and production interference is significant, and the production dynamic evolution law is extremely complex, which seriously restricts the optimization of development plan and the improvement of recovery efficiency.

[0003] At present, the shale oil and gas production prediction methods mainly fall into four categories: one is the decline analysis method, such as Arps model and its variants, which has the advantages of few parameters and convenient calculation, but it is difficult to accurately characterize the influence of reservoir parameter changes on production; two is the analytical method, which can obtain analytical solution under ideal boundary and simplified assumption, and has both calculation efficiency and physical meaning, but the application of analytical method in shale reservoirs faces significant challenges, mainly limited by the strong heterogeneity of the reservoir, multi-scale flow mechanism and complex physical and chemical processes; three is the numerical simulation method, which describes the unsteady seepage process by numerically solving the seepage differential equation set, and is suitable for complex system analysis of multi-physical field coupling problems, but has problems such as complicated modeling, parameter sensitivity and high calculation cost; four is the machine learning method, which is gradually becoming an important means of shale oil and gas production prediction due to its strong non-linear modeling ability and good adaptability to unstructured data.

[0004] In addition, the existing production prediction and interference analysis methods generally have difficulty in systematically considering the artificial fracture distribution characteristics, the spatial relationship between multiple wells and the constraints of physical laws, resulting in deficiencies in model generalization and prediction accuracy. Under the mode of three-dimensional development of shale oil and gas, the interwell pressure and production interference mechanism is extremely complex, and it is difficult to quantitatively characterize the interference of fracture network by using traditional shale oil and gas production prediction methods.

[0005] Therefore, there is an urgent need to propose a shale oil and gas production dynamic prediction method based on GCN-GRU, which can not only reflect the spatial relationship of multiple wells and the characteristics of artificial fracture distribution by integrating the physical mechanism, but also efficiently process nonlinear and dynamic characteristics in the shale oil and gas production prediction process. SUMMARY

[0006] The present application aims to solve the above problems and provides a shale oil and gas production dynamic prediction method based on GCN-GRU, which predicts shale oil and gas production by integrating graph convolutional neural network, gated recurrent neural network and volume reconstruction area spatial relationship and artificial fracture distribution characteristics, quantifies interwell interference, and realizes high-precision prediction of shale oil and gas production under complex fracture network.

[0007] To achieve the above purpose, the present application adopts the following technical solutions:

[0008] The shale oil and gas production dynamic prediction method based on GCN-GRU integrates the volume reconstruction area spatial relationship and artificial fracture distribution characteristics of shale reservoirs, including the following steps:

[0009] Step 1, select a shale oil and gas well platform, obtain the basic geological parameters and engineering parameters of each fractured horizontal well in the shale oil and gas well platform, calculate the interwell interference coefficient according to the volume reconstruction area spatial relationship and fracture distribution characteristics of shale reservoirs;

[0010] Step 2, calculate the interference coefficient between each fractured horizontal well in the shale oil and gas well platform, and construct a multi-well adjacency matrix;

[0011] Step 3, according to the production of each fractured horizontal well in the shale oil and gas well platform, obtain the production dynamic data, distribute the production dynamic data to the training set and the test set, and construct a GCN-GRU model at the same time;

[0012] Step 4, input the production dynamic data and multi-well adjacency matrix in the training set into the GCN-GRU model, train the GCN-GRU model, use the GCN model in the GCN-GRU model to extract spatial features and input them into the GRU model to extract time features, train the GCN-GRU model to predict the production of fractured horizontal wells, and obtain the trained GCN-GRU model;

[0013] Step 5, input the production dynamic data in the test set into the trained GCN-GRU model for prediction, and use the production dynamic data in the test set to verify the accuracy of the trained GCN-GRU model prediction.

[0014] Preferably, the interwell interference coefficient calculation process is:

[0015] The first crack of the first fracturing horizontal well in the shale reservoir is connected with the wellbore, and the connected position is taken as the coordinate origin. A Cartesian coordinate system is established with the horizontal well direction as the x axis, the crack extension direction as the y axis, and the vertical reservoir direction as the z axis. The grid is divided in the shale reservoir, and the coordinates of the center position of the first crack of the first fracturing horizontal well are x1, y1, and z1. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2. The coordinates of the center position of the first crack of the second fracturing horizontal well are x2, y2, and z2.

[0016]

[0017] In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate.

[0018] In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate.

[0019] In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. , the plane in which the volume of the second fracturing horizontal well is located is expressed as , the plane is expressed as , the plane expression is , the spatial straight line intersects the plane at , and the plane at ;

[0020] In order to simplify the calculation, the coordinates of the point on the spatial straight line are simplified as , and the coordinates of the point are simplified as , wherein is the simplified horizontal coordinate of the point , , is the simplified vertical coordinate of the point , , is the simplified horizontal coordinate of the point , , is the simplified vertical coordinate of the point , ; when the SRV regions of the two fracturing horizontal wells do not intersect, the coordinates of the intersection point are , and the coordinates of the intersection point are ; when the SRV regions of the two fracturing horizontal wells intersect, the coordinates of the intersection point are , and the coordinates of the intersection point are .

[0021] Preferably, the interference coefficient of the first fracturing horizontal well to the second fracturing horizontal well is:

[0022] ;

[0023] wherein

[0024] ;

[0025] ;

[0026] In the formula, the total number of cracks in the first fracturing horizontal well is ; is the total number of cracks in the second fracturing horizontal well; The first fractured horizontal well The fracture in the No. 2 fracturing horizontal well The interference coefficient of the crack; The second horizontal well with fracturing Half the length of a crack; The first fractured horizontal well The fracture in the No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment; The first horizontal well in the fracturing process Half the length of a crack; The first horizontal well in the fracturing process Crack No. Section and No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment;

[0027] When the SRV regions of two fractured horizontal wells do not intersect, the first fractured horizontal well... Crack No. Section and No. 2 fracturing horizontal well Crack No. Interference coefficient of segment The calculation formula is:

[0028] ;

[0029] in,

[0030] ;

[0031] ;

[0032] ;

[0033] In the formula, For point With point The distance between them; For point Intersection point The distance; For point Intersection point The distance; Intersection point Intersection point The distance; The permeability of the volumetric stimulation zone of the No. 1 fracturing horizontal well. The permeability of the volumetric stimulation zone of the No. 2 fracturing horizontal well; For matrix permeability;

[0034] When the SRV regions of two fractured horizontal wells intersect, the first fractured horizontal well... Crack No. Section and No. 2 fracturing horizontal well Crack No. Interference coefficient of segment The calculation formula is:

[0035] ;

[0036] in,

[0037] ;

[0038] ;

[0039] ;

[0040] In the formula, The permeability of the SRV intersection region.

[0041] Preferably, the GCN-GRU model comprises a GCN model and a GRU model connected in sequence;

[0042] The GCN model is used to aggregate the features of surrounding nodes based on the multi-well adjacency matrix and feature matrix to update its own node. The GCN model incorporates a graph convolutional network, and the propagation pattern between the structural layers within the graph convolutional network is as follows:

[0043] ;

[0044] In the formula, For layer number, For the first in a graph convolutional network The feature input of the layer, For the first in a graph convolutional network Layer feature input; It is the sigmoid function; It is a degree matrix; It is a self-connected adjacency matrix. , It is an adjacency matrix; It is the identity matrix;

[0045] The GRU model is built on a Long Short-Term Memory neural network and is used to extract the temporal features of the GCN model output features, including the update gate. and reset door The GRU model integrates the forget gate and input gate of the Long Short-Term Memory Neural Network into an update gate, and uses a gating mechanism to model time series data, in order to capture long-term dependencies and alleviate the gradient vanishing problem.

[0046] Preferably, when predicting the production of the shale reservoir at the time point t, the GRU model first inputs the production of the shale reservoir at the time point t and the hidden state of the GRU model at the previous time point t-1 to obtain the gating signals of the update gate and the reset gate, calculates the update gate and the reset gate, and compresses the update gate and the reset gate to the interval [0, 1] by using a Sigmoid function, wherein the reset gate is used to control the influence of the previous state on the current candidate state, if the reset gate is close to 0, the historical information is ignored, and the candidate state is generated based on the current input only; and the update gate is used to dynamically balance the proportion of the old state and the new candidate state, to determine the reserved part and the absorbed part, and output the hidden state at the time point t.

[0047] The calculation formula of the GRU model is as follows:

[0048]

[0049] In the formula, r represents the reset gate, u represents the update gate, x represents the input of the GRU model at the time point t, h represents the output of the GRU model at the time point t, h' represents the updated cell state, W r represents the weight matrix of the reset gate, W u represents the weight matrix of the update gate, W c represents the weight matrix of the current memory content, b r represents the bias matrix of the reset gate, b u represents the bias matrix of the update gate, b c represents the bias matrix of the current memory content, and tanh represents the hyperbolic tangent function.

[0050] ​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Preferably, in step 4, the production dynamic data in the test set and the multi-well adjacency matrix constructed by the interference coefficients between the wells are input into the GCN-GRU model, the GCN-GRU model is trained to predict the production of the fractured horizontal well at the next time according to the interference coefficients between the fractured horizontal wells and the production of the fractured horizontal well at the current time, and the prediction result of the GCN-GRU model on the production of the fractured horizontal well is compared with the original production data of the fractured horizontal well in the production dynamic data , the mean absolute error , the prediction accuracy , and the average relative error When the root mean square error , the mean absolute error , the prediction accuracy , and the average relative error of the GCN-GRU model reach the preset accuracy value, the training of the GCN-GRU model is stopped, and the trained GCN-GRU model is obtained.

[0051] The trained GCN-GRU model can predict the production at a specified future time according to the production of the fractured horizontal well at the current time, and obtain the predicted value of the production of each fractured horizontal well in the shale oil and gas well platform.

[0052] Preferably, the root mean square error is calculated according to the following formula:

[0053]

[0054] The mean absolute error is calculated according to the following formula:

[0055]

[0056] The prediction accuracy is calculated according to the following formula:

[0057]

[0058] The average relative error is calculated according to the following formula:

[0059]

[0060] In the formula, is the actual production of the fractured horizontal well; is the predicted value of the production of the fractured horizontal well output by the GCN-GRU model; is the serial number of the production data of the fractured horizontal well; is the total number of the production data of the fractured horizontal well.​​​​

[0061] The beneficial technical effects brought by the present application are:

[0062] The present application proposes a shale oil and gas production dynamic prediction method based on GCN-GRU, which fuses a graph convolutional neural network, a gated recurrent neural network, and volume reconstruction region spatial relationship and artificial fracture distribution characteristics to construct a GCN-GRU model with physical constraints, which not only integrates the interference relationship between multiple wells, but also can efficiently process nonlinear and dynamic characteristics, quantizes the interwell interference, fully considers the volume reconstruction region spatial relationship and artificial fracture distribution characteristics, considers the interwell interference of each fracturing horizontal well, and realizes high-precision prediction of shale oil and gas production, laying a foundation for intelligent development of unconventional oil and gas reservoirs. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 The flowchart of the shale oil and gas production dynamic prediction method based on GCN-GRU of the present application.

[0064] Figure 2 The schematic diagram of multiple fracturing horizontal wells in a shale oil and gas well station.

[0065] Figure 3 The schematic diagram of two fracturing horizontal well volume reconstruction regions not intersecting.

[0066] Figure 4 The schematic diagram of two fracturing horizontal well volume reconstruction regions intersecting.

[0067] Figure 5 The schematic diagram of shale oil and gas well station production dynamic data.

[0068] Figure 6 The comparison diagram of prediction results of GCN-GRU models trained by different methods.

[0069] In the figure, W1 is a first fracturing horizontal well, W2 is a second fracturing horizontal well, W3 is a third fracturing horizontal well, W4 is a fourth fracturing horizontal well, W5 is a fifth fracturing horizontal well, and W6 is a sixth fracturing horizontal well. DETAILED DESCRIPTION

[0070] The present application will be further described in detail below in combination with the drawings and examples.

[0071] Taking a shale oil and gas well station in a shale reservoir research area as an example, a shale oil and gas production dynamic prediction method based on GCN-GRU and fusing shale reservoir volume reconstruction region spatial relationship and artificial fracture distribution characteristics is proposed, as shown in Figure 1 The method comprises the following steps:

[0072] Step 1, select a shale oil and gas well platform, the shale oil and gas well platform is provided with a plurality of fracturing horizontal wells, as shown in Figure 2 In the embodiment, the shale oil and gas well platform is provided with six fracturing horizontal wells, namely No. 1 fracturing horizontal well W1, No. 2 fracturing horizontal well W2, No. 3 fracturing horizontal well W3, No. 4 fracturing horizontal well W4, No. 5 fracturing horizontal well W5 and No. 6 fracturing horizontal well W6. The basic geological parameters of each fracturing horizontal well in the shale oil and gas well platform and the engineering parameters of the fracturing horizontal well are obtained. The basic geological parameters of the fracturing horizontal well include well location and unmodified area permeability. The engineering parameters of the fracturing horizontal well include the included angle between the fracture and the horizontal well, the fracture half length, the first section perforation distance of the fracturing horizontal well, the volume modified area permeability, the fracturing fracture spacing, the well fracturing cluster number, the fracture distribution, and the fracture opening, as shown in Table 1.

[0073] Table 1 Engineering parameter table of fracturing horizontal well

[0074] .

[0075] According to the volume modified area and the spatial distribution of the fracture network of the shale reservoir, the interwell interference coefficient is calculated. The calculation process of the interwell interference coefficient is as follows:

[0076] Taking the connection position of the first fracture of the No. 1 fracturing horizontal well in the shale reservoir with the wellbore as the coordinate origin, establishing a Cartesian coordinate system with the horizontal well direction as the x-axis, the fracture extension direction as the y-axis, and the vertical reservoir direction as the z-axis, discretizing the fracture into a grid with a length of one meter, a width of the fracture width, and a height of the fracture height, each fracture occupies grid, the coordinates of the center position of the segment of the th fracture in the No. 1 fracturing horizontal well are , the coordinates of the connection position of the first fracture of the No. 2 fracturing horizontal well with the wellbore are , the coordinates of the center position of the segment of the th fracture in the No. 2 fracturing horizontal well are , the point of the spatial straight line passing through the two center positions is

[0077] ;

[0078] In the formula, x is the horizontal coordinate, y is the vertical coordinate, and z is the vertical coordinate. , are the fracture numbers. , are the segment numbers. ​​​The fracture spacing of the No. 1 fracturing horizontal well. The fracture spacing of the No. 2 fracturing horizontal well; The first horizontal well in the fracturing process The angle between the fracture and the No. 1 fractured horizontal well. The second horizontal well with fracturing The angle between the fracture and the No. 2 fractured horizontal well; This refers to the perforation distance of the No. 1 fracturing horizontal well. This refers to the perforation distance of the No. 2 fracturing horizontal well; The well spacing is along the x-axis. The well spacing is along the y-axis. The well spacing is in the z-axis direction.

[0079] Furthermore, the volumetric stimulation zones of fractured horizontal well No. 1 and fractured horizontal well No. 2 were determined, and the half-length of the volumetric stimulation zone of fractured horizontal well No. 1 was obtained. and half the length of the volumetric stimulation zone of the No. 2 fracturing horizontal well Where the half-length of the volume modification region is half the length of the volume modification region; when the volume modification regions of the two do not intersect, such as Figure 3 As shown, at this time When the volume modification regions of the two intersect, such as Figure 4 As shown, at this time .

[0080] The plane in which the volumetric stimulation zone of the No. 1 fracturing horizontal well is located is represented as follows: The plane representing the volumetric stimulation zone of the No. 2 fracturing horizontal well is shown as follows: ,flat The expression is Plane expression for spatial straight line With plane The intersection point is , and plane The intersection point is .

[0081] To simplify calculations, use spatial straight lines On point The coordinates are simplified to ,point The coordinates are simplified to ,in, For point Simplified x-axis, , For point Simplified ordinate, , For point Simplified x-axis, , For point Simplified ordinate, When the SRV regions of two fractured horizontal wells do not intersect, the intersection point The coordinates are Intersection point The coordinates are When the SRV regions of two fractured horizontal wells intersect, the intersection point The coordinates are Intersection point The coordinates are .

[0082] Furthermore, the interference coefficient of the No. 1 fractured horizontal well to the No. 2 fractured horizontal well... for:

[0083] ;

[0084] in,

[0085] ;

[0086] ;

[0087] In the formula, This represents the total number of fractures in the No. 1 fractured horizontal well. This represents the total number of fractures in the No. 2 fracturing horizontal well. The first horizontal well in the fracturing process The fracture in the No. 2 fracturing horizontal well The interference coefficient of the crack; The second horizontal well with fracturing Half the length of a crack; The first fractured horizontal well The fracture in the No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment; The first horizontal well in the fracturing process Half the length of a crack; The first horizontal well in the fracturing process Crack No. Section and No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment.

[0088] When the SRV regions of two fractured horizontal wells do not intersect, the first fractured horizontal well... Crack No. Section and No. 2 fracturing horizontal well The interference coefficient of the first fracture in the first section The interference coefficient of the first fracture in the first section The calculation formula is:

[0089] ;

[0090] Wherein,

[0091] ;

[0092] ;

[0093] ;

[0094] In the formula, The distance between the point And the intersection point ; The distance between the point And the intersection point ; The distance between the point And the intersection point ; The distance between the intersection point And the intersection point ; The permeability of the first fractured horizontal well volume reconstruction area, The permeability of the second fractured horizontal well volume reconstruction area; The matrix permeability.

[0095] When the SRV areas of the two fractured horizontal wells intersect, the interference coefficient of the first fracture in the first section of the first fractured horizontal well and the first fracture in the first section of the second fractured horizontal well is The interference coefficient of the first fracture in the first section The interference coefficient of the first fracture in the first section The interference coefficient of the first fracture in the first section The calculation formula is:

[0096] ; Wherein,

[0097]

[0098] ;

[0099] ;

[0100] ; In the formula,

[0101] The permeability of the SRV intersection area.

[0102] ​Step 2: Calculate the interference coefficient between each fractured horizontal well in the shale oil and gas well platform, and construct a multi-well adjacency matrix, which is expressed as:

[0103] ;

[0104] In the formula, The first shale oil and gas well platform Fracturing horizontal well for the first Fracturing horizontal wells Interference coefficient of fractured horizontal wells.

[0105] The multi-well adjacency matrix constructed in this embodiment is as follows:

[0106] .

[0107] To highlight the superiority of the method of this invention, a 0-1 matrix is ​​constructed using common methods, resulting in:

[0108] .

[0109] Step 3: Based on the production output of each fractured horizontal well in the shale oil and gas well platform, obtain dynamic production data, such as... Figure 5 As shown, in this embodiment, the ratio of the amount of dynamic data generated in the training set to the amount of data generated in the test set is 8:2, and a GCN-GRU model is constructed.

[0110] In this embodiment, the GCN-GRU model includes a GCN model and a GRU model connected in sequence.

[0111] The GCN model consists of an input layer, hidden layers, and an output layer. The hidden layer is the core of the GCN model, and it includes multiple graph convolutional layers. Each graph convolutional layer operates by applying an adjacency matrix... With characteristic matrix Convolutional operations are performed to update the node feature representations. The propagation pattern between the structural layers within the graph convolutional network is as follows:

[0112] ;

[0113] In the formula, For layer number, For the first in a graph convolutional network The feature input of the layer, For the first in a graph convolutional network Layer feature input; It is the sigmoid function; It is a degree matrix; It is a self-connected adjacency matrix. , It is an adjacency matrix; It is an identity matrix.

[0114] The GRU model is based on a Long Short-Term Memory (LSTM) neural network, a variant of LSTM, used to extract temporal features from the output features of the GCN model, including the update gate. and reset door This approach solves the vanishing gradient problem while preserving time-series characteristics. The GRU model improves training speed by integrating the forget gate and input gate of the Long Short-Term Memory (LSTM) neural network into an update gate, and uses a gating mechanism to model time-series data, effectively capturing long-term dependencies and mitigating the vanishing gradient problem.

[0115] When using the GRU model for prediction When calculating the production of a shale reservoir, the GRU model first inputs the current shale reservoir production data. The hidden state of the GRU model at the previous time step Get Updates and reset door The gate control signal is used to calculate and update the gate. and reset door And it is compressed to the [0,1] interval using the Sigmoid function, wherein the reset gate Used to control the previous state for the current candidate state. The impact of resetting the door If the value is close to 0, historical information is ignored, and candidate states are generated only based on the current input; the update gate The ratio of old states to new candidate states used to dynamically balance the states determines the portion to be retained. and absorption portion and output the hidden state at the current moment. .

[0116] Furthermore, the calculation formula for the GRU model is as follows:

[0117] ;

[0118] In the formula, To reset the door, To update the door; for The input to the GRU model at time step; for The output of the GRU model at time 10:00 for The output of the GRU model at time 10:00 The updated cell state; To reset the weight matrix of the gate, To update the gate weight matrix, is a weight matrix of the current memory content; is a bias matrix of the reset gate, is a bias matrix of the update gate, is a bias matrix of the current memory content; is a Hadamard product operator; is a hyperbolic tangent function.

[0119] Step 4, input the production dynamic data in the training set and the multi-well adjacency matrix into the GCN-GRU model, train the GCN-GRU model, input the spatial features extracted by the GCN model in the GCN-GRU model into the GRU model to extract the time features, train the GCN-GRU model to predict the production of the fractured horizontal well, and obtain the trained GCN-GRU model.

[0120] In this embodiment, the production dynamic data in the test set and the multi-well adjacency matrix constructed by the interference coefficient between each well are input into the GCN-GRU model, the GCN-GRU model is trained to predict the production of the fractured horizontal well at the next time according to the adjacency matrix and the production of the fractured horizontal well at the current time, and the prediction result of the GCN-GRU model on the production of the fractured horizontal well is compared with the original production data of the fractured horizontal well in the production dynamic data. The root mean square error of the GCN-GRU model is calculated each time , the mean absolute error , the prediction accuracy and the average relative error of the GCN-GRU model, the calculation formula of the root mean square error is:

[0121] ;

[0122] The calculation formula of the mean absolute error is:

[0123] ;

[0124] The calculation formula of the prediction accuracy is:

[0125] ;

[0126] The calculation formula of the average relative error is:

[0127] ;

[0128] In the formula, is the actual production of the fractured horizontal well; is the predicted value of the production of the fractured horizontal well output by the GCN-GRU model. is the serial number of the fractured horizontal well production data; is the total number of the fractured horizontal well production data.

[0129] the root mean square error of the GCN-GRU model , the mean absolute error , the prediction accuracy , and the average relative error , the smaller the values of the root mean square error , the mean absolute error , and the average relative error , and the closer the value of the prediction accuracy to 1, the better the prediction effect of the GCN-GRU model.

[0130] In this embodiment, when the root mean square error , the mean absolute error , the prediction accuracy , and the average relative error of the GCN-GRU model reach the preset precision value, the training of the GCN-GRU model is stopped, and the trained GCN-GRU model is obtained.

[0131] The trained GCN-GRU model can predict the production of the future specified time according to the production of the current time of the fractured horizontal well, and obtain the predicted value of the production of each fractured horizontal well in the shale oil and gas well platform.

[0132] Step 5: input the production dynamic data in the test set into the trained GCN-GRU model for prediction, and verify the accuracy of the prediction of the trained GCN-GRU model by using the production dynamic data in the test set.

[0133] Further, the root mean square error , the mean absolute error , the prediction accuracy , and the average relative error of the GCN-GRU models obtained by using the conventional method and the method of the present application are compared, the conventional method uses a 0-1 matrix to train the GCN-GRU model, and the method of the present application uses a multi-well adjacency matrix fused with the volume reconstruction region of the shale reservoir to train the GCN-GRU model, as shown in Table 2.

[0134] Table 2 Comparison table of evaluation indexes of two methods

[0135] .

[0136] The prediction results of the GCN-GRU model trained by using the conventional method and the method of the present application are compared, and it is found that the prediction error of the GCN-GRU model trained by using the method of the present application is smaller than that of the GCN-GRU model trained by using the conventional method, that is, the method of the present application quantitatively characterizes the interference between the fractured horizontal wells by fusing the shale reservoir volume reconstruction region spatial relationship and the artificial fracture distribution characteristics, and realizes the accurate prediction of the shale oil and gas production, which is beneficial to the optimization of the shale reservoir development plan and the improvement of the recovery rate. Figure 6

[0137] Of course, the above description is not a limitation of the present application, and the present application is not limited to the above examples. Changes, modifications, additions or replacements made by those skilled in the art within the essential scope of the present application should also be within the protection scope of the present application.​

Claims

1. A shale oil and gas production dynamic prediction method based on GCN-GRU, characterized in that, Fusing the volume reconstruction region space relationship of shale reservoir and the fracture distribution characteristics of artificial fractures, comprising the following steps: Step 1, selecting a shale oil and gas well station, obtaining the basic geological parameters and engineering parameters of each fractured horizontal well in the shale oil and gas well station, and calculating the interwell interference coefficient according to the volume reconstruction region space relationship of shale reservoir and the fracture distribution characteristics; Step 2, calculate the interference coefficient between each fractured horizontal well in the shale oil and gas well station, and construct a multi-well adjacency matrix; Step 3, according to the production of each fractured horizontal well in the shale oil and gas well station, obtain the production dynamic data, distribute the production dynamic data to the training set and the test set, and construct a GCN-GRU model; Step 4, input the production dynamic data and multi-well adjacency matrix in the training set into the GCN-GRU model, train the GCN-GRU model, use the GCN model in the GCN-GRU model to extract spatial features and input them into the GRU model to extract time features, train the GCN-GRU model to predict the production of the fractured horizontal well, and obtain the trained GCN-GRU model; Step 5, input the production dynamic data in the test set into the trained GCN-GRU model for prediction, and verify the accuracy of the trained GCN-GRU model prediction by using the production dynamic data in the test set; The interwell interference coefficient calculation process is: Using the point where the first fracture connects to the wellbore in the No. 1 fractured horizontal well in the shale reservoir as the origin, a Cartesian coordinate system is established with the horizontal well direction as the x-axis, the fracture extension direction as the y-axis, and the vertical reservoir direction as the z-axis. A grid is then drawn within the shale reservoir to obtain the first fracture in the No. 1 fractured horizontal well. Crack No. The coordinates of the center position of the segment are The coordinates of the connection point between the first fracture and the wellbore in the No. 2 fracturing horizontal well are: The second horizontal well with fracturing Crack No. The coordinates of the center position of the segment are Then the spatial straight line passing through the two center positions The point-directed form is: ； In the formula, is the horizontal coordinate, is the vertical coordinate, is the vertical coordinate; , are both fracture numbers; , are both segment numbers; is the fracture spacing of the first fractured horizontal well, is the fracture spacing of the second fractured horizontal well; is the angle between the first fracture and the first fractured horizontal well, is the angle between the first fracture and the first fractured horizontal well, is the angle between the first fracture and the first fractured horizontal well; is the angle between the first fracture and the first fractured horizontal well; is the perforation distance of the first fractured horizontal well, is the perforation distance of the second fractured horizontal well; is the well spacing in the x-axis direction; is the well spacing in the y-axis direction; is the well spacing in the z-axis direction; Determine the volume reconstruction region of the first fractured horizontal well and the second fractured horizontal well, obtain the half length of the volume reconstruction region of the first fractured horizontal well and the half length of the volume reconstruction region of the second fractured horizontal well When the volume reconstruction regions of the two are not intersected, When the volume reconstruction regions of the two are intersected, ; The plane in which the volumetric stimulation zone of the No. 1 fracturing horizontal well is located is represented as follows: The plane representing the volumetric stimulation zone of the No. 2 fracturing horizontal well is shown as follows: ,flat The expression is Plane expression for spatial straight line With plane The intersection point is , and plane The intersection point is ; For simplifying calculation, the coordinates of the straight line in space are simplified as the coordinates of the point are simplified as the coordinates of the point , wherein is the simplified horizontal coordinate of the point , , is the simplified vertical coordinate of the point , , is the simplified horizontal coordinate of the point , , is the simplified vertical coordinate of the point , ; when the SRV areas of the two fractured horizontal wells do not intersect, the coordinates of the intersection point are , and the coordinates of the intersection point are ; when the SRV areas of the two fractured horizontal wells intersect, the coordinates of the intersection point are , and the coordinates of the intersection point are ; The interference coefficient of the first fracturing horizontal well to the second fracturing horizontal well is: ； Wherein, ； ； In the formula, This represents the total number of fractures in the No. 1 fractured horizontal well. This represents the total number of fractures in the No. 2 fracturing horizontal well. The first horizontal well in the fracturing process The fracture in the No. 2 fracturing horizontal well The interference coefficient of the crack; The second horizontal well with fracturing Half the length of a crack; The first horizontal well in the fracturing process The fracture in the No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment; The first horizontal well in the fracturing process Half the length of a crack; The first horizontal well in the fracturing process Crack No. Section and No. 2 fracturing horizontal well Crack No. The interference coefficient of the segment; When the SRV regions of two fractured horizontal wells do not intersect, the first fractured horizontal well... Crack No. Section and No. 2 fracturing horizontal well Crack No. Interference coefficient of segment The calculation formula is: ； Wherein, ； ； ； wherein is the distance between point and point ; is the distance between point and intersection point ; is the distance between point and intersection point ; is the distance between intersection point and intersection point ; is the permeability of the first fractured horizontal well volume reformation region, is the permeability of the second fractured horizontal well volume reformation region; is the matrix permeability; When the SRV regions of two fractured horizontal wells intersect, the first fractured horizontal well... Crack No. Section and No. 2 fracturing horizontal well Crack No. Interference coefficient of segment The calculation formula is: ； Wherein, ； ； ； In the formula, is the permeability of the SRV intersection area.

2. The GCN-GRU-based shale oil and gas production dynamic prediction method of claim 1, wherein, The GCN-GRU model comprises a GCN model and a GRU model connected in sequence; The GCN model is used to update its node by aggregating the features of surrounding nodes according to the multi-well adjacency matrix and the feature matrix, and a graph convolution network is arranged in the GCN model, and the propagation mode between each structure layer in the graph convolution network is: ； In the formula, is a layer number, is a feature input of a layer in the graph convolution network, is a feature input of a layer in the graph convolution network; is a sigmoid function; is a degree matrix; is a self-connected adjacency matrix, , is an adjacency matrix; is an identity matrix; The GRU model is constructed based on a long short-term memory neural network, and is used for extracting time features of GCN model output features, including an update gate and a reset gate The GRU model integrates the forgetting gate and the input gate of the long short-term memory neural network into the update gate, and uses a gating mechanism to model the time series data, so as to capture long-term dependencies and relieve the gradient vanishing problem.

3. The GCN-GRU based shale oil and gas production dynamic prediction method of claim 2, wherein, When using the GRU model for prediction When calculating the production of a shale reservoir, the GRU model first inputs the current shale reservoir production data. The hidden state of the GRU model at the previous time step Get Updates and reset door The gate control signal is used to calculate and update the gate. and reset door And it is compressed to the [0,1] interval using the Sigmoid function, wherein the reset gate Used to control the previous state for the current candidate state. The impact of resetting the door If the value is close to 0, historical information is ignored, and candidate states are generated only based on the current input; the update gate The ratio of old states to new candidate states used to dynamically balance the states determines the portion to be retained. and absorption portion and output the hidden state at the current moment. ; The calculation formula of the GRU model is: ； wherein, is a reset gate, is an update gate; is input to the GRU model at time t; is output of the GRU model at time t, is output of the GRU model at time t, is the updated cell state; is a weight matrix for the reset gate, is a weight matrix for the update gate, is a weight matrix for the current memory content; is a bias matrix for the reset gate, is a bias matrix for the update gate, is a bias matrix for the current memory content; is a Hadamard product operator; is a hyperbolic tangent function.

4. The GCN-GRU-based shale oil and gas production dynamic prediction method of claim 3, wherein, In step 4, the production dynamic data in the test set and the multi-well adjacency matrix constructed by the interference coefficients between the wells are input into the GCN-GRU model, the GCN-GRU model is trained to predict the production of the fractured horizontal well at the next moment according to the interference coefficients between the fractured horizontal wells and the production of the fractured horizontal well at the current moment, and the prediction result of the GCN-GRU model on the production of the fractured horizontal well is compared with the original production data of the fractured horizontal well in the production dynamic data , the mean absolute error , the prediction accuracy , and the average relative error , when the root mean square error , the mean absolute error , the prediction accuracy , and the average relative error of the GCN-GRU model reach the preset precision value, the training of the GCN-GRU model is stopped, and the trained GCN-GRU model is obtained; The trained GCN-GRU model can predict the production at a specified time according to the current production of the fractured horizontal well, and obtain the predicted value of the production of each fractured horizontal well in the shale oil and gas well station.

5. The GCN-GRU-based shale oil and gas production dynamic prediction method according to claim 4, characterized in that, the root mean square error The calculation formula is: ； the average absolute error The calculation formula is: ； The prediction accuracy The calculation formula is: ； the average relative error The calculation formula is: ； In the formula, is the actual production of the fractured horizontal well; is the predicted value of the fractured horizontal well production output by the GCN-GRU model; is the serial number of the fractured horizontal well production data; is the total number of the fractured horizontal well production data.

Images