A method, medium and system for predicting productivity of horizontal wells in tight oil reservoirs

Abstract

This invention provides a method, medium, and system for predicting the productivity of horizontal wells in tight oil reservoirs, belonging to the field of petroleum productivity prediction technology. The method includes: collecting data affecting the productivity of horizontal wells in tight oil reservoirs and historical daily productivity data of the corresponding wells; preprocessing and dimensionality reduction of the data affecting the productivity of horizontal wells and the historical daily productivity data of the corresponding wells to obtain a training dataset; constructing a productivity prediction model for horizontal wells in tight oil reservoirs based on a deep learning neural network, the model including an input module, a spatiotemporal feature extraction module, a feature fusion module, and a regression prediction module; training the productivity prediction model using the training dataset; inputting the data to be predicted into the productivity prediction model to obtain the productivity prediction result; and deploying and adjusting oil extraction projects based on the obtained productivity prediction result.

Description

Technical Field

[0001] This invention belongs to the field of petroleum production capacity prediction technology, and specifically relates to a method, medium and system for predicting the production capacity of horizontal wells in tight oil reservoirs. Background Technology

[0002] As an important successor to conventional oil and gas, my country has abundant tight oil resources with great development potential. Accurately predicting tight oil production capacity is key to the effective development and management of tight oil resources, but the strong heterogeneity of tight oil reservoirs makes accurate production capacity prediction extremely difficult.

[0003] Traditional production capacity prediction methods are divided into static and dynamic methods. Static production capacity prediction methods mainly rely on well logging information, which primarily reflects the static characteristics of the reservoir and cannot directly reflect the impact of the reservoir's dynamic characteristics on production capacity. Dynamic production capacity prediction models include analytical methods, semi-analytical methods, and numerical simulation methods. However, dynamic methods are derived based on the assumption of formation homogeneity, and are not accurate enough for tight oil reservoirs with strong heterogeneity.

[0004] Data mining-based production capacity prediction research can be broadly categorized into two types: traditional machine learning and deep learning. Traditional machine learning struggles to model complex nonlinear relationships and extract complex features from input data. While deep learning can extract these features, most existing deep learning-based production capacity prediction methods focus on extracting temporal features from the input data to predict well production capacity, neglecting the impact of the strong heterogeneity of tight oil reservoirs on production capacity prediction. This results in inaccurate production capacity predictions for tight oil reservoirs due to the failure to simultaneously extract the spatiotemporal characteristics of the production capacity data. Summary of the Invention

[0005] In view of this, the present invention provides a method, medium and system for predicting the productivity of horizontal wells in tight oil reservoirs, which solves the problem of inaccurate productivity prediction in tight oil reservoirs caused by the spatiotemporal characteristics of not simultaneously extracting productivity data.

[0006] This invention is implemented as follows:

[0007] The first aspect of this invention provides a method for predicting the productivity of horizontal wells in tight oil reservoirs, comprising the following steps:

[0008] S10: Collect data on the production capacity of tight oil horizontal wells and the historical daily production capacity data of the corresponding wells. The data on the production capacity of tight oil horizontal wells include: fracturing stage, sand content per cluster, number of fracturing clusters, fracture spacing, fluid content per cluster, sand content per segment, porosity, permeability, oil saturation, Poisson's ratio, Young's modulus, brittleness index, horizontal segment length, oil layer penetration rate, thickness of penetrated oil layer, and penetration rate of Class I oil layer.

[0009] S20: Preprocess and reduce the dimensionality of the data affecting the productivity of tight oil horizontal wells and the historical daily productivity data of the corresponding wells to obtain the training dataset;

[0010] S30: Construct a production capacity prediction model for horizontal wells in tight oil reservoirs based on deep learning neural networks. The production capacity prediction model for horizontal wells in tight oil reservoirs includes an input module, a spatiotemporal feature extraction module, a feature fusion module, and a regression prediction module.

[0011] S40: Train the production capacity prediction model for the horizontal well in the tight oil reservoir using the training dataset;

[0012] S50: Input the data to be predicted into the production capacity prediction model of the horizontal well in the tight oil reservoir to obtain the production capacity prediction results;

[0013] S60: Deploy and adjust oil extraction projects based on the obtained production capacity forecast results.

[0014] Based on the above technical solution, the method for predicting the productivity of horizontal wells in tight oil reservoirs of the present invention can be further improved as follows:

[0015] The input module first divides the raw data according to the reservoir, and constructs a Spatial-Temporal matrix based on the data of the same reservoir as the input of the deep neural network.

[0016] The spatiotemporal feature extraction module extracts spatiotemporal features from input data by constructing a Conv-LSTM-SAT network based on convolution operations, long short-term memory networks, and self-attention mechanisms.

[0017] The feature fusion module integrates spatiotemporal features with fracturing data.

[0018] The regression prediction module uses a fully connected layer to perform regression prediction on the fused features and outputs the predicted production capacity value.

[0019] The method for dimensionality reduction of the data affecting the productivity of tight oil horizontal wells and the actual productivity data of tight oil horizontal wells is as follows: calculate the weighted correlation coefficient between each variable factor and the productivity of the oil well, and discard variable factors that are below the threshold.

[0020] The method for preprocessing the data affecting the productivity of tight oil horizontal wells and the actual productivity data of tight oil horizontal wells is as follows: First, the data is standardized using the Z-score method; then, outliers are identified and removed according to the 3σ principle; missing values ​​are filled using linear interpolation; and the data is normalized using the minimax method.

[0021] Furthermore, the method for constructing the Spatial-Temporal matrix in the input module is to select the production sequences of some oil wells and the target oil well to construct the Spatial-Temporal matrix as the input of the neural network, wherein some oil wells and the target oil well are located in the same reservoir.

[0022] A second aspect of the present invention provides a computer-readable storage medium storing computer program instructions; when the computer program instructions are executed by a processor, they implement the above-described method for predicting the productivity of a horizontal well in a tight oil reservoir.

[0023] A third aspect of the present invention provides a production capacity prediction system for horizontal wells in tight oil reservoirs, comprising the aforementioned computer-readable storage medium.

[0024] Compared with existing technologies, the beneficial effects of the method, medium, and system for predicting the productivity of horizontal wells in tight oil reservoirs provided by this invention are as follows: Based on the extraction of temporal features of productivity data, this invention considers the impact of strong heterogeneity of tight oil reservoirs on productivity prediction, and obtains the spatial features of productivity data. The original data is divided into reservoirs using the K-means algorithm to address the impact of strong inter-reservoir heterogeneity on productivity prediction; a Spatial-temporal matrix is ​​constructed based on the divided productivity sequences to aggregate spatiotemporal information, and a Conv-LSTM network is built to extract intra-reservoir spatial and temporal dependency information from the input data; a self-attention mechanism is introduced for the first time for tight oil productivity prediction, by assigning different weights to the productivity sequences, enabling the Conv-LSTM network to focus on learning data features more relevant to the current productivity, thus enhancing the accuracy of feature extraction. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart of a method for predicting the productivity of horizontal wells in tight oil reservoirs provided by the present invention;

[0027] Figure 2 The diagram shows the predictive model structure of a method for predicting the productivity of horizontal wells in tight oil reservoirs provided by this invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0030] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0031] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0032] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0033] like Figure 1 The diagram shown is a flowchart of a method for predicting the productivity of horizontal wells in tight oil reservoirs, provided by the first aspect of this invention. Figure 2 This is an overall model diagram of a horizontal well productivity prediction model for tight oil reservoirs. The method includes the following steps:

[0034] S10: Collect data on factors affecting the productivity of tight oil horizontal wells and the historical daily productivity data of the corresponding wells. Among them, the data affecting the productivity of tight oil horizontal wells include: fracturing stage, sand content per cluster, number of fracturing clusters, fracture spacing, fluid content per cluster, sand content per segment, porosity, permeability, oil saturation, Poisson's ratio, Young's modulus, brittleness index, horizontal segment length, oil layer penetration rate, thickness of penetrated oil layer, and penetration rate of Class I oil layer.

[0035] S20: Preprocess and reduce the dimensionality of the data affecting the productivity of tight oil horizontal wells and the historical daily productivity data of the corresponding wells to obtain the training dataset;

[0036] S30: Construct a production capacity prediction model for horizontal wells in tight oil reservoirs based on deep learning neural networks. The production capacity prediction model for horizontal wells in tight oil reservoirs includes an input module, a spatiotemporal feature extraction module, a feature fusion module, and a regression prediction module.

[0037] S40: Train the production capacity prediction model for horizontal wells in tight oil reservoirs using the training dataset;

[0038] S50: Input the data to be predicted into the production capacity prediction model for horizontal wells in tight oil reservoirs that is applicable to the actual situation, and obtain the production capacity prediction results;

[0039] S60: Deploy and adjust oil extraction projects based on the obtained production capacity forecast results.

[0040] In the above technical solution, the input module first divides the original data according to the reservoir, and constructs a Spatial-Temporal matrix based on the data of the same reservoir as the input of the deep neural network.

[0041] The specific steps in S30 include:

[0042] Step 1: Cluster the reservoir property data using the K-means algorithm. Divide the original data into reservoirs and randomly select a portion of oil wells that have the same clustering results as the target oil well. Construct a Spatial-Temporal matrix based on the production sequences of the selected oil wells and the target oil wells as the network input.

[0043] The Spatial-Temporal matrix is ​​as follows:

[0044]

[0045] In the formula, f represents the spatial-temporal matrix aggregated at time t for the well p to be predicted. t p This represents the production data of oil well p at time t. This represents the production data of oil well p from time tn to time t. n=8, m=9 indicates that the daily production sequence of 9 oil wells in the same reservoir, including the prediction well, over the past 9 days is selected as input data.

[0046] The second step is to construct a Conv-LSTM-SAT network based on convolutional operations, long short-term memory networks, and self-attention mechanisms. SAT assigns different weights to the production sequence, enabling Conv-LSTM to extract the spatiotemporal features of the input data more accurately.

[0047] The calculation process of Conv-LSTM is as follows:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] In the formula, input X t Let C be a time series vector with t time steps. t For storage units, H t For the output sequence, W xi ~W co For the convolution kernel, b i ~b o For bias terms, * indicates convolution operation. This represents the Hadamard product.

[0054] Step 3: Establish a feature fusion layer to connect and fuse spatiotemporal features with fracturing data as input to the fully connected layer.

[0055] Step 4: Construct a fully connected neural network to regress and fit the productivity of horizontal wells.

[0056] In the above technical solution, the spatiotemporal feature extraction module constructs a Conv-LSTM-SAT network based on convolution operations, long short-term memory networks, and self-attention mechanisms to extract the spatiotemporal features of the input data.

[0057] In the above technical solution, the feature fusion module connects and fuses spatiotemporal features with fracturing data.

[0058] In the above technical solution, the regression prediction module uses a fully connected layer to perform regression prediction on the fused features and outputs the predicted production capacity value.

[0059] In the above technical solution, the method for reducing the dimensionality of the data affecting the productivity of tight oil horizontal wells and the actual productivity data of tight oil horizontal wells is as follows: calculate the weighted correlation coefficient between each variable factor and the productivity of the oil well, and discard variable factors that are below the threshold.

[0060] In the above technical solution, the method for preprocessing the data affecting the productivity of tight oil horizontal wells and the actual productivity data of tight oil horizontal wells is as follows: First, the data is standardized using the Z-score method, and then outliers are identified and removed according to the 3σ principle; missing values ​​are filled using linear interpolation; and the data is normalized using the minimax method.

[0061] The formula for calculating the Z-score is as follows:

[0062]

[0063] In the formula, x represents the sample data. Let be the mean of the sample data, s be the standard deviation of the sample data, and z be the standardized data. Based on the 3σ principle, data with z > 3 or z < -3 are considered outliers and removed, and then the data are inversely standardized.

[0064] The formula for linear interpolation is as follows:

[0065]

[0066] In the formula, (X1,Y1), (X2,Y2), X represents known values, and Y1 represents missing values.

[0067] The formula for normalization using the max-min method is:

[0068]

[0069] In the formula, x′ represents the normalized sample data, x represents the data to be normalized, and x' represents the normalized sample data. min and x max These are the maximum and minimum values ​​of the data to be normalized.

[0070] Furthermore, in the above scheme, the method for reducing the dimensionality of the data in step S10 is to calculate the weighted correlation coefficient between each variable factor and the oil well production capacity, and discard variable factors that are below the threshold.

[0071] The formula for calculating the weighted correlation coefficient is as follows:

[0072]

[0073] In the formula, n is the total number of variable factors, I(Y, X) i ) represents the mutual information between the i-th variable and output. Let P be the absolute value of the Pearson correlation coefficient between the i-th variable and output. i This is the weighted correlation coefficient.

[0074] Furthermore, in the above technical solution, the method for constructing the Spatial-Temporal matrix in the input module is to select the production sequences of some oil wells and the target oil well to construct the Spatial-Temporal matrix as the input of the neural network, wherein some oil wells and the target oil well are located in the same reservoir.

[0075] A second aspect of the present invention provides a computer-readable storage medium storing computer program instructions; when the computer program instructions are executed by a processor, they implement the above-described method for predicting the productivity of a horizontal well in a tight oil reservoir.

[0076] A third aspect of the present invention provides a production capacity prediction system for horizontal wells in tight oil reservoirs, comprising the aforementioned computer-readable storage medium.

[0077] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting the productivity of horizontal wells in tight oil reservoirs, characterized in that, Includes the following steps: S10: Collect data on factors affecting the productivity of tight oil horizontal wells and the historical daily productivity data of the corresponding wells. Among them, the data affecting the productivity of tight oil horizontal wells include: fracturing stage, sand content per cluster, number of fracturing clusters, fracture spacing, fluid content per cluster, sand content per segment, porosity, permeability, oil saturation, Poisson's ratio, Young's modulus, brittleness index, horizontal segment length, oil layer penetration rate, thickness of penetrated oil layer, and penetration rate of Class I oil layer. S20: Preprocess and reduce the dimensionality of the data affecting the productivity of tight oil horizontal wells and the historical daily productivity data of the corresponding wells to obtain the training dataset; S30: A production capacity prediction model for horizontal wells in tight oil reservoirs is constructed based on deep learning neural networks, including an input module, a spatiotemporal feature extraction module, a feature fusion module, and a regression prediction module. S40: Train the production capacity prediction model for the horizontal well in the tight oil reservoir using the training dataset; S50: Input the data to be predicted into the production capacity prediction model of the horizontal well in the tight oil reservoir to obtain the production capacity prediction results; S60: Deploy and adjust oil extraction projects based on the obtained production capacity forecast results; The specific steps in S30 include: Step 1: Use the K-means algorithm to cluster the reservoir property data. Divide the original data according to the reservoir, randomly select a portion of oil wells with the same clustering results as the target oil well, and construct a Spatial-Temporal matrix based on the production sequences of the selected oil wells and the target oil wells as the network input. The Spatial-Temporal matrix is ​​as follows: ； In the formula, Indicates the well to be predicted exist Spatial-Temporal matrix aggregated at each time step. Indicates oil well exist Real-time production data, Indicates oil well from Time's up Real-time production data, , This means that the daily production sequence of 9 oil wells in the same reservoir, including the predicted well, over the past 9 days is selected as input data. Step 2: Conv-LSTM-SAT network is constructed based on convolution operation, long short-term memory network and self-attention mechanism. SAT assigns different weights to the production sequence, so that Conv-LSTM can extract the spatiotemporal features of the input data more accurately. The calculation process of Conv-LSTM is as follows: ； ； ； ； ； In the formula, the input for A time-series vector at each time step. For storage units, For the output sequence, For convolution kernel, For bias terms, This represents the convolution operation. Represents the Hadamard product; Step 3: Establish a feature fusion layer to connect and fuse spatiotemporal features with fracturing data as input to the fully connected layer; Step 4: Construct a fully connected neural network to regress and fit the productivity of horizontal wells.

2. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The input module first divides the raw data according to the reservoir, and constructs a Spatial-Temporal matrix based on the data of the same reservoir as the input of the deep neural network.

3. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The spatiotemporal feature extraction module extracts spatiotemporal features from the input data by constructing a Conv-LSTM-SAT network based on convolution operations, long short-term memory networks, and self-attention mechanisms.

4. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The feature fusion module connects and fuses spatiotemporal features with fracturing data.

5. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The regression prediction module uses a fully connected layer to perform regression prediction on the fused features and outputs the predicted production capacity value.

6. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The dimensionality reduction method in step S20 is to calculate the weighted correlation coefficient between each variable factor and the oil well production capacity, and discard variable factors that are below the threshold.

7. The method for predicting the productivity of horizontal wells in tight oil reservoirs according to claim 1, characterized in that, The preprocessing method in step S20 is as follows: First, the data is standardized using the Z-score method; then, outliers are identified and removed according to the 3σ principle; missing values ​​are filled using linear interpolation; and the data is normalized using the min-max method.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, they implement a method for predicting the productivity of a horizontal well in a tight oil reservoir as described in any one of claims 1 to 7.

9. A production capacity prediction system for horizontal wells in tight oil reservoirs, characterized in that, It includes the computer-readable storage medium as described in claim 8.

Images