Oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network

Abstract

The present invention provides an oil production prediction method based on a Voronoi diagram and a spatiotemporal graph convolutional neural network, and the oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network comprises: step 1, obtaining target well network data; step 2, establishing a structure diagram according to the location information of the well network; step 3, processing the adjacency matrix based on the Voronoi diagram; step 4, establishing a spatiotemporal graph convolutional neural network model; step 5, performing network model training and oil production prediction. The oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network can not only take into account the influence of relevant factors on oil production, but also take into account the temporal correlation of data, and has high prediction accuracy, and can be used as a new method for predicting oil production in oilfield production.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development, and in particular to an oil production prediction method based on a Voronoi diagram and a spatiotemporal graph convolutional neural network. Background Art

[0002] The artificial intelligence algorithms currently used in oil production prediction of oil field production wells are mainly based on back-propagation neural networks and support vector machine regression algorithms. However, these two methods cannot take into account the temporal correlation of data during the prediction process.

[0003] In the Chinese patent application with application number: CN201910917091.8, a method and system for detecting oil and gas productivity is involved, which is related to the field of oil and gas reservoir exploration and development. The method includes: inputting numerical data, categorical numerical data and structured data into a preset hybrid neural network model respectively; processing the numerical data and categorical numerical data through a multi-layer perceptron network to obtain a first output result; processing the structured data through a convolutional neural network to obtain a second output result; processing the first output result and the second output result through a fully connected layer and an activation function to obtain the productivity of the oil and gas well.

[0004] In the Chinese patent application with application number: CN202010604568.X, an artificial intelligence-based oil production prediction method is involved. The method includes: inputting oil production-related information at multiple historical time points into a trained time series prediction model, outputting oil production-related information at the next time point in sequence, and then obtaining a time series prediction value; for the time series prediction value, judging the abnormal event and the corresponding time point, and with the goal of reducing the loss caused by abnormal fluctuations, calculating the predicted value of the subsequent oil production-related information at the abnormal event time point through a sliding average model; replacing the predicted value of the corresponding time point output by the time series model with the calculated predicted value of the oil production-related information, and obtaining the oil production prediction result.

[0005] In the Chinese patent application with application number: CN201510535506.7, an improved neural network oil well production prediction method and device are involved, which relates to the field of oil production. It optimizes the hidden layer connection weights and output layer connection weights in the neural network through an improved particle swarm optimization algorithm, and then uses the neural network and historical oil well basic parameters to establish an oil and gas production prediction model, and finally obtains the oil well production prediction value under the sample value by inputting the sample value.

[0006] The above existing technologies are significantly different from the present invention and fail to solve the technical problem we want to solve. For this reason, we have invented a new oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network. Summary of the invention

[0007] The purpose of the present invention is to provide an oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network with high prediction accuracy, which can be applied to oil production prediction in oil fields.

[0008] The purpose of the present invention can be achieved by the following technical measures: an oil production prediction method based on a Voronoi diagram and a time-space graph convolutional neural network, the oil production prediction method based on a Voronoi diagram and a time-space graph convolutional neural network comprises:

[0009] Step 1, obtaining target well pattern data;

[0010] Step 2, establishing a structure diagram according to the location information of the well pattern;

[0011] Step 3, processing the adjacency matrix based on the Voronoi diagram;

[0012] Step 4, establish a spatiotemporal graph convolutional neural network model;

[0013] Step 5: Conduct network model training and oil production prediction.

[0014] The purpose of the present invention can also be achieved by the following technical measures:

[0015] In step 1, various information of the target production well network for oil production prediction is obtained, including oil production rate, water injection rate, well location, perforation information, formation thickness, permeability, fluid viscosity, pressure data, and the status of the switch wells at different times.

[0016] In step 2, according to the obtained location information of the well network, the water injection wells and production wells in the well network are used as data structure nodes, and the connections between the water injection wells and production wells are used as edges to establish a structural diagram corresponding to the target well network.

[0017] In step 3, the Voronoi diagram is composed of a set of continuous polygons consisting of perpendicular bisectors of straight lines connecting two adjacent points. N distinct points on the plane divide the plane according to the nearest neighbor principle.

[0018] In step 3, in different regions to which the nodes belong in the Voronoi diagram, the average data in the region is used to replace the data of the nodes.

[0019] In step 3, the known data is substituted into formula 1 to calculate the qij between different nodes to represent the adjacency matrix between two nodes:

[0020] q ij =T ij (p i -p j ) Formula 1

[0021]

[0022] In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, and pj represents pressure at node j. In formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

[0023] In step 4, the oil production rate of the production wells at M moments from t-M+1 to t, the injection rate of the injection wells, the pressure data of each well, and the switch well status are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, wherein each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn, and the output layer includes a one-dimensional convolution in time and a fully connected layer; thereby obtaining the oil production rate and water content of the production wells at H moments from t+1 to t+H, and optimizing the error by calculating the mean square error (MSE) with the known data.

[0024] In step 5, the data at multiple moments are divided into a training set and a test set. The error is calculated and optimized through the training set to obtain the spatiotemporal graph convolutional neural network model, and its prediction accuracy is verified on the test set.

[0025] The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network in the present invention predicts the oil production of production wells by using the spatiotemporal graph convolutional neural network and Voronoi diagram, and the historical production data, well location data and permeability data of the well network. The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network can not only take into account the influence of relevant factors on oil production, but also take into account the temporal correlation of data, and has high prediction accuracy. It can be used as a new method for predicting the oil production of production wells in oil fields, and as a method with high prediction accuracy and high computational efficiency, it can be used for predicting the oil production of production wells in oil fields, providing technical support for further improving the recovery rate of water-driven oil reservoirs in the ultra-high water cut period. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of a specific embodiment of the oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network of the present invention;

[0027] Figure 2 This is a schematic diagram of the well pattern structure in a specific embodiment 1 of the present invention;

[0028] Figure 3A schematic diagram of dividing different nodes according to a Voronoi diagram in a specific embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the well pattern structure in a specific embodiment 2 of the present invention;

[0030] Figure 5 It is a schematic diagram of prediction results of some wells in the test set in a specific embodiment 2 of the present invention;

[0031] Figure 6 This is a schematic diagram of the well network structure in a specific embodiment 3 of the present invention. DETAILED DESCRIPTION

[0032] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0033] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.

[0034] like Figure 1 As shown, Figure 1 The flowchart of the oil production prediction method based on the Voronoi diagram and the time-space graph convolutional neural network of the present invention. The oil production prediction method based on the Voronoi diagram and the time-space graph convolutional neural network includes the following steps:

[0035] Step 101: Obtain target well pattern data

[0036] Obtain various information of the target production well network for oil production prediction, including oil production rate, water injection rate, well location, perforation information, formation thickness, permeability, fluid viscosity, pressure data and the status of well opening and closing at different times.

[0037] Step 102: Create a structure diagram based on the location information of the well network

[0038] According to the location information of the well network obtained, the structural diagram corresponding to the target well network is established with the injection wells and production wells in the well network as data structure nodes and the connections between the injection wells and production wells as edges.

[0039] Step 103: Process the adjacency matrix based on the Voronoi diagram

[0040] The Voronoi diagram consists of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two adjacent points. N distinct points on the plane divide the plane according to the nearest neighbor principle.

[0041] In the Voronoi diagram, the average data in the region is used to replace the data of the node. The known data is substituted into Formula 1 to calculate the qij between different nodes, which is used to represent the adjacency matrix between two nodes.

[0042] q ij =T ij (p i -p j ) Formula 1

[0043]

[0044] In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, in formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

[0045] Step 104: Build a spatiotemporal graph convolutional neural network model

[0046] The oil production rate of the production wells at M moments from t-M+1 to t, the water injection rate of the water injection wells, the pressure data of each well, and the switch well status (0 means closing the well, 1 means opening the well) are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, where each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn. The output layer includes a one-dimensional convolution in time and a fully connected layer. Thus, the oil production rate and water content of the production wells at H moments from t+1 to t+H are obtained, and the error is optimized by calculating the MSE (mean square error) with the known data.

[0047] Step 105: Network model training and oil production prediction

[0048] The data at multiple moments are divided into training sets and test sets. The error is calculated and optimized through the training set to obtain the spatiotemporal graph convolutional neural network model, and its prediction accuracy is verified on the test set.

[0049] In the specific embodiment 1 of the present invention, the oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network includes the following steps:

[0050] Step 1: Obtain target well pattern data

[0051] Obtain various information of the target production well network for oil production prediction, including oil production rate, water injection rate, well location, perforation information, formation thickness, permeability, fluid viscosity, pressure data and the status of well opening and closing at different times.

[0052] Step 2: Create a structure diagram based on the location information of the well network

[0053] According to the location information of the well network, the injection wells and production wells in the well network are used as data structure nodes, and the connection between the injection wells and production wells is used as an edge to establish the structure diagram corresponding to the target well network, such as Figure 2 As shown (1, 3, 5, 7 are production wells, 2, 4, 6, 8 are water injection wells).

[0054] Step 3: Process the adjacency matrix based on the Voronoi diagram

[0055] The Voronoi diagram is composed of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two adjacent points. N points that are distinct on the plane are divided into planes according to the nearest neighbor principle, such as Figure 3 As shown, it is divided into 8 polygons according to W1, W2, W3, W4, W5, W6, W7, and W8.

[0056] In the Voronoi diagram, the average data in the region is used to replace the data of the node. The known data is substituted into Formula 1 to calculate the qij between different nodes, which is used to represent the adjacency matrix between two nodes.

[0057] q ij =T ij (p i -p j ) Formula 1

[0058]

[0059] In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, in formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

[0060] Step 4: Build a spatiotemporal graph convolutional neural network model

[0061] The oil production rate of the production wells at M moments from t-M+1 to t, the water injection rate of the water injection wells, the pressure data of each well, and the switch well status (0 means closing the well, 1 means opening the well) are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, where each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn. The output layer includes a one-dimensional convolution in time and a fully connected layer. Thus, the oil production rate and water content of the production wells at H moments from t+1 to t+H are obtained, and the error is optimized by calculating the MSE (mean square error) with the known data.

[0062] Step 5: Network model training and oil production prediction.

[0063] The data at multiple moments are divided into training sets and test sets. The error is calculated and optimized through the training set to obtain the spatiotemporal graph convolutional neural network model, and its prediction accuracy is verified on the test set.

[0064] In the specific embodiment 2 of the present invention, a study on oil production prediction is carried out for a water injection development oil reservoir in Shengli. The oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network includes the following steps:

[0065] Step 1: Obtain target well pattern data

[0066] Obtain various information of the target production well network for oil production prediction, including the oil production rate, water injection rate, well location, nine-point average pressure, well permeability, well porosity and the status of the opening and closing wells at different times of the 49 wells.

[0067] Step 2: Create a structure diagram based on the location information of the well network

[0068] According to the location information of the well network, the injection wells and production wells in the well network are used as data structure nodes, and the connection between the injection wells and production wells is used as an edge to establish the structure diagram corresponding to the target well network. The well network structure diagram at a certain moment is as follows: Figure 4 As shown (the solid circles are production wells, and the hollow circles with arrows are injection wells).

[0069] Step 3: Process the adjacency matrix based on the Voronoi diagram

[0070] The Voronoi diagram consists of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two adjacent points.49 Distinct points on the plane, partition the plane according to the nearest neighbor principle.

[0071] In the Voronoi diagram, the average data in the region is used to replace the data of the node. The known data is substituted into Formula 1 to calculate the qij between different nodes, which is used to represent the adjacency matrix between two nodes.

[0072] q ij =T ij (p i -p j ) Formula 1

[0073]

[0074] In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, in formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

[0075] Step 4: Build a spatiotemporal graph convolutional neural network model

[0076] The oil production rate of the production wells at M moments from t-M+1 to t, the water injection rate of the water injection wells, the pressure data of each well, and the switch well status (0 means closing the well, 1 means opening the well) are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, where each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn. The output layer includes a one-dimensional convolution in time and a fully connected layer. Thus, the oil production rate and water content of the production wells at H moments from t+1 to t+H are obtained, and the error is optimized by calculating the MSE (mean square error) with the known data.

[0077] Step 5: Network model training and oil production prediction.

[0078] In this example, there are 748 time steps in total, 70% of the data set is selected as the training set and 30% as the validation set; the input time step is 12 and the predicted time step is 3. The spatiotemporal graph convolutional neural network model is obtained by error calculation and optimization on the training set, and its prediction accuracy is verified on the test set. The results show that the prediction accuracy of 83% of the wells reaches more than 85%. Some of the prediction results are shown in Figure 5 .

[0079] In the specific embodiment 3 of the present invention, a study on oil production prediction is carried out for a water injection development oil reservoir in Shengli. The oil production prediction method based on the Voronoi diagram and the spatiotemporal graph convolutional neural network includes the following steps:

[0080] Step 1: Obtain target well pattern data

[0081] Obtain various information of the target production well network for oil production prediction, including oil production rate, water injection rate, well location, nine-point average pressure, well permeability, well porosity and the status of the opening and closing wells at different times of the 81 wells.

[0082] Step 2: Create a structure diagram based on the location information of the well network

[0083] According to the location information of the well network, the injection wells and production wells in the well network are used as data structure nodes, and the connection between the injection wells and production wells is used as an edge to establish the structure diagram corresponding to the target well network. The schematic diagram of the well network structure at a certain moment is as follows: Figure 6 As shown (the solid circles are production wells, and the hollow circles with arrows are injection wells).

[0084] Step 3: Process the adjacency matrix based on the Voronoi diagram

[0085] The Voronoi diagram consists of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two adjacent points. 81 points that are distinct on the plane, partition the plane according to the nearest neighbor principle.

[0086] In the Voronoi diagram, the average data in the region is used to replace the data of the node. The known data is substituted into Formula 1 to calculate the qij between different nodes, which is used to represent the adjacency matrix between two nodes.

[0087] q ij =T ij (p i -p j ) Formula 1

[0088]

[0089] In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, in formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

[0090] Step 4: Build a spatiotemporal graph convolutional neural network model

[0091] The oil production rate of the production wells at M moments from t-M+1 to t, the water injection rate of the water injection wells, the pressure data of each well, and the switch well status (0 means closing the well, 1 means opening the well) are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, where each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn. The output layer includes a one-dimensional convolution in time and a fully connected layer. Thus, the oil production rate and water content of the production wells at H moments from t+1 to t+H are obtained, and the error is optimized by calculating the MSE (mean square error) with the known data.

[0092] Step 5: Network model training and oil production prediction.

[0093] In this example, there are 548 time steps in total, 70% of the data set is selected as the training set and 30% as the validation set; the input time step is 12 and the predicted time step is 3. The spatiotemporal graph convolutional neural network model is obtained by error calculation and optimization on the training set, and its prediction accuracy is verified on the test set. The results show that the prediction accuracy of 76% of the wells reaches more than 85%.

[0094] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0095] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network is characterized by: The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network includes: Step 1, obtaining target well pattern data; Step 2, establishing a structure diagram according to the location information of the well pattern; Step 3, processing the adjacency matrix based on the Voronoi diagram; Step 4, establish a spatiotemporal graph convolutional neural network model; Step 5, network model training and oil production prediction; In step 1, various information of the target production well network for which oil production prediction is required is obtained, including oil production rate, water injection rate, well point location, perforation information, formation thickness, permeability, fluid viscosity, pressure data, and the status of the opening and closing wells at different times; In step 2, according to the obtained location information of the well network, the water injection wells and production wells in the well network are used as data structure nodes, and the connection between the water injection wells and the production wells is used as an edge to establish a structure diagram corresponding to the target well network; In step 3, the Voronoi diagram is composed of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two neighboring points. N points that are distinct on the plane divide the plane according to the nearest neighbor principle; In step 3, in different regions to which the nodes belong in the Voronoi diagram, the average data in the region is used to replace the data of the nodes; In step 3, the known data is substituted into formula 1 to calculate the qij between different nodes to represent the adjacency matrix between two nodes: q ij =T ij (p i -p j ) Formula 1 In formula 1, qij represents flow rate, Tij represents conductivity between node i and node j, pi represents pressure at node i, and pj represents pressure at node j. In formula 2, Kij represents average permeability between node i and node j, μij represents fluid viscosity between node i and node j, Lij represents distance between node i and node j, and Aij represents thickness of formation of perforated layer between node i and node j multiplied by boundary length between two points in Voronoi diagram.

2. The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network according to claim 1 is characterized in that: In step 4, the oil production rate of the production wells at M moments from t-M+1 to t, the injection rate of the injection wells, the pressure data of each well, and the switch well status are input as node features, and the input data passes through two spatiotemporal convolution blocks and an output layer, wherein each spatiotemporal convolution block performs a one-dimensional convolution in time, a graph convolution in space, and a one-dimensional convolution in time in turn, and the output layer includes a one-dimensional convolution in time and a fully connected layer; thereby obtaining the oil production rate and water content of the production wells at H moments from t+1 to t+H, and optimizing the error by calculating the mean square error (MSE) with the known data.

3. The oil production prediction method based on Voronoi diagram and spatiotemporal graph convolutional neural network according to claim 2 is characterized in that: In step 5, the data at multiple moments are divided into a training set and a test set. The error is calculated and optimized through the training set to obtain the spatiotemporal graph convolutional neural network model, and its prediction accuracy is verified on the test set.

Images