KR20200084447A - Decision system and method of well completion and hydraulic fracturing liquid based on the artificial neural network in shale gas reservoir - Google Patents

Abstract

The present invention relates to an expert system for deciding well completion and a hydraulic fracturing liquid when designing hydraulic fracturing for production of shale gas and, more specifically, to a system for deciding well completion and a hydraulic fracturing liquid in a shale gas reservoir based on an artificial neural network and a decision method using the same, which can accurately decide types of well completion and hydraulic fracturing liquid by using property data on geological feature information, mineralogical feature information, and rock mechanical feature information of the shale gas at the time of hydraulic fracturing of the shale gas.

Description

Decision system and method of well completion and hydraulic fracturing liquid based on the artificial neural network in shale gas reservoir}

The present invention relates to an expert system for selecting an oil well completion method and a hydraulic fracturing solution when designing hydraulic fracturing for the production of shale gas, and more specifically, geological property information, mineralogical property information of shale gas during hydraulic fracturing of shale gas, By using the physical properties data for the rock dynamics information, the hydraulic crushing fluid and oil well completion selection system of artificial neural network based shale gas reservoir that can more accurately select the type of oil well completion method and the type of hydraulic crushing fluid, and the selection method using the same It is about.

The shale gas reservoir layer is formed when the shale containing organic materials is deposited, and then forms a source rock through thermal maturation, and the generated gas does not migrate to a porous medium and remains. Since shale gas is present in the extremely dense permeable body and shows a wide distribution in the form of a blanket without following the existing petroleum system, it is essential to apply horizontal drilling and hydraulic fracturing in the development and production process.

In order to develop such shale gas, it is necessary to select the well completion method and the type of hydraulic fracturing solution through simulation based on the geological and mineral data of the reservoir. However, when applied to the hydraulic fracturing design based on the result of relying on empirical judgment, there is a reliability problem with the result, and when determining the main factors through comprehensive analysis of various reservoir information, the analysis process is not only complicated, but also a lot of time Is consumed.

Therefore, it is necessary to develop a technology that is not only reliable, but also capable of selecting an appropriate type of oil well completion method and a type of hydraulic fracturing solution when designing a hydraulic fracturing method by determining a major influencer through comprehensive analysis of various reservoir information. This exists.

Republic of Korea Patent Registration No. 10-1657889

The present inventors completed the present invention by developing an oil well completion method and a hydraulic fracturing fluid selection system of an artificial neural network based shale gas reservoir that can be applied when designing a hydraulic fracturing method.

Accordingly, the object of the present invention is the depth, thickness (thickness), total carbon content (total carbon contents: TOC), adsorption gas amount among various geological property information, mineralogical property information, and rock epidemiological property information of the shale gas reservoir. (adsorbed gas contents), gas contents, quartz contents, clay contents, carbonate contents and brittleness index, oil well completion method and types of hydraulic fracturing solution Constructing an artificial neural network (ANN) in consideration of the correlation of, and using only simple shale gas reservoir information within a very short time through the artificial neural network, types of well completion methods and hydraulic fracturing solutions that match the characteristics of the shale gas reservoir layer It is to provide an oil well completion method and a hydraulic fracturing fluid selection system and a selection method for an artificial neural network based shale gas reservoir capable of selecting a type.

Another object of the present invention is to establish an oil well completion method and a hydraulic fracturing fluid selection system and a selection method for an artificial neural network-based shale gas reservoir that can be used as a tool in designing an optimal hydraulic fracturing system to meet shale layer characteristics in establishing a shale gas development plan. Is to provide.

The object of the present invention is not limited to the above-mentioned object, and even if not explicitly mentioned, the object of the invention that can be recognized by those skilled in the art from the description of the detailed description of the invention to be described later may also be included naturally. .

In order to achieve the object of the present invention described above, the present invention provides geological properties information, mineral properties information and shale gas reservoir layer affecting oil well completion method and hydraulic fracturing solution design when designing a hydraulic fracturing method for shale gas production. An input unit for inputting reservoir characteristic information including petrochemical characteristic information; Performing learning and validity check through updating the connection weight value through an artificial neural network set to predict result information including the type of oil well completion method and the type of hydraulic fracturing solution according to the reservoir characteristic information input by the input unit Data processing department; And an output unit showing result information including the type of oil well completion method predicted by the data processing unit and the type of hydraulic fracturing fluid; and an artificial neural network-based shale gas reservoir layer hydraulic well completion method and hydraulic fracturing fluid selection system. do.

In a preferred embodiment, the data processing unit determines geological properties and minerals determined to affect the well completion method and the selection of the hydraulic fracturing solution when designing a hydraulic fracturing method for shale gas production through a survey of known physical properties of shale gas reservoirs. It further includes a shale gas reservoir characteristic information DB that is constructed of a plurality of reservoir characteristic information sets including medical characteristics and petrochemical characteristics to provide a number of empirical information for training the artificial neural network.

In a preferred embodiment, the artificial neural network includes an input layer having a plurality of input nodes associated with the reservoir characteristic information; An output layer having at least one output node representing the result information based on the storage layer characteristic information; And two or more hidden layers having at least one hidden node having processing elements for storage layer characteristic information and result information between the input layer and the output layer.

In a preferred embodiment, the artificial neural network is set to assign a connection weight calculated by an empirically trained model to at least one of between the input node and the hidden node and between the hidden node and the output node.

In a preferred embodiment, learning through updating the connection weight value is performed by a backpropagation learning algorithm.

In a preferred embodiment, among the characteristic information of the reservoir layer, the characteristic information affecting the prediction of the type of oil well completion method is a group consisting of the depth, thickness, total carbon content, quartz content, gas content, gas adsorption amount, and clay content of the reservoir layer. It is one or more selected from.

In a preferred embodiment, the type of the well completion method represented by the result information in the output unit according to the characteristic information is any one of a single stage vertical tablet, a multi-stage vertical tablet, and a horizontal tablet.

In a preferred embodiment, the characteristic information influencing the selection of the type of hydraulic fracturing fluid among the storage layer characteristic information is selected from the group consisting of depth, pressure change according to depth change, temperature reference, and brittleness index during shale gas development. It is one or more.

In a preferred embodiment, the type of the hydraulic fracturing solution represented by the result information on the output unit according to the characteristic information is any one of carbon dioxide foam containing propane, nitrogen foam containing propane, and water-based fracturing fluid.

In a preferred embodiment, when the hydraulic fracturing type is water-based hydraulic fracturing, it is further classified into water containing a chemical additive, a linear gel, a crosslinking gel, and any one determined according to the brittleness index in the further classified state. Appears as the result information.

In a preferred embodiment, the brittleness index is a value obtained by dividing the relationship between Young's modulus, Poisson's ratio, porosity, or quartz content by adding all of the quartz content, clay content, and carbonate content.

In addition, the present invention comprises the steps of constructing a shale gas reservoir characteristic information DB that provides empirical information on the shale gas reservoir characteristic information that affects the oil well completion method and the selection of the hydraulic fracturing solution when designing a hydraulic fracturing method for shale gas production; Constructing an artificial neural network capable of correlating the contribution to the result information including the type of oil well completion method and the type of hydraulic fracturing according to the reservoir characteristic information and the reservoir characteristic information: the artificial neural network type and water pressure of the well completion method Training the artificial neural network by a backpropagation learning algorithm using the shale gas reservoir characteristic information DB to predict result information including a type of crushing liquid; And estimating result information by the artificial neural network. An oil well completion method of an artificial neural network based shale gas reservoir and a method of selecting a hydraulic fracturing solution are provided.

In a preferred embodiment, the step of predicting the result information by the artificial neural network includes the step of performing learning and feasibility check by the artificial neural network by updating the connection weight value.

In a preferred embodiment, the step of constructing the artificial neural network includes single-stage vertical and multi-stage vertical tablets for storage layer characteristic information related to the depth, thickness, total carbon content, quartz content, gas content, gas adsorption, and clay content of the reservoir layer. , And associating any one of the horizontal wells with a type of well completion method; And carbon dioxide foam containing propane, nitrogen foam containing propane, and water-based crushing liquid for depth information related to depth, pressure change according to depth change, temperature standards, and brittleness index It is carried out including the step of associating the type of hydraulic fracturing solution.

In a preferred embodiment, the training of the artificial neural network may include repeatedly learning to predict the type of well completion method by determining the contribution of each variable using a Bayesian regularization algorithm; And using the Levenberg-Marquart algorithm to determine the contribution of each variable to repeatedly learn to predict the type of hydraulic fracturing fluid.

In a preferred embodiment, the contribution of each variable is determined by assigning a connection weight based on empirical information and error correction provided from the reservoir characteristic information DB, and the connection weight is adjusted multiple times so as to be properly set.

According to the present invention described above, the depth, thickness, total carbon content (TOC), and the amount of adsorbed gas (depth) among various geological property information, mineralogical property information, and rock epidemiological property information of the shale gas reservoir layer ( Types of adsorbed gas contents, gas contents, quartz contents, clay contents, carbonate contents and brittleness index and types of well completion and hydraulic fracturing solutions Constructing an artificial neural network (ANN) in consideration of correlation, and using only the simple shale gas reservoir information within a very short time through the artificial neural network, the type of well completion method and hydraulic fracturing solution that match the characteristics of the shale gas reservoir layer Can be selected.

In addition, the selection system and the selection method of the present invention can be used as a tool in designing an optimal hydraulic fracturing that conforms to the characteristics of the shale layer in establishing a shale gas development plan. There is this.

These technical effects of the present invention are not limited to the above-mentioned ranges, and even if not explicitly stated, the effects of the invention that can be recognized by those of ordinary skill from the description of specific contents for the practice of the invention described below are also Of course it is included.

1 is a block diagram of an artificial neural network based oil well completion method and a hydraulic fracturing fluid selection system according to an embodiment of the present invention.
FIG. 2 is a flow chart for the selection of a well completion method established for the development of an oil well completion method and a hydraulic fracturing solution selection system for an artificial neural network based shale gas reservoir of the present invention.
3 is a hydraulic crushing fluid selection flow chart constructed for the development of an oil well completion method and a hydraulic crushing fluid selection system for an artificial neural network based shale gas reservoir of the present invention.
4 is a graph showing the results of analyzing the sensitivity of the target value and the output value of the system of the well completion method selection training data set using Bayesian regularization.
5 is a schematic diagram of an artificial neural network (ANN) structure for selecting an oil well completion method applied to an oil well completion method and a hydraulic fracturing fluid selection system of an artificial neural network based shale gas reservoir according to the present invention when there are two hidden layers.
6 is a graph showing the result of analyzing the sensitivity of the target value and the output value of the system of the hydraulic fracturing fluid selection training data set using Levenberg-Marquart.
7 is a schematic diagram of a hydraulic crushing fluid selection artificial neural network (ANN) structure applied to an oil well completion method and a hydraulic crushing fluid selection system of an artificial neural network based shale gas storage layer according to the present invention when there are two hidden layers.
8 is an exemplary view showing an embodiment of the panel of the hydraulic well completion method and hydraulic fracturing solution selection system based on the artificial neural network-based shale gas reservoir according to the present invention implemented with a graphical user interface (GUI).
Figure 9 is another embodiment of a GUI according to the present invention, the oil well completion method of the artificial gas network based shale gas reservoir, and the pressure change amount according to the temperature, depth change, and the panel according to the temperature of the hydraulic fracturing solution selection system. It is an exemplary diagram showing.
FIG. 10 shows a panel based on petrochemical properties implemented as a graphical user interface (GUI) when water-based crushing fluid is selected in an oil well completion method and a hydraulic crushing fluid selection system of an artificial neural network-based shale gas reservoir according to the present invention. It is an exemplary view showing another embodiment.
FIG. 11 is a GUI (graphic user interface) of panels according to geological and mineral properties when water-based crushing fluid is selected in the well completion method of the artificial neural network-based shale gas reservoir and the hydraulic crushing fluid selection system according to the present invention. It is an exemplary view showing another implementation.
12 is another embodiment of performing an on-site applicability evaluation using an oil well completion method and a hydraulic fracturing fluid selection system of an artificial neural network-based shale gas reservoir according to the present invention.
13 is another embodiment of performing an on-site applicability evaluation using an oil well completion method and a hydraulic fracturing solution selection system for an artificial neural network-based shale gas reservoir according to the present invention.

The terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as “include” or “have” are intended to indicate the presence of features, numbers, steps, actions, components, parts or combinations thereof described in the description of the invention, one or more other It should be understood that features or numbers, steps, actions, components, parts or combinations thereof are not excluded in advance.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from other components. For example, the first component may be referred to as a second component without departing from the scope of the present invention, and similarly, the second component may also be referred to as a first component.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art to which the present invention pertains. Terms, such as those defined in a commonly used dictionary, should be interpreted as having meanings consistent with meanings in the context of related technologies, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined in the present invention. Does not.

In interpreting the components, it is interpreted as including the error range even if there is no explicit description. In particular, when the terms "about", "substantially", etc. of degree are used, it can be interpreted as being used in or close to the value when manufacturing and substance tolerances specific to the stated meaning are given. .

In the case of a description of a time relationship, for example,'after','following','~after','~before', etc. When the temporal sequential relationship is described,'right' or'directly' It also includes cases that are not continuous unless' is used.

Hereinafter, the technical configuration of the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments.

However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. The same reference numerals used to describe the present invention throughout the specification indicate the same components.

Technical features of the present invention is a variety of data collection and a variety of information that affects the determination of the type of oil well completion method and hydraulic crushing fluid, among a variety of geological property information, mineralogical property information and rock epidemiological property information of the shale gas reservoir. Depth, thickness, total carbon content (TOC), adsorbed gas contents, which are selected and selected through the simulation of conditions and selected geological and mineral properties information. Correlation between gas contents, quartz contents, clay contents, carbonate contents, and brittleness index, which is information on rock dynamics, well completion and hydraulic fracturing Constructing an artificial neural network (ANN) in consideration of the relationship, and using only the simple shale gas reservoir layer information in a very short time through the artificial neural network, select the type of well completion method and hydraulic fracturing solution that match the characteristics of the shale gas reservoir layer. It is in the well completion method and hydraulic crushing fluid selection system and selection method of the shale gas reservoir based on the artificial neural network that can be selected.

That is, according to the existing studies, in order to determine the well completion method, the geological properties of the reservoir such as depth, thickness, total carbon contents (TOC), gas content, etc. Although it was considered to be the most important in carrying out, there was a problem in that a smaller reservoir volume (SRV) than expected was shown. In addition, in the case of the hydraulic fracturing solution, only the stability of the crushing liquid was considered important in the past, so the depth of the shale layer, the pressure of the reservoir layer, and the temperature have been considered, but in the case of the water-soluble hydraulic fracturing liquid, This is because the difference in the crushing volume of the reservoir layer is very large, so additional characteristic information for correcting it needs to be considered. However, in the present invention, by considering the quartz contents and clay contents as characteristic information that influences the determination of the type of oil well completion method, an oil well completion method capable of increasing the crushing volume of the reservoir can be selected. In addition, by additionally considering the brittleness as characteristic information that affects the determination of the type of hydraulic fracturing solution, it is possible to minimize the difference in the storage layer fracturing volume according to the additives of the fracturing solution.

Therefore, the oil well completion method and the hydraulic fracturing fluid selection system of the artificial neural network-based shale gas reservoir of the present invention are used to design the hydraulic fracturing method for shale gas production. An input unit for inputting reservoir characteristic information including characteristic information, mineralogical characteristic information and petrochemical characteristic information; Performing learning and validity check through updating the connection weight value through an artificial neural network set to predict result information including the type of oil well completion method and the type of hydraulic fracturing solution according to the reservoir characteristic information input by the input unit Data processing department; And an output unit that displays result information including the type of oil well completion method predicted by the data processing unit and the type of hydraulic fracturing solution.

In addition, the oil well completion method and the hydraulic fracturing fluid selection method of the artificial neural network-based shale gas storage layer of the present invention are applied to the shale gas storage layer characteristic information that affects the oil well completion method and the hydraulic fracturing fluid selection when designing the hydraulic fracturing method for shale gas production. Constructing a shale gas reservoir characteristic information DB that provides empirical information about; Constructing an artificial neural network capable of correlating the contribution to the result information including the type of oil well completion method and the type of hydraulic fracturing according to the reservoir characteristic information and the reservoir characteristic information: the artificial neural network type and water pressure of the well completion method Training the artificial neural network by a backpropagation learning algorithm using the shale gas reservoir characteristic information DB to predict result information including a type of crushing liquid; And predicting result information by the artificial neural network.

1 is a block diagram of an artificial neural network-based well completion method and a hydraulic fracturing fluid selection system according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating the selection of a well completion method established for the development of an oil well completion method and a hydraulic fracturing solution selection system for an artificial neural network-based shale gas storage layer of the present invention, and FIG. 3 is an artificial neural network based shale gas storage layer of the present invention. It is a flow chart of hydraulic fracturing solution selection, which was established to develop a well completion method and hydraulic fracturing solution selection system.

Referring to Figure 1, the artificial neural network-based well completion method and hydraulic fracturing fluid selection system according to an embodiment of the present invention includes an input unit 100, a data processing unit 200 and an output unit 300.

As shown, the input unit 100 is a component for inputting the shale gas reservoir characteristic information, as shown in FIG. 1, the input value 110 and the hydraulic fracturing liquid which are characteristic information influencing the selection of the well completion method. The input value 120, which is characteristic information affecting the data, may be classified and input. Here, the input value 110, which is characteristic information influencing the selection of the well completion method, is the reservoir characteristic information including geological characteristic information, mineralogical characteristic information, and petrochemical characteristic information of the shale gas reservoir input to the input unit 100. It may be one or more selected from the group consisting of depth, thickness, total carbon content, quartz content, gas content, gas adsorption amount, and clay content of the reservoir, and input values (120) that are characteristic information affecting the selection of hydraulic fracturing fluid (120) ) May be one or more selected from the group consisting of depth, pressure change according to depth change, temperature standard, and brittleness index when shale gas is developed.

As described above, the characteristic information input to the input unit 100, that is, the input values 110 and 120, collects geological, mineralogical, and rock-mechanical data of a region requiring hydraulic fracturing for shale gas development, and based on this, each region It is a measurement value input to the input layer of the artificial neural network so as to determine the well completion method and the type of hydraulic fracturing solution used in the artificial neural network included in the data processing unit 200 to be described later according to the characteristics of the.

The data processing unit 200 is predicted according to the input value 110, which is characteristic information influencing the selection of the well completion method input to the input unit 100, and the input value 120, which is characteristic information influencing the selection of hydraulic fracturing solution. It includes an artificial neural network that derives result information such as a well completion method and a type of hydraulic fracturing solution.

Here, the artificial neural network is a multi-layer perceptron (MLP) that implements information processing in the form of a network that connects a large number of simple function processors using a model of a human neural tissue as a model. An input layer having an input node; An output layer having one or more output nodes representing result information based on the storage layer characteristic information; And two or more hidden layers having at least one hidden node having processing elements for storage layer characteristic information and result information between the input layer and the output layer.

The input layer is a set consisting of 7 input nodes including depth, thickness, total carbon content, quartz content, gas content, gas adsorption content, and clay content, which are property information that affects the selection of the well completion method among the property information of the reservoir layer. When developing shale gas, which is characteristic information that affects the selection of hydraulic fracturing solution, it includes set 2 consisting of 4 input nodes including depth, pressure change according to depth change, temperature standard, and brittleness index. Characteristic information input to the input layer may be input by the input unit 100.

The output layer is a processing element representing result information based on the storage layer characteristic information input to the input layer, and includes a set of one or more output nodes, and the hidden layer is a processing element for storage layer characteristic information and result information between the input layer and the output layer. It includes a set consisting of one or more hidden nodes having, at least two hidden layers may be formed.

As such, when an artificial neural network consisting of an input layer, a hidden layer, and an output layer is executed, the input value 110, which is characteristic information influencing the selection of the well completion method, and the input value 120, which is characteristic information influencing the selection of hydraulic fracturing solution, The processing elements or nodes are interconnected so that the relationship between the predicted well completion method selection well output method 310 and the predicted hydraulic fracturing fluid selection hydraulic pressure selection output value 320 can be simply calculated. It is done.

The artificial neural network is set to assign a connection weight calculated by an empirically trained model to at least one of the input node of the input layer and the hidden node of the hidden layer and between the hidden node of the hidden layer and the output node of the output layer. As, the product of the connection weight between the input value input to the input layer and the hidden layer is added to the neuron of the hidden layer, and the inputted input value is output through a non-linear activation function in the neuron to output a different value from the input value. The product of the connection weights between the output layers is added to perform a process of transferring to other neurons of the other hidden layer or output layer. In this case, the process of revising the connection weight value in the neural network repeatedly using a large amount of input and output database is defined as learning or training. Here, learning through updating the connection weight value may be performed by a backpropagation learning algorithm.

For learning or training of the artificial intelligence network, the data processing unit 200 may further include a shale gas reservoir characteristic information DB. The shale gas reservoir characteristic information DB 210 is a component for providing a number of empirical information for training an artificial neural network, and the well is completed when designing a hydraulic fracturing method for shale gas production through a survey of known physical properties of the shale gas reservoir. It is constructed with a number of reservoir properties information sets, including geological properties, mineralogical properties and rock mechanics properties that have been determined to influence the choice of method and hydraulic fracturing solution. Here, the material data survey of known shale gas reservoirs was conducted by gathering various information such as academic literature and actual developed developer fees, including papers on known sales gas reservoirs, including the five largest shale gas reservoirs in the United States. Among the characteristic information of the shale gas reservoir, the input value (110), which is characteristic information that affects the selection of the well completion method, and the input value (120), which is characteristic information that affects the selection of hydraulic fracturing liquid, are used to determine the properties of the proposed North American shale gas reservoir. It was determined through simulation of various conditions while considering the well completion method and hydraulic fracturing solution selection chart.

The data processing unit 200 performs regression analysis using a data set that is not used for learning among datasets stored in the shale gas reservoir characteristic information DB, which is built to confirm feasibility, as described later. The output value can be verified.

The artificial neural network configured as described above may be trained in advance to mimic the selection characteristics according to the type of hydraulic well completion method and hydraulic fracturing solution predicted according to the variation of the input value.

When the artificial neural network is trained, the weight included in the connection line between the processing elements is information related to the relationship between the characteristic information of the sale gas reservoir (input value) and the result information (output value) including the type of well completion method and the type of hydraulic fracturing solution. Will have Data input to the input layer of the artificial neural network is input by the input unit 100. The result calculated by the artificial neural network is output through the output unit 300. The data processed by the artificial neural network may be stored in the characteristic information DB or a separate data unit.

The output unit 300 is a component showing result information including the type of the well completion method predicted by the data processing unit 200 and the type of hydraulic fracturing solution, as shown in FIG. 1, the type of well completion method The oil well completion method selection output value 310 and the hydraulic fracturing liquid selection output value 320 indicating the type of hydraulic fracturing liquid may be classified and output.

Here, the output value 310 for selecting a well completion method is result information indicating the type of a well completion method, which is one of a single-stage vertical well, a multi-stage vertical well, and a horizontal well. Referring to FIG. 2, the selection of the well-complete method is influenced. Depth (111), thickness (112), total carbon content (113), quartz content (114), gas content (115), gas adsorption amount (116), clay content (117) that affect the input value (110) According to the oil well completion method output value 310, it can be seen that the single-stage vertical well 311, the multi-stage vertical well 312, and the horizontal well 313 are output.

In addition, the hydraulic fracturing fluid selection output value 320 is result information indicating the type of hydraulic fracturing fluid, which is any one of carbon dioxide foam including propane, nitrogen foam including propane, and water-based fracturing fluid. When developing shale gas, which is an input value (120) that affects the selection of hydraulic fracturing solution, carbon dioxide foam including propane (depth 121), pressure change amount (122) according to the depth change, and output value (320) according to the temperature standard (123) (321), it can be seen that the nitrogen foam 322 containing propane, water-based crushing solution 323 is determined. Particularly, when the type of the hydraulic fracturing fluid is predicted to be a water-based hydraulic fracturing fluid, the data processing unit 200 includes water 324 containing a chemical additive, a linear gel 325 ), can be additionally classified as a crosslinked gel 326, and finally, in accordance with the brittleness index 124 in the additionally classified state, any one hydraulic fracturing solution is determined and displayed as result information through the output unit 300 Can. Here, brittleness is a property that is rarely accompanied by plastic deformation when a load is applied to the object, and the brittleness index (124) is expressed in relation to Young's modulus, Poisson's ratio, and porosity in terms of rock mechanics. From a mineralogical point of view, it may be a value obtained by dividing the quartz content by adding the quartz content, clay content, and carbonate content.

Next, an artificial neural network based well completion method and a prediction method of a hydraulic fracturing solution selection system according to the present invention having the above-described configuration will be described in detail.

First, since the artificial neural network-based well completion method and the hydraulic fracturing fluid selection system of the present invention form an artificial neural network by a back propagation method, an oil well is designed when a hydraulic fracturing method for shale gas production is designed to obtain empirical information for training the artificial neural network. A step of constructing a shale gas reservoir characteristic information DB that provides empirical information about the shale gas reservoir characteristic information affecting the completion method and the selection of the hydraulic fracturing solution may be performed. As an embodiment, the shale gas reservoir characteristic information DB includes input values 110, which are characteristic information influencing selection of a well completion method determined through simulation of various conditions, and input values, which are characteristic information influencing the selection of hydraulic fracturing solution ( 120).

When the shale gas reservoir characteristic information DB is constructed, an artificial neural network capable of correlating the contribution to the resulting information including the reservoir characteristic information and the type of oil well completion method and the type of hydraulic fracturing fluid according to the reservoir characteristic information is constructed. As an embodiment, the artificial neural network includes one or more input nodes for inputting the input value 110 affecting the selection of the well completion method, which is the characteristic information of the reservoir layer, and the input value 120 affecting the selection of the hydraulic fracturing solution. An artificial neural network may be configured as an output layer including at least two hidden layers including an input layer and one or more hidden nodes for each measurement value, and an output layer including one or more output nodes for deriving a result according to the input value. have. The weight set based on the shale gas reservoir characteristic information DB is given between the input node and the hidden node, and between the hidden node and the output node. The nodes of each neural network are simply sums of the weights from the preceding nodes, and can be expressed by a nonlinear output function. Therefore, the steps of constructing the artificial neural network include single-stage vertical, multi-stage vertical, and horizontal wells for the storage layer characteristic information related to the depth, thickness, total carbon content, quartz content, gas content, gas adsorption, and clay content of the reservoir layer. Associating any one kind of well completion method; And carbon dioxide foam containing propane, nitrogen foam containing propane, and water-based crushing liquid for depth information related to depth, pressure change according to depth change, temperature standards, and brittleness index It may be performed, including the step of associating the type of hydraulic fracturing solution.

When the artificial neural network is completed, the artificial neural network trains the artificial neural network by a backpropagation learning algorithm using a shale gas reservoir characteristic information DB so as to predict result information including a type of well completion method and a type of hydraulic fracturing solution. You can do it like this: The backpropagation learning algorithm means a method of overcoming errors by performing supervised learning through a correlation between input data and output data in an algorithm composed of multiple layers. That is, it means a method of calculating an output according to an input, calculating an error between a calculated output value and a desired output value, and then assigning weights to reduce the error. According to this backpropagation learning algorithm, it is possible to derive the desired output value through multiple error adjustments.

In performing artificial neural network learning, the accuracy varies depending on the number of hidden layers and neurons, and since there is no general rule between the number of nodes and layers and the accuracy, most of the trial and error methods allow the trainer to select the structure of the neural network subjectively. . Therefore, in the present invention, in order to design an optimal system, a sensitivity analysis using a Bayesian regularization algorithm, which is a learning algorithm, and Levenberg-Marquart was performed to select the number of hidden layers and to design an optimal algorithm.

Here, the Bayesian regularization algorithm used in the well-completion method has the disadvantage of relatively long learning time because learning is not terminated before the number of iterations, but it is over-fitting than other learning algorithms through the regularization process. Because it shows the correct predictive power by generating less, it is mainly used for learning neural networks with a large number of specific factors, and is most suitable as a learning algorithm for well-finished systems with many specific factors.

In addition, the Levenberg-Marquart algorithm used in the hydraulic fracturing system is one of the neural network's gradient-based learning algorithms, showing a fast convergence rate and a low mean squared error (MSE), which has a small number of factors affecting it. It is most suitable as a learning algorithm for hydraulic fracturing systems.

As an embodiment, the step of training the artificial neural network may include repeatedly learning to predict the type of a well completion method by determining the contribution of each variable using a Bayesian regularization algorithm; And iterative learning to predict the type of hydraulic fracturing by determining the contribution of each variable using the Levenberg-Marquart algorithm, wherein the contribution of each variable is based on empirical information and error correction provided from the reservoir characteristic information DB. It can be determined by assigning the basis connection weight, and the weight adjustment is performed multiple times to ensure that the connection weight is set appropriately. Here, the connection weight value is the input value 110 that affects the selection of the well completion method input to the input node of the input layer and the input value 120 that affects the selection of the hydraulic fracturing fluid type and the type of hydraulic fracturing solution. It means the weight given between each node to calculate how it affects the result information including and to reduce the error with the desired result.

It is possible to construct the network that has the minimum test error by repeatedly training the artificial neural network at various times to form the most optimal model set. That is, the relationship between the input value and the output value can be generalized as an independent sample by an optimal model set.

Figure 4 shows the results of analyzing the sensitivity of the target value and the output value of the artificial neural network-based well completion method and the hydraulic fracturing solution selection system using the Bayesian regularization method using 1200 data sets in the constructed shale gas reservoir characteristic information DB. It is a graph. As a result of system sensitivity analysis, the correlation coefficient of the Bayesian regularized learning algorithm is 0.91, which confirms that the output value of the system matches the target value well.

5 is a schematic diagram of an artificial neural network structure for selecting an oil well completion method applied to an artificial neural network based oil well completion method and a hydraulic fracturing fluid selection system according to the present invention. In order to construct an artificial neural network system using a Bayesian regularized learning algorithm, the number of hidden layers and the number of neurons with the highest accuracy were selected by changing the number of hidden layers and the number of neurons. In order to select the optimal number of hidden layers and the number of neurons, the structure with the highest accuracy was selected through a trial and error method. As a result of selection, 7 input neurons exist due to 7 input variables, and 12 neurons each exist in 2 hidden layers, and 1-12 neurons exist in the output layer. .

6 is a result of analyzing the sensitivity of the target value and the output value of the artificial neural network based well completion method and the hydraulic fracturing solution selection system using the Levenberg-Marquart method using 1200 data sets in the constructed shale gas reservoir characteristic information DB. It is a graph showing. As a result of system sensitivity analysis, the correlation coefficient of Levenberg-Marquart is 0.92, which confirms that the output value of the system matches the target value well.

7 is a schematic diagram of a structure of an artificial neural network for selecting hydraulic fracturing fluid applied to an artificial neural network based oil well completion method and a hydraulic fracturing fluid selection system according to the present invention. To construct an artificial neural network system using the Levenberg-Marquart learning algorithm, an artificial neural network module with the highest accuracy was designed using a trial and error method by changing the number of hidden layers and neurons. As a result of design, 3 input neurons exist due to 3 input variables, 8 and 3 neurons exist in 2 hidden layers, and 3-8-3-1 multi-layer artificial neural network with 1 neuron in the output layer. Made up.

When the training of the artificial neural network is completed, through the verification data set that is not used for training, through the input of factors affecting the well completion method and the hydraulic crushing solution during hydraulic crushing as in training, and through the output of the well completion method and hydraulic crushing solution The trained neural network is verified. In the embodiment of the present invention, a back propagation artificial neural network is used. In addition, various techniques of the artificial neural network may be used.

In order to verify the validity of the system, in the case of well completion, 1020 and 180 of the 1200 learning materials are used to determine the learning data and usefulness, respectively, and the validity is verified by comparing the output values of the simulation with the target values. do. In addition, in the case of the selection of hydraulic fracturing fluid, 600 learning materials are separated into 420, 90, and 90 learning data, verifiable data, and usefulness data, respectively. Validation is performed through the comparison result.

As a result of verifying the validity of the result information for the type of well completion method, it was confirmed that the output value and the target value matched well except for a few cases. Among these, ??0.33 and 0.33 values generate some noise, but this is only a small part of 1200 data, so it is considered that the impact on the reliability of the developed system is insignificant. In addition, as a result of learning the artificial neural network, the mean square error was 0.004, and after predicting the accuracy of the neural network after training using training and usefulness judgment data, all showed high correlation coefficient of 0.99 or higher.

In the case of selecting the hydraulic fracturing solution that was developed using the Levenberg-Marquart learning algorithm, the validity was verified by matching the output value and the target value of the simulation. As a result of comparing the output values of the simulation and the target values, the output values matched well with the target values, except for a few cases, or the difference between the output values and the target values existed within the allowable category. As a result of learning the artificial neural network, the mean square error showed a relatively low value of 0.0114, and as a result of regression analysis on the prediction accuracy of the neural network after training using verification and usability judgment data, all showed high correlation coefficient of 0.97 or higher. In particular, the verification showed a high correlation of 0.99 or more, confirming that the accuracy of the artificial neural network model was high.

Therefore, if the artificial neural network included in the artificial neural network-based well completion method and the hydraulic fracturing solution selection system of the present invention is trained, the input value 110 affecting the selection of the well completion method and the input value affecting the hydraulic fracturing solution selection Changes in the result information including the type of oil well completion method and the type of hydraulic fracturing according to the change in (120) can be approximated with the actual experimental values, and the type and water pressure of oil well completion method can be input only by inputting the reservoir characteristics information by the trained artificial neural network. Result information including the type of crushing liquid can be predicted.

FIG. 8 illustrates a graphical user interface (GUI) using matlab software to facilitate user access to an artificial neural network-based well completion method and hydraulic fracturing solution selection system according to an embodiment of the present invention. As an example, the upper part of the well completion module panel is designed to input specific factors of the reservoir, and the artificial neural network structure applied to the system is inserted in the center left. In addition, it is designed by arranging the panel showing the most suitable well completion method in the center right of the figure as a result of the simulation, and placing the panel output as a letter at the bottom.

9 is another embodiment of constructing a graphical user interface of a hydraulic crushing fluid selection system of an artificial neural network based oil well completion method and a hydraulic crushing fluid selection system according to an embodiment of the present invention. When the output value is derived using an artificial neural network as in the selection system, the output value is designed to output in the same structure as the well completion method when the output value is nitrogen foam or carbon dioxide foam.

10 and 11 are panels that appear when a water-based crushing liquid is derived as an output value of the output unit. FIG. 10 is a panel that can use petrochemical factors to select a hydraulic fracturing solution such as water, a linear gel, and a crosslinked gel containing chemical additives by calculating the brittleness using Young's modulus and Poisson's ratio, and FIG. 11 Is a panel that can display the hydraulic fracturing solution of water, linear gel, and crosslinked gel containing chemical additives as an output value using a formula consisting of quartz, clay mineral, and mineral acid content from a geological and mineralogical point of view.

12 is another result applied to an oil well completion method selection GUI program constructed using field data in an artificial neural network based oil well completion method and a hydraulic fracturing fluid selection system according to an embodiment of the present invention, and the US Horn River as a well completion module The oil well completion method derived by inputting the properties of the shale reservoir is represented by pictures and letters.

13 is another result applied to a hydraulic fracturing fluid selection GUI program constructed using field data in an artificial neural network-based well completion method and a hydraulic fracturing fluid selection system according to an embodiment of the present invention. The hydraulic crushing fluid selection module shows the primary output value of water-soluble hydraulic crushing fluid as shown in the left side of the middle panel through the storage layer properties such as depth, and the results of inputting the contents of quartz, carbonate minerals, and clay minerals are shown in pictures and characters.

12 and 13, it can be seen that the accuracy of matching the result with the predicted value is high, so it can be confirmed that the artificial neural network-based well completion method and the hydraulic fracturing solution selection system of the present invention have excellent field usability and suitability. In addition, these results, if using the artificial neural network-based well completion method and hydraulic fracturing fluid selection system of the present invention, as well as experts, non-experts can select an appropriate well completion method and hydraulic fracturing fluid when designing the hydraulic fracturing method if only shale reservoir characteristics are provided. It shows that there is.

The present invention has been illustrated and described with reference to preferred embodiments, as described above, but is not limited to the above-described embodiments, and to those skilled in the art to which the present invention pertains without departing from the spirit of the present invention. Various changes and modifications will be possible.

100: input unit 200: data processing unit
300: output

Claims (16)

When designing the hydraulic fracturing method for the production of shale gas, to enter the reservoir characteristics information including geological properties information, mineral properties information, and petrochemical properties information of the shale gas reservoir affecting the selection of the hydraulic well completion method and the hydraulic fracturing solution. Inputs;
Performing learning and validity check through updating the connection weight value through an artificial neural network set to predict result information including the type of oil well completion method and the type of hydraulic fracturing solution according to the reservoir characteristic information input by the input unit Data processing department; And
An output unit showing result information including the type of the well completion method predicted by the data processing unit and the type of hydraulic fracturing fluid; an artificial neural network-based shale gas reservoir comprising a well completion method and a hydraulic fracturing fluid selection system.
According to claim 1,
The data processing unit is a geological property, mineralogical property, and petromechanical property determined to influence the oil well completion method and the selection of a hydraulic crushing solution when designing a hydraulic fracturing method for shale gas production through data analysis of the physical properties of a known shale gas reservoir. A well-established method of artificial neural network based shale gas reservoir comprising a shale gas reservoir characteristic information DB which is constructed with a plurality of reservoir characteristic information sets including a plurality of empirical information for training the artificial neural network. And hydraulic fracturing solution selection system.
According to claim 1,
The artificial neural network may include an input layer having a plurality of input nodes associated with the storage layer characteristic information; An output layer having at least one output node representing the result information based on the storage layer characteristic information; And an at least two hidden layer having at least one hidden node having processing elements for storage layer characteristic information and result information between the input layer and the output layer, and the well completion method of an artificial neural network based shale gas storage layer. Hydraulic fracturing solution selection system.
The method of claim 3,
The artificial neural network is configured to assign a connection weight calculated by an empirically trained model to at least one of the input node and the hidden node and between the hidden node and the output node. Oil well completion method of gas reservoir and hydraulic fracturing solution selection system.
According to claim 1,
The training through the update of the connection weight value is performed by a backpropagation learning algorithm, and the well completion method of the artificial neural network based shale gas reservoir and the hydraulic fracturing fluid selection system.
The method according to any one of claims 1 to 5,
At least one selected from the group consisting of the depth, thickness, total carbon content, quartz content, gas content, gas adsorption amount, and clay content of the reservoir layer, which affects the prediction of the type of oil well completion method An artificial neural network based oil well completion method and hydraulic fracturing solution selection system based on an artificial neural network.
The method of claim 6,
The type of the well completion method represented by the result information in the output unit according to the characteristic information is one of a single-stage vertical well, a multi-stage vertical well, and a horizontal well. And hydraulic fracturing solution selection system.
The method according to any one of claims 1 to 5,
Among the characteristic information of the reservoir layer, the characteristic information influencing the selection of the type of hydraulic fracturing fluid is characterized in that at least one selected from the group consisting of depth, pressure change according to depth change, temperature standard, and brittleness index during shale gas development. The artificial neural network based oil well completion method and hydraulic fracturing fluid selection system for shale gas reservoirs.
The method of claim 8,
According to the characteristic information, the type of the hydraulic fracturing solution represented by the result information on the output unit is an artificial neural network characterized by being one of carbon dioxide foam containing propane, nitrogen foam containing propane, and water-based fracturing fluid. Oil well completion method and hydraulic fracturing solution selection system for the base shale gas reservoir.
The method of claim 9,
When the type of the hydraulic fracturing solution is a water-based hydraulic fracturing solution, it is further classified into water containing a chemical additive, a linear gel, and a crosslinked gel, and any one determined according to the brittleness index in the additional classified state appears as the result information. An artificial neural network based oil well completion method and hydraulic fracturing solution selection system based on an artificial neural network.
The method of claim 10,
The brittleness index is a relationship between Young's modulus, Poisson's ratio, porosity or quartz content divided by the sum of quartz content, clay content and carbonate content, and the well completion method and hydraulic crushing of an artificial neural network based shale gas reservoir. Liquid selection system.
When designing the hydraulic fracturing method for shale gas production, constructing a shale gas reservoir characteristic information DB that provides empirical information about the well completion method and the characteristic of shale gas reservoir characteristics that affect the selection of the hydraulic fracturing solution;
Constructing an artificial neural network capable of correlating the contribution to the result information including the type of oil well completion method and the type of hydraulic fracturing solution according to the reservoir characteristic information and the reservoir characteristic information:
Training the artificial neural network by a backpropagation learning algorithm using the shale gas reservoir characteristic information DB so that the artificial neural network predicts result information including a type of well completion method and a type of hydraulic fracturing fluid; And
Prediction of the result information by the artificial neural network; Oil well completion method of the artificial neural network-based shale gas reservoir comprising a method of selecting a hydraulic fracturing solution.
The method of claim 12,
The step of predicting the result information by the artificial neural network includes the step of performing learning and feasibility check through the updating of the connection weight value of the artificial neural network, and the well completion method and hydraulic pressure of the artificial neural network based shale gas reservoir. How to select crushing solution.
The method of claim 12,
The step of constructing the artificial neural network includes any one of a single-stage vertical tablet, a multi-stage vertical tablet, and a horizontal tablet for the storage layer characteristic information related to the depth, thickness, total carbon content, quartz content, gas content, gas adsorption amount, and clay content of the reservoir layer. Associating one kind of well completion method; And carbon dioxide foam containing propane, nitrogen foam containing propane, and water-based crushing liquid for depth information related to depth, pressure change according to depth change, temperature standards, and brittleness index Oil well completion method of artificial neural network based shale gas reservoir and method of selecting hydraulic fracturing liquid, characterized in that it is carried out including the step of associating the type of hydraulic fracturing liquid.
The method of claim 12,
The step of training the artificial neural network may include repeatedly learning to predict the type of well completion method by determining the contribution of each variable using a Bayesian regularization algorithm; And Leverberg-Marquart algorithm to determine the contribution of each variable to iteratively learn to predict the type of hydraulic fracturing fluid; and an oil well completion method and a hydraulic fracturing fluid based on artificial neural network based shale gas reservoir. Selection method.
The method of claim 15,
The contribution of each variable is determined by assigning a connection weight based on empirical information and error correction provided from the reservoir characteristic information DB. The connection weight is based on an artificial neural network characterized by performing multiple weight adjustments to be properly set. Oil well completion method of shale gas reservoir and hydraulic fracturing method selection method.

Images