A prediction method for hydrocarbon drainage potential in in-situ mining of shale by heat injection based on neural network

Abstract

The present invention provides a method for predicting the hydrocarbon drainage potential of in-situ shale mining by heat injection based on a neural network, belonging to the technical field of predicting the development effect of oil fields. The method for predicting the hydrocarbon drainage potential of in-situ shale mining by heat injection based on a neural network includes the following steps: collecting static and dynamic data of the reservoir in the target area, and extracting the influencing factors of in-situ mining of heat injection fluid in the shale reservoir; establishing a numerical simulation model for in-situ heat injection and mining of heat injection fluid in the shale that conforms to the target block according to the influencing factors; generating simulation schemes by using the Latin hypercube sampling method, and combining with the numerical simulation model for in-situ heat injection and mining of heat injection fluid in the shale to construct a basic sample database; optimizing the basic sample database by combining the 3σ principle and the correlation analysis method; using the neural network algorithm to construct a prediction model for the hydrocarbon drainage potential of in-situ heat injection and mining of heat injection fluid in the shale.

Description

Technical Field

[0001] The present invention belongs to the technical field of oilfield development effect prediction. Specifically, it relates to a method for predicting the hydrocarbon expulsion potential of in-situ shale mining by heat injection based on a neural network. Background Art

[0002] The success of the shale oil and gas revolution in the United States has promoted China's exploration of shale oil and gas. China has a huge amount of organic-rich shale resources, and significant progress has been made in the exploration and development of medium to high maturity shale oil. Commercial exploitation can be achieved by combining horizontal wells and multi-stage fracturing technologies. However, even with mature horizontal well and artificial fracturing technologies, it is impossible to achieve large-scale exploitation of continental immature and medium to low maturity shale oil. In-situ heating exploitation will become the main technology for its commercial utilization. According to different heating methods, in-situ exploitation technologies are mainly divided into conductive heating, convective heating, and radiative heating, etc. Existing research has proved that convective heating of heat injection fluid is an effective method for shale oil exploitation. However, the research on hydrocarbon generation and expulsion during in-situ shale heat injection exploitation is not deep enough, and there is no method to quickly and accurately evaluate the hydrocarbon expulsion potential of in-situ shale exploitation.

[0003] In most of the existing technologies at present, shale pyrolysis experiments are carried out based on powders, granules or shale rock columns. The scale of the pyrolysis experiment has a great influence on the pyrolysis of shale organic matter to generate and expel hydrocarbons, which makes it difficult to accurately explain the technical problems of the in-situ exploitation mechanism of shale at the field scale with the existing experimental results. Summary of the Invention

[0004] In view of this, the present invention provides a method for predicting the hydrocarbon expulsion potential of in-situ shale mining by heat injection based on a neural network, which can solve the technical problems that in most of the existing technologies, shale pyrolysis experiments are carried out based on powders, granules or shale rock columns, and the scale of the pyrolysis experiment has a great influence on the pyrolysis of shale organic matter to generate and expel hydrocarbons, which makes it difficult to accurately explain the in-situ exploitation mechanism of shale at the field scale with the existing experimental results.

[0005] The present invention is implemented as follows:

[0006] The present invention provides a method for predicting the hydrocarbon expulsion potential of in-situ shale mining by heat injection based on a neural network, which includes the following steps:

[0007] S10: Collect static and dynamic data of the reservoir in the target area, and extract the influencing factors of in-situ exploitation of heat injection fluid in the shale reservoir;

[0008] S20: Based on the influencing factors, establish a numerical simulation model for in-situ exploitation of heat injection fluid in shale that conforms to the target block;

[0009] S30: Use the Latin hypercube sampling method to generate simulation schemes, and combine with the numerical simulation model for in-situ exploitation of heat injection fluid in shale to construct a basic sample database;

[0010] S40: Optimize the basic sample database by combining the 3σ principle and the correlation analysis method;

[0011] S50: Use the neural network algorithm to build the prototype of the prediction model for the hydrocarbon drainage potential of in-situ exploitation of shale by injecting heat fluid;

[0012] S60: Test the performance of the built prototype of the prediction model, optimize the prototype of the prediction model for the hydrocarbon drainage potential of in-situ exploitation of shale by injecting heat fluid, and obtain the prediction model for the hydrocarbon drainage potential of in-situ exploitation of shale by injecting heat fluid.

[0013] Based on the above technical solutions, a method for predicting the hydrocarbon drainage potential of in-situ exploitation of shale based on neural network of the present invention can also be improved as follows:

[0014] Among them, the static data of the target area reservoir includes the reservoir characteristics, fluid properties and field monitoring of the target block; the dynamic data of the target area reservoir includes the oil testing and production history of the reservoir.

[0015] Furthermore, the influencing factors of in-situ exploitation of shale reservoir by injecting heat fluid include the average porosity of the reservoir, average permeability, natural fracture spacing, initial kerogen concentration, heat injection rate, heat injection temperature and production pressure.

[0016] Among them, the specific steps of establishing a numerical simulation model for in-situ exploitation of shale by injecting heat fluid in line with the target block according to the influencing factors are as follows:

[0017] Use the CMG-STARS module to establish a numerical simulation model, including a geological model, a pyrolysis reaction kinetics model and a numerical simulation model;

[0018] The input parameters for establishing the geological model include reservoir burial depth, average thickness, reservoir temperature, reservoir pressure, rock volume heat capacity, rock thermal conductivity, average porosity of the reservoir, average permeability, natural fracture spacing and initial kerogen concentration;

[0019] The input parameters for establishing the pyrolysis reaction kinetics model include pyrolysis reaction equations, reaction order, reaction frequency factor and activation energy;

[0020] The input parameters for the numerical simulation model include well pattern well spacing, the number of injection wells and production wells, well locations, heat injection rate, heat injection temperature and production pressure.

[0021] Among them, the specific steps of generating simulation schemes by using the Latin hypercube sampling method and constructing a basic sample database in combination with the numerical simulation model for in-situ exploitation of shale by injecting heat fluid are as follows:

[0022] A large number of numerical simulation schemes are generated using the Latin hypercube sampling method. All the simulation schemes are run in combination with the numerical simulation software CMG, and the hydrocarbon production at 1500 d of in-situ steam injection is extracted as the target response; the hydrocarbon production is the equivalent of the produced oil, and 975 sm 3 hydrocarbon gas is defined as the unit of oil equivalent. The hydrocarbon production is calculated by comprehensively considering the produced oil volume and the produced hydrocarbon gas volume, and the basic sample database is jointly composed of influencing factors and corresponding response values.

[0023] Furthermore, the specific operation steps for optimizing the basic sample database by combining the 3σ principle and the correlation analysis method are as follows:

[0024] The 3σ principle is used to detect outliers in the sample database, that is, abnormal data is removed according to the mean value μ and the standard deviation σ; the abnormal data is all data values whose deviation from the mean value exceeds 3.0σ; if all data in the sample database appear within the interval [μ - 3.0σ, μ + 3.0σ], there are no outliers in the basic sample database.

[0025] Among them, the specific steps for constructing the prototype of the prediction model for the hydrocarbon production potential of in-situ heating of shale by hot fluid using the neural network algorithm are as follows:

[0026] S51: Divide the model training set and test set according to a ratio, determine the input and output data sets, and perform normalization processing; based on the input and output characteristics of the model, determine the number of nodes in the network input layer and output layer;

[0027] S52: Set the neural network structure and parameters, including the training function, activation function, number of hidden layers and hidden neurons, maximum number of training times, learning rate, and target error;

[0028] S53: Initialize the connection weights between the layers of the neural network, as well as the thresholds of the hidden layer and the output layer;

[0029] S54: Calculate the outputs of the hidden layer and the output layer, and calculate the network prediction error based on the output value of the output layer and the expected output value;

[0030] S55: Update the network connection weights and thresholds based on the network training error;

[0031] S56: Determine whether the algorithm iteration reaches the termination condition; if satisfied, the prediction model for the hydrocarbon production potential of in-situ shale mining is obtained, otherwise, return to step S54.

[0032] Furthermore, the termination condition is reaching the maximum number of training times or reaching the target training error.

[0033] Furthermore, the calculation of the hidden layer is as follows:

[0034]

[0035] Among them, y is the normalized input information; n and l are the numbers of nodes in the input layer and the hidden layer respectively; υ ij is the connection weight between the i-th neuron in the input layer and the j-th neuron in the hidden layer; θ is the hidden layer threshold; 1 is the hidden layer activation function.

[0036] Furthermore, the output layer is calculated as:

[0037]

[0038] Among them, m is the number of nodes in the output layer; ω jk is the connection weight between the j-th neuron in the hidden layer and the k-th neuron in the output layer; γ is the output layer threshold; f2 is the output layer activation function.

[0039] Compared with the prior art, the beneficial effects of a method for predicting hydrocarbon drainage potential in in-situ shale heating extraction based on a neural network provided by the present invention are as follows: Based on the characteristics of typical shale reservoirs, the present invention establishes a numerical simulation model for in-situ heating extraction of shale by injecting heat-carrying fluid, realizing in-situ conversion of shale by injecting heat at the mine scale, and truly simulating the process of heat-carrying fluid heating and displacement in shale; Based on the simulation results of a large number of in-situ shale heating extraction schemes, a basic sample database is constructed for neural network training and testing, enabling the prediction results of the neural network model to more accurately explain the actual hydrocarbon drainage effect of in-situ development of shale by injecting heat-carrying fluid; The present invention establishes a model for predicting hydrocarbon drainage potential in in-situ shale heating extraction based on a neural network algorithm, which can quickly and efficiently evaluate the hydrocarbon drainage potential of different in-situ shale heating extractions by injecting heat-carrying fluid, thereby providing certain theoretical guidance for the large-scale exploitation of shale. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 is the flowchart of the steps of the present invention;

[0042] Figure 2 is the flowchart of the steps of S50 in the present invention;

[0043] Figure 3 is the shale numerical simulation model diagram in an embodiment of the present invention;

[0044] Figure 4 is the histogram of the data distribution of each variable in the sample library in an embodiment of the present invention;

[0045] Figure 5 It is a correlation coefficient matrix diagram between characteristic factors based on the Pearson correlation analysis method in an embodiment of the present invention;

[0046] Figure 6 It is a prediction result diagram of the training set and the test set based on the optimal neural network structure in an embodiment of the present invention. Detailed implementation manners

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0050] As Figure 1 shown, it is a step flow chart of a method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network provided by the present invention, including the following steps:

[0051] S10: Collect static and dynamic data of the target area reservoir, and extract the influencing factors of in-situ heat injection fluid exploitation in the shale reservoir;

[0052] S20: Based on the influencing factors, establish a numerical simulation model for in-situ heat injection fluid heating exploitation in shale that conforms to the target block;

[0053] S30: Use the Latin hypercube sampling method to generate simulation schemes, and combine with the numerical simulation model for in-situ heat injection fluid heating exploitation in shale to construct a basic sample database;

[0054] S40: Optimize the basic sample database in combination with the 3σ principle and the correlation analysis method;

[0055] S50: Use the neural network algorithm to construct the prototype of a prediction model for the hydrocarbon drainage potential of in-situ heat injection fluid heating in shale;

[0056] S60: Test the prototype performance of the constructed prediction model, optimize the prototype of the prediction model for the hydrocarbon drainage potential of in-situ exploitation by injecting heat fluid into shale, and obtain the prediction model for the hydrocarbon drainage potential of in-situ exploitation by injecting heat fluid into shale.

[0057] Among them, in the above technical solution, the static data of the target area reservoir includes the reservoir characteristics, fluid properties and field monitoring of the target block; the dynamic data of the target area reservoir includes the oil well testing and production history of the reservoir.

[0058] Furthermore, in the above technical solution, the influencing factors of in-situ exploitation by injecting heat fluid into shale reservoir include the average porosity of the reservoir, average permeability, natural fracture spacing, initial kerogen concentration, heat injection rate, heat injection temperature and production pressure.

[0059] Among them, in the above technical solution, the specific steps to establish a numerical simulation model for in-situ exploitation by injecting heat fluid into shale that conforms to the target block according to the influencing factors are as follows:

[0060] An embodiment of the present invention is based on the reservoir characteristic parameters of a typical shale oil reservoir in the Ordos Basin, and uses the CMG-STARS module to establish a numerical simulation model for in-situ exploitation by injecting steam into shale, including a geological model, a pyrolysis reaction kinetics model and a numerical simulation model.

[0061] Set the reservoir burial depth to 1585.34 m, the average thickness to 15 m, and the reservoir temperature and pressure to 50 °C and 10.89 MPa respectively; set the rock volume heat capacity and thermal conductivity to 5.0×10 6 J / (m 3 ·°C), 216000 J / (m·day·°C); for the setting of other input parameters, including the main influencing factors determined in step S10, namely the average porosity, average permeability, natural fracture spacing, initial kerogen concentration, and subsequent determination according to their respective parameter ranges, thus establishing a shale geological model.

[0062] The pyrolysis reaction of organic matter in shale reservoir is complex. Referring to the Braun-Burnham pyrolysis model and experimental data, it mainly includes three alternating reactions: kerogen pyrolysis, heavy oil cracking and light oil cracking. In order to simplify the pyrolysis reaction model, the pseudo-components Kerogen, Prechar, IC 37 、IC 13IC1 and IC2 represent kerogen, solid residue, heavy oil, light oil, and hydrocarbon gas respectively. The pyrolysis reaction kinetic equation is commonly described by the Arrhenius formula, as shown in Equation (1). The pyrolysis reaction kinetic model is composed of the pyrolysis reaction equation, reaction order, reaction frequency factor, activation energy, etc. The specific pyrolysis reaction kinetic parameters in an embodiment of the present invention are shown in Table 1. In the CMG-STARS module, seven components, namely CO2, H2O, Kerogen, Prechar, IC 37 、IC 13 and IC2, are defined. Then, three alternating reactions in Table 1 are added in sequence, and the pyrolysis kinetic parameters of each chemical reaction are input, thereby establishing a kinetic simulation model for shale pyrolysis reaction.

[0063]

[0064] Among them, x is the conversion rate of the reactant at temperature T; is the change rate of the reactant conversion rate with time; A is the frequency factor, s -1 ; m is the reaction order; E is the apparent activation energy, J·mol-1; R is the universal gas constant, 8.314 J·(mol·K) -1 .

[0065] Table 1 Pyrolysis reaction kinetic model parameter table

[0066]

[0067] Referring to the in-situ electric heating well pattern of Nantaohuamu, it includes two layers of hexagonal well patterns and one layer of quadrilateral well pattern, with a total of 16 injection wells, and two production wells are drilled in the center. In order to effectively restore the in-situ steam injection process for shale exploitation, in an embodiment of the present invention, one-quarter of the original reservoir is simulated using a logarithmic grid, with 6 injection wells and 1 production well set. The established numerical simulation model for in-situ exploitation of shale by steam injection heating and the well distribution are as Figure 3 shown, Figure 3 where the left side in the figure is the well layout position of in-situ electric heating in Nantaohuamu, and the right side is the plan view of the numerical simulation model for in-situ exploitation of shale by steam injection heating finally established in an embodiment of the present invention. For the setting of other input parameters, including some main influencing factors determined in step S10, namely injection temperature, injection rate, and production pressure, they are determined according to their respective parameter ranges subsequently.

[0068] Among them, in the above technical solution, the specific steps for generating a simulation scheme using the Latin hypercube sampling method and constructing a basic sample database in combination with the numerical simulation model for in-situ exploitation of shale by heat injection fluid heating are as follows:

[0069] In an embodiment of the present invention, the influencing factors and the value ranges of the influencing factors for the hydrocarbon drainage potential in the in-situ steam injection heating of shale are shown in Table 2. The Latin hypercube sampling method has the advantages of uniform stratification and comprehensiveness. Therefore, based on the selected influencing factors and value ranges, 40 numerical simulation schemes are generated using the Latin hypercube sampling method. All the simulation schemes are run in combination with the numerical simulation software CMG, and the hydrocarbon drainage volume at 1500 d of in-situ steam injection mining is extracted as the target response. Here, the hydrocarbon drainage volume is defined as the discharged oil equivalent, and it is specified that 975 sm 3 hydrocarbon gas is the unit oil equivalent, and the hydrocarbon drainage volume is calculated by comprehensively considering the discharged oil volume and the discharged hydrocarbon gas volume; the basic sample database is jointly composed of the influencing factors and the corresponding response values.

[0070] Table 2 Table of influencing factors and value ranges

[0071] Influencing factor Value taken Average porosity / % 2-10 <![CDATA[Average permeability / (10 -3 μm 2 )]]> 0.015-1 Natural fracture spacing / m 0.1-1 <![CDATA[Initial kerogen concentration / (mol·m -3 )]]> 6000-20000 <![CDATA[Steam injection rate / (m 3 ·d -1 )]]> 10-100 Steam injection temperature / °C 550-650 Production pressure / MPa 5-10

[0072] Further, in the above technical solution, the specific operation steps for optimizing the basic sample database by combining the 3σ principle and the correlation analysis method are as follows:

[0073] Histograms of the seven influencing factors and the response parameters are plotted, and it is found that they all conform to the normal distribution or approximately normal distribution. Therefore, the 3σ principle can be used to detect outliers in the sample data, that is, abnormal data are removed according to the mean value (μ) and the standard deviation (σ), and it is considered that all measured values with a deviation from the mean value exceeding 3.0σ are outliers. After analysis, all the data in the sample library fall within the interval [μ - 3.0σ, μ + 3.0σ], which means that there are no outliers in the basic sample database. The distribution histograms of each parameter are as Figure 4 shown.

[0074] Too many input variables will lead to a large number of network parameters and a slow convergence speed, and when the input factors are highly correlated, it will cause feature redundancy. Considering that each parameter conforms to the normal distribution or approximately normal distribution, in an embodiment of the present invention, the Pearson correlation analysis method is selected to determine the correlation degree between the factors to avoid information redundancy. The Pearson correlation coefficient is an index to evaluate the correlation degree between variables. The correlation coefficients between the variables in the basic sample database can be calculated using formula (2). With the help of MATLAB software, the Pearson correlation coefficients between the influencing factors are obtained, and the correlation coefficient matrix is plotted as Figure 5 shown. After analysis, it can be determined that there is no highly linear correlation relationship between the influencing factors, and the feature overlap is weak. Therefore, all the influencing factors can be retained as feature variables for the establishment of the neural network model.

[0075]

[0076] Among them, r is the data pair (a t,b t The correlation coefficient of )(t = 1, 2, …, q). The value range of the correlation coefficient r is [-1, 1]. When r < 0, it indicates that the two groups of variables a and b are negatively correlated; when r > 0, it indicates that the two groups of variables are positively correlated. And it is defined that |r| in the range of 0 - 0.3 indicates low correlation between factors, 0.3 - 0.8 belongs to medium correlation, and 0.8 - 1.0 indicates high correlation between factors.

[0077] Such as Figure 2 , in the above technical solution, using the neural network algorithm, the specific steps to construct the prototype of the prediction model for the hydrocarbon drainage potential of in-situ mining by shale hydrothermal fluid heating are as follows:

[0078] S51: Divide the model training set and test set according to a ratio, determine the input and output data sets, and perform normalization processing; based on the input and output characteristics of the model, determine the number of nodes in the network input layer and output layer;

[0079] S52: Set the neural network structure and parameters, including the training function, activation function, number of hidden layers and hidden neurons, maximum number of training times, learning speed, and target error;

[0080] S53: Initialize the connection weights between the layers of the neural network, as well as the thresholds of the hidden layer and output layer;

[0081] S54: Calculate the outputs of the hidden layer and output layer, and calculate the network prediction error based on the output value of the output layer and the expected output value;

[0082] S55: Update the network connection weights and thresholds based on the network training error;

[0083] S56: Determine whether the algorithm iteration reaches the termination condition; if satisfied, obtain the prediction model for the hydrocarbon drainage potential of in-situ shale mining, otherwise, return to step S54.

[0084] Furthermore, in the above technical solution, the termination condition is reaching the maximum number of training times or reaching the target training error.

[0085] In an embodiment of the present invention, an error backpropagation neural network is selected to construct a prediction model for the hydrocarbon drainage potential of in-situ mining by shale hydrothermal fluid heating, and its algorithm is given, specifically including:

[0086] Divide the model training set and test set according to a ratio, determine the input and output data sets, and perform normalization processing. Based on the input and output characteristics of the model, determine the number of nodes in the network input layer and output layer;

[0087] Randomly select 80% of the data in the optimized sample library as the training set and 20% as the test set. Normalize the data of the training set and the test set according to formula (3), and the original data can be mapped to the range of [0, 1]. Determine that the number of nodes in the input layer of the neural network is 7 and the number of nodes in the output layer is 1 according to the influencing factors and response parameters.

[0088]

[0089] Among them, x is the original data before normalization; y is the corresponding parameter value after normalization; x max , x min are the maximum and minimum values in the original dataset respectively; y max , y min are the maximum and minimum values of the normalized dataset respectively;

[0090] Given the neural network structure and parameters, including the training function, activation function, number of hidden layers and hidden neurons, maximum number of training times, learning rate, and target error;

[0091] Calculate the outputs of the hidden layer and the output layer. Based on the output value of the output layer and the expected output value, calculate the network prediction error;

[0092] Calculation of the hidden layer output:

[0093]

[0094] Calculation of the output layer output:

[0095]

[0096] Calculation of the network prediction error value:

[0097] e k = Y k - O k , k = 1, 2,..., m;

[0098] Among them, H is the output of the hidden layer, O is the output of the output layer, e is the network calculation error; y is the input information after normalization; n, l, m are the number of nodes in the input layer, hidden layer, and output layer respectively; v ij is the connection weight between the i-th neuron in the input layer and the j-th neuron in the hidden layer, ω jk is the connection weight between the j-th neuron in the hidden layer and the k-th neuron in the output layer; θ, γ are the thresholds of the hidden layer and the output layer respectively; f1, f2 are the activation functions of the hidden layer and the output layer respectively; Y is the expected output.

[0099] Based on the network training error, update the network connection weights and thresholds;

[0100] Weight update:

[0101]

[0102] i = 1, 2, …, n; j = 1, 2, …, l;

[0103] ω jk = ω jk + ηH j e k , j = 1, 2, …, l;

[0104] k = 1, 2, …, m;

[0105] Threshold update:

[0106]

[0107] j = 1, 2, …, l;

[0108] γ k = γ k + e k , k = 1, 2, …, m;

[0109] where η is the network learning rate.

[0110] Judge whether the algorithm iteration reaches the termination condition. If satisfied, obtain the prototype of the shale in-situ production hydrocarbon expulsion potential prediction model; otherwise, return to step S54.

[0111] Furthermore, in the above technical solution, the specific steps to test the performance of the prototype of the constructed prediction model and optimize the prototype of the shale in-situ production hydrocarbon expulsion potential prediction model by injecting heated fluid are as follows:

[0112] Set the minimum error of the training target to 10 -3 , the hidden layer uses the Sigmoid activation function, the output layer uses the linear activation function, and use the error backpropagation to update the weights and thresholds to minimize the difference between the network output and the expected output. Under the given neural network hyperparameter combination, the optimal neural network structure is often not obtained. Therefore, multiple above-mentioned neural network structures need to be established, and multiple repeated trainings are carried out. Finally, a neural network hyperparameter combination that makes the prediction effect of the test set optimal is determined, including: 2 hidden layers, the number of hidden neurons is 38 and 23 respectively, the maximum number of training times is 1000, the learning rate is 0.03, and the widely used Levenberg-Marquardt algorithm is used as the training function. The prediction results of the training set and the test set obtained under the optimal neural network structure are as Figure 6As shown, the scatter points in the cross-plot of simulated data and predicted data are all distributed near the 45° line, indicating that the established neural network model can accurately predict the hydrocarbon drainage potential of in-situ shale mining.

[0113] Select a shale reservoir porosity of 10%, a permeability of 0.15×10 -3 μm 2 , a natural fracture spacing of 0.5 m, an initial kerogen concentration of 16000 mole / m 3 , a steam injection rate of 80 m 3 / d, a steam injection temperature of 600 °C, and a production pressure of 8 MPa. Use the CMG-STARS simulation software and the shale hydrocarbon drainage potential prediction model based on neural network to calculate the produced oil equivalent to be 157.09 m 3 and 163.94 m 3 , respectively. The prediction error of the neural network model is less than 5%, verifying the accuracy of the neural network model, which can be used for predicting the hydrocarbon drainage potential of in-situ heat injection in a certain number of shale reservoirs.

[0114] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all of them should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network, characterized in that, It includes the following steps: S10: Collect static and dynamic data of the target area reservoir, and extract the influencing factors for in-situ exploitation of heat-injected fluid in shale reservoirs; S20: Based on the influencing factors, establish a numerical simulation model for in-situ exploitation of heat-injected fluid in shale by heating that conforms to the target block; S30: Use the Latin hypercube sampling method to generate simulation schemes, and combine with the numerical simulation model for in-situ exploitation of heat-injected fluid in shale by heating to construct a basic sample database; S40: Optimize the basic sample database by combining the 3σ principle and the correlation analysis method; S50: Use the neural network algorithm to construct the prototype of a prediction model for the hydrocarbon expulsion potential of in-situ exploitation of heat-injected fluid in shale by heating; S60: Test the performance of the constructed prototype of the prediction model, optimize the prototype of the prediction model for the hydrocarbon expulsion potential of in-situ exploitation of heat-injected fluid in shale by heating, and obtain the prediction model for the hydrocarbon expulsion potential of in-situ exploitation of heat-injected fluid in shale by heating; The S20 includes the following steps: Use the CMG-STARS module to establish a numerical simulation model, including a geological model, a pyrolysis reaction kinetic model, and a numerical simulation model; The parameters input for establishing the geological model include reservoir burial depth, average thickness, reservoir temperature, reservoir pressure, rock volume heat capacity, rock thermal conductivity, average reservoir porosity, average permeability, natural fracture spacing, and initial kerogen concentration; The parameters input for establishing the pyrolysis reaction kinetic model include pyrolysis reaction equations, reaction order, reaction frequency factor, and activation energy; The parameters input for the numerical simulation model include well pattern well spacing, the number of injection wells and production wells, well locations, heat injection rate, heat injection temperature, and production pressure.

2. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 1, characterized in that, The static data of the target area reservoir include the reservoir characteristics, fluid properties, and field monitoring of the target block; the dynamic data of the target area reservoir include well testing and production history of the reservoir.

3. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 2, characterized in that, The influencing factors for in-situ exploitation of heat-injected fluid in shale reservoirs include average reservoir porosity, average permeability, natural fracture spacing, initial kerogen concentration, heat injection rate, heat injection temperature, and production pressure.

4. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 1, characterized in that, The specific steps for using the Latin hypercube sampling method to generate simulation schemes and combining with the numerical simulation model for in-situ exploitation of heat-injected fluid in shale by heating to construct a basic sample database are as follows: Generate a numerical simulation plan using the Latin hypercube sampling method, run all the simulation plans in combination with the numerical simulation software CMG, and extract the hydrocarbon production at 1500 d of in-situ steam injection as the target response; the hydrocarbon production is the equivalent of produced oil, and 975 sm 3 hydrocarbon gas is used as the unit of oil equivalent, the hydrocarbon production is calculated by comprehensively considering the produced oil volume and the produced hydrocarbon gas volume, and the basic sample database is jointly composed of influencing factors and corresponding response values.

5. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 4, characterized in that, The specific operation steps for optimizing the basic sample database by combining the 3σ principle and the correlation analysis method are as follows: Use the 3σ principle to detect outliers in the sample database, that is, remove abnormal data according to the mean μ and standard deviation σ; the abnormal data are all data values whose deviation from the mean exceeds 3.0σ; if all data in the sample library appear within the interval [μ - 3.0σ, μ + 3.0σ], then there are no outliers in the basic sample database.

6. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 1, characterized in that, The specific steps for using the neural network algorithm to construct the prototype of a prediction model for the hydrocarbon expulsion potential of in-situ exploitation of heat-injected fluid in shale by heating are as follows: S51: Divide the model training set and test set according to a ratio, determine the input and output data sets, and perform normalization processing; based on the input and output characteristics of the model, determine the number of nodes in the network input layer and output layer; S52: Set the neural network structure and parameters, including the training function, activation function, number of hidden layers and hidden neurons, maximum number of training times, learning rate, and target error; S53: Initialize the connection weights between the layers of the neural network, as well as the thresholds of the hidden layer and the output layer; S54: Calculate the outputs of the hidden layer and the output layer, and calculate the network prediction error based on the output value of the output layer and the expected output value; S55: Update the network connection weights and thresholds based on the network training error; S56: Determine whether the algorithm iteration reaches the termination condition; if satisfied, obtain the prediction model for the hydrocarbon drainage potential of in-situ shale mining, otherwise, return to step S54.

7. The method for predicting the hydrocarbon drainage potential of in-situ heat injection in shale based on a neural network according to claim 6, characterized in that, The termination condition is reaching the maximum number of training times or reaching the target training error.

8. A method for predicting the hydrocarbon drainage potential of in-situ shale mining by heat injection based on a neural network according to claim 6, characterized in that, The calculation of the hidden layer is as follows: Among them, y is the normalized input information; n and l are the numbers of neurons in the input layer and the hidden layer respectively; v ij is the connection weight between the i-th neuron in the input layer and the j-th neuron in the hidden layer; θ is the hidden layer threshold; f1 is the hidden layer activation function.

9. A method for predicting the hydrocarbon drainage potential of in-situ shale mining by heat injection based on a neural network according to claim 6, characterized in that, The calculation of the output layer is as follows: where m is the number of nodes in the output layer; ω jk is the connection weight between the j-th neuron in the hidden layer and the k-th neuron in the output layer; γ is the output layer threshold; f2 is the output layer activation function.

Images