Method for predicting shale oil production capacity by particle swarm optimization CNN

Abstract

The present disclosure provides a model training method and device, and a shale oil productivity prediction method and device. The model training method comprises: constructing a shale oil productivity mathematical model, the shale oil productivity mathematical model being used to represent the relationship between characteristic parameters and target parameters; obtaining a plurality of groups of corresponding characteristic parameters and target parameters as model training data based on the shale oil productivity mathematical model; establishing a convolutional neural network model for predicting shale oil productivity, and obtaining optimal weight values and optimal bias values of the convolutional neural network model by using a particle swarm algorithm; and training the convolutional neural network model based on the model training data to obtain a shale oil productivity prediction model. The shale oil productivity prediction method uses the shale oil productivity prediction model to realize shale oil productivity prediction.

Description

Technical Field

[0001] The present disclosure relates to the technical field of oilfield development, and in particular to a model training method and device, a shale oil production capacity prediction method and device, and a computer-readable storage medium. Background Art

[0002] Shale oil and gas reservoirs have extremely low porosity and permeability, small seepage channels, low pressure coefficients, and insufficient development energy. This results in low oil production, rapid decline in production, difficulty replenishing energy, and low recovery rates. Currently, there are two main approaches to shale oil development: one is to increase the scale of reconstruction using multi-stage horizontal well fracturing to increase production and efficiency; the other is to shut down the well after fracturing and use dialysis replacement to improve oil recovery efficiency. Post-fracturing reservoir thickness, production pressure differential, and fracture conductivity all affect production capacity. Therefore, the factors influencing production capacity are numerous and complex, and the efficient prediction of shale oil production capacity is an urgent issue.

[0003] Traditional methods for predicting production capacity are primarily based on reservoir engineering principles and physical flow simulation techniques, including reservoir engineering methods, analytical solutions, and numerical simulations. These methods require extensive reservoir and fluid data, are difficult to implement on a large scale in the field, are time-consuming, and offer uncertain accuracy.

[0004] In recent years, artificial intelligence (AI) technology has rapidly advanced, particularly neural networks in deep learning, which have been widely applied across various industries. Inspired by biological neural networks, neural networks can extract hidden relationships from large amounts of data, thereby approximating target functions and providing solutions to many previously difficult and complex engineering problems. Summary of the Invention

[0005] In response to the problems existing in the above-mentioned prior art, the embodiments of the present disclosure provide a model training method and device, a shale oil production capacity prediction method and device, and a computer-readable storage medium to more accurately predict shale oil production capacity.

[0006] In one aspect, a model training method is provided for training a shale oil production capacity prediction model, the model training method comprising:

[0007] Constructing a mathematical model for shale oil productivity, wherein the mathematical model is used to characterize the relationship between characteristic parameters and target parameters; wherein the characteristic parameters include oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number; and the target parameter includes production;

[0008] Based on the shale oil productivity mathematical model, multiple sets of corresponding characteristic parameters and target parameters are obtained as model training data;

[0009] Establishing a convolutional neural network model for predicting shale oil production capacity, and using a particle swarm algorithm to obtain the optimal weights and optimal bias values ​​of the convolutional neural network model;

[0010] The convolutional neural network model is trained based on the model training data to obtain a shale oil production capacity prediction model.

[0011] In some embodiments, the shale oil productivity mathematical model includes a mass conservation equation, a motion equation, a state equation, and an auxiliary equation; wherein the motion equation includes a motion equation of the fluid in the main fracture and a fluid motion equation in the matrix-fracture network zone.

[0012] In some embodiments, the step of obtaining multiple sets of corresponding characteristic parameters and target parameters based on the shale oil productivity mathematical model as model training data includes:

[0013] Several sets of orthogonal experiments were designed to calculate the corresponding production under different oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume and cluster number conditions based on the shale oil productivity mathematical model. The data from these several sets of orthogonal experiments were used as model training data.

[0014] In some embodiments, the model training method further includes: performing outlier detection on the model training data using a box method, and deleting outliers in the model training data.

[0015] In some embodiments, after obtaining the model training data, the model training method further includes:

[0016] Dividing the model training data into training group data and test group data;

[0017] Accordingly, training the convolutional neural network model based on the model training data includes:

[0018] The convolutional neural network model is trained based on the training set data, and the effectiveness of the convolutional neural network model is verified using the test set data.

[0019] On the other hand, a model training device is provided, which includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps in the model training method described in any of the above embodiments are executed.

[0020] On the other hand, a computer-readable storage medium is provided, in which computer program instructions are stored. When the computer program instructions are executed by a processor of a user device, the user device executes the model training method described in any of the above embodiments.

[0021] In another aspect, a shale oil production capacity prediction method is provided, the shale oil production capacity prediction method comprising:

[0022] Obtain characteristic parameter data for the target shale reservoir; these parameters include reservoir thickness, fracture half-length, matrix permeability, production pressure differential, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number;

[0023] Inputting the characteristic parameter data into a shale oil productivity prediction model to obtain the predicted productivity of the target shale oil reservoir;

[0024] The shale oil production capacity prediction model is obtained by the model training method described in any of the above embodiments.

[0025] On the other hand, a shale oil production capacity prediction device is provided, which includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps of the shale oil production capacity prediction method described in any of the above embodiments are executed.

[0026] On the other hand, a computer-readable storage medium is provided, characterized in that computer program instructions are stored in the storage medium. When the computer program instructions are executed by a processor of a user device, the user device executes the shale oil production capacity prediction method as described in any of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings illustrate exemplary embodiments of the present disclosure and together with the description serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0028] Figure 1 is a flowchart of a model training method according to some embodiments;

[0029] Figure 2 is a flowchart of another model training method according to some embodiments;

[0030] Figure 3 Schematic diagram of the results of detecting production capacity anomalies using the box method according to some embodiments;

[0031] Figure 4 The present invention is a flowchart of a shale oil production capacity prediction method according to some embodiments. DETAILED DESCRIPTION

[0032] The present disclosure will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the relevant content and are not intended to limit the present disclosure. It should also be noted that, for ease of description, only the portions relevant to the present disclosure are shown in the accompanying drawings.

[0033] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0034] It should be noted that the step numbers in this article are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are executed.

[0035] The method provided in this embodiment can be executed by a related server, and the following description will be made using the server as an example of the execution subject. The execution subject can be adjusted according to the specific case, such as a controller, electronic device, computer, etc.

[0036] As described in the background, productivity is affected by numerous factors and is complex. Traditional methods for predicting productivity require extensive data on reservoirs and fluids, making them difficult to implement on a large scale, time-consuming, and uncertain about the accuracy of prediction results. Therefore, this disclosure provides a shale oil productivity prediction method, apparatus, and computer-readable storage medium based on a particle swarm optimization convolutional neural network. These methods can be applied to shale oil reservoir productivity prediction, offering high prediction accuracy, a simple method, and fast computational speed.

[0037] like Figure 1 and Figure 2 As shown, some embodiments of the present disclosure provide a model training method for training a shale oil production capacity prediction model, and the model training method includes S1 to S4.

[0038] S1, construct a mathematical model of shale oil production capacity.

[0039] The mathematical model for shale oil productivity characterizes the relationship between characteristic parameters, including reservoir thickness, fracture half-length, matrix permeability, production differential pressure, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number, and target parameters, including production.

[0040] In some embodiments, the shale oil productivity mathematical model includes a mass conservation equation, an equation of motion, an equation of state, and auxiliary equations.

[0041] The mass conservation equation:

[0042]

[0043] Where: v is the seepage velocity, unit is m / s;

[0044] φ is the porosity;

[0045] ρ is the fluid density, in kg / m 3 ;

[0046] t is the production time, in seconds;

[0047] o represents the oil phase;

[0048] w represents the water phase;

[0049] S is saturation.

[0050] The equations of motion include the equations of motion of the fluid in the main fracture and the equations of motion of the fluid in the matrix-fracture network region.

[0051] The seepage in the main fracture zone follows the two-phase Darcy law. Considering the strong stress sensitivity of the fracture and the relative permeability characteristics of the fracture, its motion equation is:

[0052]

[0053] Its boundary conditions are:

[0054]

[0055] Where: q is the flow rate of the main fracture, that is, the output or production capacity;

[0056] w f is the crack width;

[0057] h is the reservoir thickness;

[0058] r w is the wellbore radius;

[0059] r w For crack half length;

[0060] p e is the supply pressure;

[0061] p w is the bottom hole pressure;

[0062] μ is the fluid viscosity;

[0063] p is pressure;

[0064] k f is the fracture permeability, unit is 10 -3 μm 2 ;

[0065] x fFor crack half length;

[0066] p f is the fluid pressure in the fracture.

[0067] The seepage in the matrix-fracture network region is a nonlinear seepage characteristic that takes into account the pseudo-starting pressure gradient, the strong stress sensitivity of the matrix-fracture region, and the oil-water two-phase seepage characteristics. Its motion equation is:

[0068]

[0069] Its boundary conditions are:

[0070]

[0071] in,

[0072] Where: p is pressure;

[0073] k f is the fracture permeability, unit is 10 -3 μm 2 ;

[0074] k rl is the relative permeability of phase l;

[0075] S l is the saturation value of phase l;

[0076] is the pressure gradient, in Pa / m;

[0077] μ l is the viscosity of phase 1, in Pa·s;

[0078] G l is the starting pressure gradient of phase l, in Pa / m;

[0079] k m is the matrix permeability, unit is 10 -3 μm 2 ;

[0080] n c is the number of crack clusters;

[0081] p e is the supply pressure;

[0082] μ l is the viscosity of phase 1;

[0083] w f is the crack width;

[0084] k mf is the permeability of the matrix-fracture network area;

[0085] b m is the width of the matrix-fracture network area of ​​a single fracturing stage.

[0086] In addition, b m ×n c It can represent the length of the horizontal well.

[0087] Equation of state: Considering that rock fluid is slightly compressible, porosity and fluid density are affected by the compressibility coefficients of rock and fluid. The specific representation is as follows:

[0088] ρ l =ρ l (p) l=o,w

[0089] φ=φ(p)

[0090] Auxiliary equation: The seepage process is oil-water two-phase, considering the effect of capillary pressure, we have:

[0091] S o +S w =1

[0092] p c (S w )=p o -p w

[0093] Where: p c is the capillary pressure, pa.

[0094] Therefore, according to the mathematical model of shale oil production capacity established above, the target parameter production can be obtained when characteristic parameters such as oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground liquid volume and cluster number are known.

[0095] S2, based on the mathematical model of shale oil production capacity, obtain multiple sets of corresponding characteristic parameters and target parameters as model training data.

[0096] As a possible implementation method, step S2 may specifically be: designing several groups of orthogonal experiments, calculating the corresponding production under different oil layer thicknesses, fracture half-lengths, matrix permeability, production pressure differentials, oil saturation, horizontal well lengths, fracture conductivity, crude oil viscosity, in-ground liquid volume, and cluster number conditions based on the shale oil productivity mathematical model, and using the data from the several groups of orthogonal experiments as model training data.

[0097] Exemplarily, different parameter values are set for the above characteristic parameters respectively to form different combinations of characteristic parameters, that is, multiple groups of orthogonal experiments are designed. A total of N groups of data are collected, that is, the total sample size after all orthogonal experiments is N groups. The dimension of each group of data is M, that is, each group of samples includes M parameters. When the characteristic parameters are reservoir thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, injection volume, and cluster number, M = 10. Thus, the data dimension of the characteristic parameters is N×M. Correspondingly, each group of data also corresponds to a target parameter, and the target parameter set is a matrix of N×1, where 1 is the number of target parameters, that is, 1 parameter - productivity.

[0098] In some embodiments, the model training method further includes: using the boxplot method to detect outliers in the model training data and deleting the outliers in the model training data.

[0099] The boxplot mainly consists of 5 points: the upper edge T max , the upper quartile Q3, the median, the lower quartile Q1, and the lower edge T min .

[0100] The boxplot can be used to identify outliers, and the standard formulas for identifying mild outliers and extreme outliers are as follows:

[0101] MA = {x|x < Q1 - 1.5I or x > Q3 + 1.5I}

[0102] EA = {x|x < Q1 - 3I or x > Q3 + 3I}

[0103] In the formula, Q1 represents the lower quartile; Q3 represents the upper quartile; I = Q3 - Q1; MA represents the set of mild outliers; EA represents the set of extreme outliers.

[0104] Using the boxplot method can intuitively identify outliers in the data, thereby improving the data quality for training the model and making the productivity prediction effect of the trained model more accurate.

[0105] After deleting the outliers in the model training data, the model training data can be divided into training group data and test group data for subsequent model training.

[0106] The division ratio is training group data: test group data = 8:2. The training group data is used to train the machine learning model, and the prediction target of the test group data is assumed to be unknown during prediction and is used to evaluate the prediction accuracy.

[0107] S3. Establish a convolutional neural network model for predicting shale oil productivity, and use the particle swarm optimization algorithm to obtain the optimal weight and optimal bias values of the convolutional neural network model.

[0108] The established convolutional neural network (CNN) model optionally uses the ReLU function as the activation function, sets the batch size to 128, and specifies the number of neurons in the input, hidden, and output layers of the CNN model. The model's input parameters include reservoir thickness, fracture half-length, matrix permeability, production pressure differential, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and number of clusters. The input layer has 10 nodes, and the output layer has only one node (production capacity). The hidden layer of the CNN model consists of two convolutional layers, two pooling layers, and one fully connected layer. The convolution kernel size is 5*5, and the pooling layer uses the max pooling method.

[0109] Among them, the convolution layer is also called a filter. The filtering of the convolution kernel can be used to extract specific features. Convolution includes continuous convolution operations and discrete convolution operations.

[0110] Define two continuous functions f1(t) and f2(t) on the interval (-∞,+∞), then the continuous convolution operation formula is:

[0111]

[0112] Among them, t is the parameter variable, p is the integral variable, and * is the convolution operation.

[0113] The discrete convolution operation formula of x(n) and h(n) is:

[0114]

[0115] Among them, n is the parameter variable, i is the independent variable, and * is the convolution operation.

[0116] Because the production capacity data is not time series data, not a continuous changing value, but a discrete point, the present disclosure adopts discrete convolution.

[0117] Neural networks are difficult to update and optimize for weights and biases, making it difficult to effectively find the optimal parameter combination. However, the particle swarm optimization algorithm simulates the foraging behavior of birds. Each particle in the algorithm represents a potential optimal solution to the optimization problem, and information exchange is achieved by incorporating the coordinates of other particles into the equations for updating their speed and position. Through this simulation, each particle in the particle swarm optimization algorithm ultimately finds the global optimal solution to the optimization problem, making its application to the selection of neural network hyperparameters a promising approach.

[0118] Preprocess the weights and bias values ​​of the CNN model and set the weights w ij and deviation value b ij Encoded into real number particles, the spatial dimension of the particles can be determined, which is equal to the sum of the weights and bias values ​​to be optimized in the neural network.

[0119] Use particle swarm optimization to update the CNN model, specifically including: encoding the weights and bias values ​​of the CNN model; initializing the population size, speed and position information, and the maximum number of iterations k max .

[0120] The prediction mean square error of the convolutional neural network is used as the fitness function to calculate the fitness value of each particle; the optimal fitness value of a single particle is compared with the optimal fitness value of the population and updated; the speed and position are updated according to the speed and position update formula; the above operation is repeated until the search result meets the stop condition (reaching the maximum number of iterations k max ). Specifically:

[0121] The fitness value of each particle is calculated according to the fitness function, which is the predicted mean square error of the convolutional neural network:

[0122]

[0123] Among them, Y i is the expected output of the i-th node in the training data set (the expected output of production capacity), is the actual output of the i-th node during the training process, and ns is the total number of network output nodes, that is, the number of training samples.

[0124] Here, it is equivalent to calculating the deviation between the predicted capacity value and the actual value based on the current particle (weight and deviation value). The calculation result is the fitness value of the particle.

[0125] After calculating the fitness value of each particle after the update, the current fitness value of each particle is compared with its historical fitness value. If the current fitness function value is less than the historical fitness function value, the current fitness value is used as the individual optimal value, otherwise it remains unchanged. If the current global optimal value is greater than the individual optimal value, the individual optimal value at this time is the global optimal value of the particle swarm.

[0126] Save the optimal position of each particle, the sum of the best fitness of all particles in the population, and the corresponding optimal positions of all particles in the population.

[0127] Update the speed and position according to the speed and position update formula.

[0128] The speed update formula is:

[0129]

[0130] The position update formula is:

[0131]

[0132] in, is the current position of the particle;

[0133] ω is the inertia factor; the calculation formula is: ω ini is the initial value of the inertia factor, usually set to 0.4; ω end The inertia factor value when the population iterates to the maximum number of evolutions, usually set to 0.9;

[0134] k max is the maximum number of iterations;

[0135] p i is the current optimal position of the particle;

[0136] p j is the global optimal position corresponding to particle i;

[0137] i=1,2,···,m, represents the number of particles;

[0138] d=1,2,···,D, represents the dimension;

[0139] k represents the current evolutionary generation;

[0140] r1 and r2 are initially random numbers between 0 and 1, which have a certain effect on the diversity of the group, so that the diversity will not be lost during the iteration process;

[0141] c1 and c2 are called learning factors (also called acceleration factors), which affect the convergence speed of the algorithm and are set to 1.5 and 2 respectively.

[0142] Determine whether the searched result meets the stopping condition (maximum number of iterations k max ), when the number of iterations reaches the maximum, the program terminates and outputs the optimal value, otherwise it repeats until the condition is met.

[0143] Finally, the optimal output value, that is, the optimized weight and bias value, is brought into the CNN network model for learning and training.

[0144] The embodiments of the present disclosure use a particle swarm optimization method to obtain optimal weights and bias values, which can improve the prediction accuracy of the model. In addition, the particle swarm optimization method is not prone to falling into local extreme values, and the algorithm is simple and easy to implement, which can improve the training speed of the model.

[0145] S4, train the convolutional neural network model based on the model training data to obtain a shale oil production capacity prediction model.

[0146] Exemplarily, the convolutional neural network model is trained based on the training set data, and the effectiveness of the convolutional neural network model is verified by the test set data.

[0147] The model training method provided by some embodiments of the present disclosure can ensure the accuracy of the production volume solved by preset parameters such as oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, underground liquid volume and cluster number by constructing an effective shale oil production capacity mathematical model. In addition, by adopting the box method, outliers in the data can be intuitively and clearly identified, thereby improving the data quality used for training the model and further improving the accuracy of the production capacity prediction effect of the trained model. In addition, the particle swarm optimization method is used to obtain the optimal weight and deviation values. The particle swarm optimization method is not easy to fall into local extreme values, and the algorithm is simple and easy to implement, which can improve the training speed of the model and improve the prediction accuracy of the model.

[0148] Some embodiments of the present disclosure also provide a model training device, which includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps in the model training method described in any of the above embodiments are executed.

[0149] Some embodiments of the present disclosure also provide a computer-readable storage medium, which stores computer program instructions. When the computer program instructions are executed by a processor of a user device, the user device executes the model training method described in any of the above embodiments.

[0150] like Figure 4 As shown, some embodiments of the present disclosure further provide a shale oil production capacity prediction method, which includes S10 to S20.

[0151] S10, obtaining characteristic parameter data of the target shale oil reservoir. The characteristic parameters include oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number;

[0152] S20, inputting the characteristic parameter data into a shale oil production capacity prediction model to obtain a predicted production capacity of the target shale oil reservoir.

[0153] The shale oil production capacity prediction model is obtained by the model training method described in any of the above embodiments.

[0154] The shale oil production capacity prediction method provided by some embodiments of the present disclosure can solve the problems that traditional methods of predicting production capacity are difficult to implement on a large scale, are time-consuming, and the accuracy of the prediction results cannot be guaranteed, as well as the problems that the updating and optimization of the weights and bias values ​​of the neural network are difficult and the optimal parameter combination cannot be effectively obtained. The method can be applied to the production capacity prediction of shale oil reservoirs, has high prediction accuracy, simple method, fast calculation speed, and has great technical value and application prospects.

[0155] The method of the present invention is further illustrated below by taking a shale oil field as an example.

[0156] First, a mathematical model for shale oil production capacity was constructed, including the mass conservation equation, the equation of motion, the equation of state, and auxiliary equations. The equations of motion include those for the fluid within the primary fracture and the fluid within the matrix-fracture network. The flow in the primary fracture zone follows Darcy's two-phase law, accounting for the strong stress sensitivity of the fractures and the relative permeability characteristics of the fractures. The flow in the matrix-fracture network zone considers the nonlinear flow characteristics of the pseudo-starting pressure gradient, the strong stress sensitivity of the matrix-fracture zone, and the oil-water two-phase flow characteristics.

[0157] Furthermore, based on the constructed mathematical model, we calculated the production rate under different conditions of reservoir thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number. We designed an orthogonal experiment with the design parameters shown in Table 1. A total of 42,352 data sets were collected, each with a dimension of 10. The dimension of the characteristic parameter data was a 42,352 × 10 matrix. Correspondingly, each data set also had a target parameter set, which was the production capacity, with a dimension of 42,352 × 1 matrix.

[0158] Table 1

[0159]

[0160]

[0161] Then use the box method to detect outliers in the data set and delete the outliers in the data set. Taking the production capacity data as an example, the results are as follows: Figure 3 shown.

[0162] The collected data is divided into training group data and test group data. The division ratio is 8:2.

[0163] A convolutional neural network (CNN) prediction model framework suitable for predicting shale oil production capacity was established. The Reluctant Unified Unit (ReLU) function was used as the activation function, the batch size was set to 128, and the number of neurons in the input, hidden, and output layers of the CNN model was determined. The model's input parameters included reservoir thickness, fracture half-length, matrix permeability, production pressure differential, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and the number of clusters. The input layer had 10 nodes, and the output layer had only one node, production capacity. The hidden layers of the CNN model consisted of two convolutional layers, two pooling layers, and one fully connected layer. The convolution kernel size was 5*5, and the pooling layer used a max pooling approach.

[0164] Use particle swarm optimization to update the CNN model. First, encode the weights and bias values ​​of the CNN model, initialize the population and its various parameters, input nodes are 10, the number of particles C = 10, and the maximum allowed number of iterations k max is 100, the maximum value of the inertia weight ω max Set to 0.9, the minimum value ω min It is set to 0.4. c1 and c2 are called learning factors (also called acceleration factors), which affect the convergence speed of the algorithm and are set to 1.5 and 2 respectively.

[0165] The prediction mean square error of the convolutional neural network is used as the fitness function to calculate the fitness value of each particle. The speed and position are repeatedly updated. The optimal fitness value of a single particle is compared with the optimal fitness value of the population and updated; until the search result meets the stopping condition (reaching the maximum number of iterations 100).

[0166] Furthermore, the optimized weights and bias values ​​are brought into the CNN network model for learning and training.

[0167] Finally, the trained CNN model is used to predict shale oil production capacity, with a prediction accuracy of over 87%. This shows that the shale oil production capacity prediction method based on the particle swarm optimization CNN network is highly feasible.

[0168] Some embodiments of the present disclosure also provide a shale oil production capacity prediction device, which includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps in the shale oil production capacity prediction method described in any of the above embodiments are executed.

[0169] Some embodiments of the present disclosure also provide a computer-readable storage medium, characterized in that computer program instructions are stored in the storage medium. When the computer program instructions are executed by a processor of a user device, the user device executes the shale oil production capacity prediction method as described in any of the above embodiments.

[0170] The computer-readable storage media provided by any embodiment of the present disclosure include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0171] An embodiment of the present disclosure further provides an electronic device, including a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and the computer program instructions are executed by the processor to execute the method disclosed in any of the above embodiments when executed.

[0172] The electronic device provided in any embodiment of the present disclosure may be a mobile phone, a computer, a tablet computer, a server, a network device, etc., or may also be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc.

[0173] For example, the electronic device may include: a processor, a memory, an input / output interface, a communication interface, and a bus, wherein the processor, the memory, the input / output interface, and the communication interface are communicatively connected to each other within the device via the bus.

[0174] The processor can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0175] The memory can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory and called and executed by the processor.

[0176] Input / output interfaces connect to input / output modules to enable information input and output. Input / output modules can be built into the device as components or externally connected to provide corresponding functionality. Input devices may include keyboards, mice, touch screens, microphones, and various sensors. Output devices may include displays, speakers, vibrators, and indicator lights.

[0177] The communication interface is used to connect the communication module to enable communication between the device and other devices. The communication module can achieve communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.).

[0178] The bus comprises a pathway that transmits information between various components of a device, such as a processor, memory, input / output interfaces, and communication interfaces.

[0179] It should be noted that although the above device only shows a processor, memory, input / output interface, communication interface, and bus, in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may only include the components necessary to implement the embodiments of this specification, and does not necessarily include all of the components described.

[0180] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that the embodiments of this specification can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solutions of the embodiments of this specification, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments of this specification.

[0181] The methods described in the above embodiments can be implemented by a computer chip or entity, or by a product having certain functions. A typical implementation device is a computer, which can be in the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email transceiver, game console, tablet computer, wearable device, or any combination of these devices.

[0182] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. The method embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, and when implementing the embodiment scheme of this specification, the functions of each module may be implemented in the same or multiple software and / or hardware. It is also possible to select some or all of the modules according to actual needs to achieve the purpose of the embodiment scheme. Those of ordinary skill in the art can understand and implement it without paying any creative work.

[0183] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments / methods or examples. In addition, those skilled in the art may combine and combine different embodiments / methods or examples described in this specification and the features of different embodiments / methods or examples, unless they are contradictory.

[0184] Furthermore, in the description of this disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined. Furthermore, in the description of this disclosure, unless otherwise specifically defined, the terms "connected" and "connection" should be understood broadly. For example, they can mean fixed connection, detachable connection, or integral connection; mechanical connection, electrical connection; direct connection, or indirect connection through an intermediary. Those skilled in the art will understand the specific meanings of the above terms in this disclosure based on specific circumstances.

[0185] Those skilled in the art will appreciate that the above embodiments are merely intended to clearly illustrate the present disclosure and are not intended to limit the scope of the present disclosure. Other changes or modifications may be made based on the above disclosure, and such changes or modifications are still within the scope of the present disclosure.

Claims

1. A model training method, characterized in that: Used to train a shale oil production capacity prediction model, the model training method includes: Constructing a mathematical model for shale oil productivity, wherein the mathematical model is used to characterize the relationship between characteristic parameters and target parameters; wherein the characteristic parameters include oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number; and the target parameter includes production; Based on the shale oil productivity mathematical model, multiple sets of corresponding characteristic parameters and target parameters are obtained as model training data; Establishing a convolutional neural network model for predicting shale oil production capacity, and using a particle swarm algorithm to obtain the optimal weights and optimal bias values ​​of the convolutional neural network model; Training the convolutional neural network model based on the model training data to obtain a shale oil production capacity prediction model; The shale oil production capacity mathematical model includes mass conservation equation, motion equation, state equation and auxiliary equation; The mass conservation equation: Where: is the seepage velocity, in units of ; is the porosity; is the fluid density in units of ; is the production time, in units of ; represents the oil phase; represents the water phase; is saturation; The motion equations include the motion equations of the fluid in the main fracture and the motion equations of the fluid in the matrix-fracture network region; The seepage in the main fracture zone follows the two-phase Darcy law. Considering the strong stress sensitivity of the fracture and the relative permeability characteristics of the fracture, its motion equation is: Its boundary conditions are: ， Where: q The flow rate of the main fracture, that is, the output or production capacity; is the crack width; h is the reservoir thickness; is the wellbore radius; For crack half length; is the supply pressure; is the bottom hole pressure; μ is the fluid viscosity; For pressure; is the fracture permeability, in units of ; x f For crack half length; p f is the fluid pressure in the fracture; The seepage in the matrix-fracture network region is a nonlinear seepage characteristic that takes into account the pseudo-starting pressure gradient, the strong stress sensitivity of the matrix-fracture region, and the oil-water two-phase seepage characteristics. Its motion equation is: Its boundary conditions are: ， in, Where: For pressure; is the fracture permeability, in units of ; for relative permeabilities of phases; for Saturation value of the phase; is the pressure gradient in units of ; for The viscosity of the phase, in units of ; for The starting pressure gradient of the phase, in units of ; is the matrix permeability, in units of ; is the number of crack clusters; is the supply pressure; μ l for Viscosity of the phase; w f is the crack width; k mf is the permeability of the matrix-fracture network area; b m is the width of the matrix-fracture network area of ​​a single fracturing stage; in addition, b m ×n c It can represent the length of the horizontal well; Equation of state: Considering that rock fluid is slightly compressible, porosity and fluid density are affected by the compressibility coefficients of rock and fluid. The specific representation is as follows: Auxiliary equation: The seepage process is oil-water two-phase, considering the effect of capillary pressure, we have: Where: is the capillary pressure, .

2. The model training method according to claim 1, characterized in that The method of obtaining multiple sets of corresponding characteristic parameters and target parameters based on the shale oil productivity mathematical model as model training data includes: Several sets of orthogonal experiments were designed to calculate the corresponding production under different oil layer thickness, fracture half-length, matrix permeability, production pressure difference, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume and cluster number conditions based on the shale oil productivity mathematical model. The data from these several sets of orthogonal experiments were used as model training data.

3. The model training method according to claim 2, characterized in that The model training method further includes: performing outlier detection on the model training data using a box method, and deleting outliers in the model training data.

4. The model training method according to any one of claims 1 to 3, characterized in that After obtaining the model training data, the model training method further includes: Dividing the model training data into training group data and test group data; Accordingly, training the convolutional neural network model based on the model training data includes: The convolutional neural network model is trained based on the training set data, and the effectiveness of the convolutional neural network model is verified using the test set data.

5. A model training device, characterized in that: The device includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps in the model training method according to any one of claims 1 to 4 are executed.

6. A computer-readable storage medium, characterized in that The storage medium stores computer program instructions. When the computer program instructions are executed by the processor of the user device, the user device executes the model training method according to any one of claims 1 to 4.

7. A shale oil production capacity prediction method, characterized in that: The shale oil production capacity prediction method includes: Obtain characteristic parameter data for the target shale reservoir; these parameters include reservoir thickness, fracture half-length, matrix permeability, production pressure differential, oil saturation, horizontal well length, fracture conductivity, crude oil viscosity, in-ground fluid volume, and cluster number; Inputting the characteristic parameter data into a shale oil productivity prediction model to obtain the predicted productivity of the target shale oil reservoir; Wherein, the shale oil production capacity prediction model is obtained by the model training method described in any one of claims 1 to 4.

8. A shale oil production capacity prediction device, characterized in that: The device includes a processor and a memory, wherein the memory stores computer program instructions suitable for execution by the processor, and when the computer program instructions are executed by the processor, the steps in the shale oil production capacity prediction method as described in claim 7 are executed.

9. A computer-readable storage medium, characterized in that The storage medium stores computer program instructions, which, when executed by a processor of a user device, enable the user device to execute the shale oil production capacity prediction method described in claim 7.

Images