CN-118029995-B - Method and system for predicting and improving recovery ratio based on neural network model

CN118029995BCN 118029995 BCN118029995 BCN 118029995BCN-118029995-B

Abstract

The invention relates to a method and a system for predicting enhanced oil recovery based on a neural network model, which belong to the technical field of deep learning and enhanced oil recovery, and comprise the steps of constructing and training the neural network model, inputting actual oil reservoir parameters of an oil field development unit into the trained neural network model, and obtaining an economic limit enhanced oil recovery value which can be achieved by the oil field development unit through heterogeneous compound flooding development through the prediction of the trained neural network model. The method comprises the steps of constructing a neural network model, wherein the neural network model comprises an input layer, a hidden layer and an output layer, the step of constructing the neural network model comprises the steps of screening oil reservoir parameters of the input layer of the neural network model, determining recovery ratio under the heterogeneous composite flooding optimal injection and production condition by adopting a particle swarm optimization algorithm, and constructing evaluation indexes of the output layer of the neural network model. According to the invention, the correlation between the prediction index and the oil reservoir parameter is not required to be manually assumed in advance, so that the problem that the prediction model has errors due to the difference of cognitive ability and experience level is avoided.

Inventors

ZHOU KANG
SU KE
HOU JIAN
ZENG QINGDONG
GE XINBO
ZHOU XILONG

Assignees

山东科技大学

Dates

Publication Date: 20260508
Application Date: 20240301

Claims (7)

1. A method for predicting enhanced oil recovery based on a neural network model, comprising: Constructing and training a neural network model; Inputting actual oil reservoir parameters of an oil field development unit into a trained neural network model, and predicting the trained neural network model to obtain an economic limit improvement recovery rate value which can be achieved by the oil field development unit through heterogeneous compound flooding development; the neural network model comprises an input layer, a hidden layer and an output layer, wherein the construction of the neural network model comprises the steps of screening oil reservoir parameters of the input layer of the neural network model; The method comprises the steps of counting the change curves of the heterogeneous composite flooding enhanced recovery value and the increment accumulated net present value along with the injection amount of the chemical agent, and taking the corresponding enhanced recovery value when the increment accumulated net present value is gradually reduced from a positive value to 0 as an evaluation index of the neural network output layer; The training process of the neural network model comprises the following steps: According to the determined oil reservoir parameters of an input layer of the neural network model, compiling orthogonal test schemes, optimizing injection and extraction parameters of heterogeneous compound flooding by adopting a particle swarm algorithm aiming at each orthogonal test scheme to obtain the maximum economic limit improvement recovery value which can be achieved by each orthogonal test scheme, taking the oil reservoir parameters in each orthogonal test scheme and the maximum economic limit improvement recovery value corresponding to the oil reservoir parameters as input parameters and output parameters respectively, and building the training sample set of the neural network model; Optimizing the network structure of the hidden layers of the neural network model, comprising the steps of optimizing and determining the number of layers of the hidden layers of the neural network model and the number of neurons of each hidden layer by comparing training fitting effects of the neural network model according to a determined training sample set of the neural network model, wherein the number of layers of the hidden layers is 1-3, and the number of neurons of each hidden layer is 5-10; the method for optimizing the injection and extraction parameters of the heterogeneous composite flooding by adopting the particle swarm optimization comprises the following steps: Firstly, initializing injection speed, viscoelastic particle injection concentration and volume, tackifier injection concentration and volume, auxiliary injection concentration and volume and liquid production speed of each production well, calling a simulator to develop heterogeneous composite flooding numerical simulation, counting heterogeneous composite flooding amount, viscoelastic particle injection amount, tackifier injection amount and auxiliary injection amount of each year in the development process, and calculating a first economic limit under the current development parameter combination to improve recovery ratio according to a calculation formula of an incremental accumulated net present value; Then, updating development dynamic parameters including injection speed of each injection well, injection concentration and volume of viscoelastic particles, injection concentration and volume of tackifier, injection concentration and volume of auxiliary agent, liquid collecting speed of each extraction well by adopting a particle swarm optimization algorithm, and then calling a simulator to develop heterogeneous composite flooding numerical simulation, and calculating a second economic limit under development parameter combination after updating by adopting the particle swarm optimization algorithm according to a calculation formula of an incremental accumulated net present value to improve recovery ratio; Finally, comparing the calculated first economic limit enhanced recovery ratio and the calculated second economic limit enhanced recovery ratio, if the difference between the two is smaller than 0.5%, using the larger one of the first economic limit enhanced recovery ratio and the second economic limit enhanced recovery ratio as the maximum economic limit enhanced recovery ratio value which can be achieved by the orthogonal test scheme after the injection and extraction parameters of the heterogeneous composite flooding are optimized by adopting a particle swarm algorithm, otherwise, assigning the second economic limit enhanced recovery ratio to the first economic limit enhanced recovery ratio, continuously adopting a particle swarm optimization algorithm to update and develop dynamic parameters, calling a simulator to develop heterogeneous composite flooding numerical simulation, and calculating a new second economic limit enhanced recovery ratio according to a calculation formula of the incremental accumulated net present value; Repeating the steps until the difference between the first economic limit enhanced recovery ratio and the second economic limit enhanced recovery ratio is less than 0.5%, and taking the larger of the two as the maximum economic limit enhanced recovery ratio value which can be achieved by the orthogonal test scheme; The first calculation formula of the economic limit enhanced oil recovery is shown as formula (3): (3) in the formula (3), the amino acid sequence of the compound, Indicating a first economic limit for enhanced recovery; Represent the first Heterogeneous composite flooding oil increasing amount; representing the development years of heterogeneous composite flooding; Representing the original geological reserves of the reservoir.
2. The method for predicting enhanced oil recovery based on a neural network model of claim 1, wherein the step of screening reservoir parameters of an input layer of the neural network model comprises the steps of: Based on the knowledge of mining field development, taking main influence factors of the heterogeneous composite flooding improvement recovery ratio as candidate parameters of an input layer of a neural network model, carrying out numerical simulation research on the influence rule of the heterogeneous composite flooding improvement recovery ratio single factor aiming at each candidate parameter, calculating the sensitivity coefficient of each candidate parameter, and determining candidate parameters with the sensitivity coefficient larger than 0.005 as oil reservoir parameters of the input layer of the neural network model; Candidate parameters of the input layer of the neural network model comprise average permeability, underground crude oil viscosity, interlayer permeability level difference, permeability variation coefficient, stratum effective thickness and reservoir pressure; the calculation formula of the sensitivity coefficient is shown in formula (1): (1) in the formula (1), the components are as follows, Representing a sensitivity coefficient; expressing the number of values adopted in the numerical simulation research of the single factor influence rule of a certain candidate parameter; indicating the heterogeneous composite flooding recovery ratio corresponding to the jth value of a certain candidate parameter; Representing a candidate parameter The values respectively correspond to the average value of the heterogeneous composite flooding improved recovery ratio.
3. The method for predicting enhanced oil recovery based on a neural network model according to claim 1, wherein the calculation formula of the incremental cumulative net present value is as shown in formula (2): (2) in the formula (2), the amino acid sequence of the compound, Representing an incremental cumulative net present value; representing commodity rate of crude oil; Represent the first Heterogeneous composite flooding oil increasing amount; representing a sales price of crude oil; representing ton oil operation costs; Represent the first Annual viscoelastic particle injection amount; Representing a viscoelastic particle purchase price; Represent the first Annual tackifier injection amount; representing a tackifier purchase price; Represent the first Annual adjuvant injection amount; representing an adjuvant purchase price; Representing a resource tax rate; representing the comprehensive tax rate; Representing a rate of return; Representing the years of development of heterogeneous composite flooding.
4. The method for predicting enhanced oil recovery based on a neural network model of claim 1, wherein optimizing the network structure of the hidden layer of the neural network model comprises: Firstly, establishing a network structure of hidden layers of a neural network model with different layers and different neuron numbers according to the number of hidden layers of the neural network model and the value range of the neuron number of each hidden layer; And substituting the input parameters in the determined neural network training sample set into the network structure of the hidden layer of each neural network model, respectively calculating to obtain economic limit improvement recovery rate values, respectively comparing the economic limit improvement recovery rate values with the output parameters in the determined neural network training sample set, and taking the network structure of the hidden layer of the neural network model corresponding to the minimum difference value as the network structure of the hidden layer of the neural network model selected to be used.
5. The method for predicting enhanced oil recovery based on a neural network model of claim 1, wherein validating the neural network model comprises: substituting the determined input parameters in the neural network training sample set into the network structure of the hidden layer of the optimized neural network model, calculating to obtain an economic limit improvement recovery rate value, comparing the economic limit improvement recovery rate value with the determined output parameters in the neural network training sample set, and if the decision coefficient between the two parameters is greater than 95%, indicating that the accuracy of the established neural network model can be used for predicting the heterogeneous composite flooding economic limit improvement recovery rate.
6. The method for predicting enhanced oil recovery based on a neural network model of claim 5, wherein the calculation formula of the decision coefficients is as shown in formula (4): (4) In the formula (4), the amino acid sequence of the compound, Representing the decision coefficients; Representing the number of samples in the neural network training sample set; representing the actual value of the output parameter of the kth training sample; Representation of Average value of output parameter actual values of the training samples; Representing predicted values calculated from the neural network model based on the input parameters of the kth training sample.
7. A system for predicting enhanced oil recovery based on a neural network model, comprising: The neural network model constructing and training module is configured to construct and train a neural network model, wherein the neural network model comprises an input layer, a hidden layer and an output layer, and the neural network model constructing comprises the steps of screening oil reservoir parameters of the input layer of the neural network model, determining the improved recovery ratio under the heterogeneous composite flooding optimal injection and production condition by adopting a particle swarm optimization algorithm, and constructing an evaluation index of the output layer of the neural network model; the prediction module is configured to input actual oil reservoir parameters of the oil field development unit into a trained neural network model, and predict the actual oil reservoir parameters to obtain an economic limit improvement recovery rate value which can be achieved by the oil field development unit through heterogeneous compound flooding development through the trained neural network model; The method comprises the steps of counting the change curves of the heterogeneous composite flooding enhanced recovery value and the increment accumulated net present value along with the injection amount of the chemical agent, and taking the corresponding enhanced recovery value when the increment accumulated net present value is gradually reduced from a positive value to 0 as an evaluation index of the neural network output layer; The training process of the neural network model comprises the following steps: According to the determined oil reservoir parameters of an input layer of the neural network model, compiling orthogonal test schemes, optimizing injection and extraction parameters of heterogeneous compound flooding by adopting a particle swarm algorithm aiming at each orthogonal test scheme to obtain the maximum economic limit improvement recovery value which can be achieved by each orthogonal test scheme, taking the oil reservoir parameters in each orthogonal test scheme and the maximum economic limit improvement recovery value corresponding to the oil reservoir parameters as input parameters and output parameters respectively, and building the training sample set of the neural network model; Optimizing the network structure of the hidden layers of the neural network model, comprising the steps of optimizing and determining the number of layers of the hidden layers of the neural network model and the number of neurons of each hidden layer by comparing training fitting effects of the neural network model according to a determined training sample set of the neural network model, wherein the number of layers of the hidden layers is 1-3, and the number of neurons of each hidden layer is 5-10; the method for optimizing the injection and extraction parameters of the heterogeneous composite flooding by adopting the particle swarm optimization comprises the following steps: Firstly, initializing injection speed, viscoelastic particle injection concentration and volume, tackifier injection concentration and volume, auxiliary injection concentration and volume and liquid production speed of each production well, calling a simulator to develop heterogeneous composite flooding numerical simulation, counting heterogeneous composite flooding amount, viscoelastic particle injection amount, tackifier injection amount and auxiliary injection amount of each year in the development process, and calculating a first economic limit under the current development parameter combination to improve recovery ratio according to a calculation formula of an incremental accumulated net present value; Then, updating development dynamic parameters including injection speed of each injection well, injection concentration and volume of viscoelastic particles, injection concentration and volume of tackifier, injection concentration and volume of auxiliary agent, liquid collecting speed of each extraction well by adopting a particle swarm optimization algorithm, and then calling a simulator to develop heterogeneous composite flooding numerical simulation, and calculating a second economic limit under development parameter combination after updating by adopting the particle swarm optimization algorithm according to a calculation formula of an incremental accumulated net present value to improve recovery ratio; Finally, comparing the calculated first economic limit enhanced recovery ratio and the calculated second economic limit enhanced recovery ratio, if the difference between the two is smaller than 0.5%, using the larger one of the first economic limit enhanced recovery ratio and the second economic limit enhanced recovery ratio as the maximum economic limit enhanced recovery ratio value which can be achieved by the orthogonal test scheme after the injection and extraction parameters of the heterogeneous composite flooding are optimized by adopting a particle swarm algorithm, otherwise, assigning the second economic limit enhanced recovery ratio to the first economic limit enhanced recovery ratio, continuously adopting a particle swarm optimization algorithm to update and develop dynamic parameters, calling a simulator to develop heterogeneous composite flooding numerical simulation, and calculating a new second economic limit enhanced recovery ratio according to a calculation formula of the incremental accumulated net present value; Repeating the steps until the difference between the first economic limit enhanced recovery ratio and the second economic limit enhanced recovery ratio is less than 0.5%, and taking the larger of the two as the maximum economic limit enhanced recovery ratio value which can be achieved by the orthogonal test scheme; The first calculation formula of the economic limit enhanced oil recovery is shown as formula (3): (3) in the formula (3), the amino acid sequence of the compound, Indicating a first economic limit for enhanced recovery; Represent the first Heterogeneous composite flooding oil increasing amount; representing the development years of heterogeneous composite flooding; Representing the original geological reserves of the reservoir.

Description

Method and system for predicting and improving recovery ratio based on neural network model Technical Field The invention belongs to the technical field of deep learning and enhanced oil recovery, and particularly relates to a method and a system for predicting enhanced oil recovery based on a neural network model. Background Heterogeneous composite flooding realizes further improvement of petroleum recovery rate of high-water-content oil fields by injecting viscoelastic particles, tackifier and auxiliary solution into oil reservoirs, and has been successfully applied to fields such as victory in China. However, the geological conditions of the high-water-content oil field are various, the heterogeneous composite flooding investment is large, the cost is high, and how to accurately evaluate the maximum enhancement recovery ratio achieved by adopting the optimal injection and production parameters under the condition of meeting the economic benefit requirement is a key for reasonably screening oil reservoir blocks in a mining field and further implementing the heterogeneous composite flooding, so that the method has important practical guiding significance. At present, the prediction method for improving the recovery ratio mainly comprises experience analogy, regression formula, deep learning and the like. The empirical analogy method is used for determining the recovery rate value according to main indexes such as the viscosity, the average permeability, the heterogeneity of the oil deposit and the like of the stratum crude oil and analogy with the developed oil deposit, but the actual oil deposit can only analogically obtain the approximate distribution range due to complex geological conditions and large difference, and the recovery rate value can not be accurately predicted. The regression formula method is used for establishing a regression formula according to the geological conditions of the developed oil reservoirs and the data of the enhanced recovery ratio in a fitting way, and then substituting the regression formula into actual oil reservoir parameters to solve the prediction, but the method needs to assume a formula form in advance, has artificial interference, and has larger prediction error when sample data are insufficient. The deep learning can obtain a prediction model between the evaluation index and the influence factor through learning training on the premise of not needing to manually presume the correlation in advance by the artificial neural network, and the prediction precision is higher. The artificial neural network is trained by a large number of samples with known input and output parameters, and the weight of the network structure is continuously adjusted through an internal self-adaptive algorithm so as to obtain the best fit prediction effect. However, heterogeneous composite flooding is an emerging enhanced recovery method, and mine sample data is limited, and no application-tested neural network structure has been reported. In addition, the currently known heterogeneous composite flooding recovery enhancement values are all obtained under a single development condition, and the maximum potential value for enhancing the recovery under the optimal development condition is not clear. Therefore, there is a need for a method and system for predicting enhanced oil recovery based on neural network models. Disclosure of Invention Aiming at the characteristics of the prior art, which are insufficient and a neural network model, the invention provides a method for predicting the recovery ratio improvement based on the neural network model, which is used for determining the oil reservoir parameters of a neural network input layer according to sensitivity analysis and screening, determining the economic limit recovery ratio improvement under the optimal injection and production condition of heterogeneous compound flooding by adopting a particle swarm optimization algorithm and using the economic limit recovery ratio improvement as a prediction index of a neural network output layer, constructing a large number of training samples based on numerical simulation calculation and optimizing and adjusting the network structure of a neural network hidden layer, thereby establishing the neural network prediction model for improving the recovery ratio. Term interpretation: orthogonal test scheme the orthogonal test design is a design method for researching multi-factor influence rule, and is a high-efficiency, quick and economic test design method by selecting partial representative scheme from comprehensive test schemes according to orthogonality for test. And (3) carrying out heterogeneous composite flooding numerical simulation, namely solving a mathematical model of the heterogeneous composite oil displacement reservoir by using a computer, simulating underground oil-water flow, giving out oil-water distribution at a certain moment, assisting in determining a reasonable development scheme an