Multi-objective optimization method for geological storage process parameters of carbon dioxide in oil reservoir

Abstract

The invention discloses an oil reservoir carbon dioxide geological sequestration process parameter multi-objective optimization method, and relates to the technical field of oil exploitation, and the method comprises the steps: employing an oil reservoir carbon dioxide geological sequestration effect prediction data set to train a variational automatic coding-deep belief neural network model, a reservoir carbon dioxide geological sequestration effect prediction agent model is obtained; constructing an oil reservoir carbon dioxide geological sequestration effect optimization model according to the safety risk coefficient, the effective sequestration coefficient and the benefit optimization coefficient; the reservoir carbon dioxide geological sequestration effect prediction agent model is combined, a non-dominated sorting genetic algorithm is utilized, and a reservoir carbon dioxide geological sequestration process parameter multi-objective optimization mathematical model is solved. The constructed reservoir carbon dioxide geological sequestration process parameter multi-objective optimization mathematical model aims at sequestration safety, potential and economy; and the optimization precision and efficiency of the oil reservoir carbon dioxide geological sequestration process parameters are improved.

Description

Technical Field

[0001] The present application relates to the technical field of oil exploitation, and particularly to a multi-objective optimization method for reservoir carbon dioxide geological storage process parameters. Background Art

[0002] As a major greenhouse gas, carbon dioxide directly affects the global climate environment. Therefore, it is very important to carry out carbon dioxide capture, utilization and storage (CCUS) projects. Geological storage of carbon dioxide is a key link in the CCUS process and an important means to reduce carbon dioxide emissions. The main geological storage sites for carbon dioxide include oil reservoirs, deep saline aquifers and coal seams. Among them, compared with other storage sites, oil reservoirs have the advantages of good geological traps and reservoir-caprock conditions, detailed geological exploration data and numerical simulation models, existing well site facilities, and the ability to improve oil recovery and bring economic benefits. During the process of carbon dioxide storage in oil reservoirs, although the amount of carbon dioxide stored in the reservoir will increase as more carbon dioxide is injected, it will also increase the reservoir pressure, increase the leakage risk, and gradually reduce the efficiency of carbon dioxide enhanced oil recovery, thus reducing economic benefits. Therefore, the key challenge in current geological storage of carbon dioxide in oil reservoirs is how to reasonably optimize the injection and production parameters of carbon dioxide storage to achieve the maximum storage volume, the strongest safety and the highest economic benefits during the carbon dioxide storage process.

[0003] The current optimization of injection and production parameters for carbon dioxide storage has the following problems: The existing optimization methods for reservoir carbon dioxide geological storage process parameters have a single optimization goal. Currently, the main optimization goal is the amount of carbon dioxide geological storage in the reservoir. However, during the process of carbon dioxide geological storage in the reservoir, the geological safety and economic benefits of storage are equally crucial. Therefore, a multi-objective optimization method for reservoir carbon dioxide geological storage process parameters based on storage safety, potential and economy needs to be proposed. In addition, the existing optimization methods for reservoir carbon dioxide geological storage process parameters have the defects of long optimization time and low efficiency. In the actual optimization process, the simulation of carbon dioxide geological storage in the reservoir needs to use a compositional model. For reservoir-scale simulation, the prediction of the storage effect of each scheme requires several hours or even longer, and the optimization process often needs to compare the effects of thousands or even more schemes, which seriously reduces the optimization efficiency. Therefore, a fast and accurate optimization method needs to be proposed. Summary of the Invention

[0004] The purpose of the present application is to provide a multi-objective optimization method for reservoir carbon dioxide geological storage process parameters, which aims at storage safety, potential and economy, and quickly and accurately optimizes the reservoir carbon dioxide geological storage process parameters.

[0005] To achieve the above object, the present application provides the following solutions: The present application provides a multi-objective optimization method for process parameters of geological carbon dioxide sequestration in oil reservoirs, including: Screen characteristic parameters and determine the value ranges of the characteristic parameters of geological carbon dioxide sequestration in oil reservoirs; Establish a geological carbon dioxide sequestration model for the oil reservoir, conduct numerical simulations of different scenarios and extract the results to generate a prediction dataset for the geological carbon dioxide sequestration effect in the oil reservoir; each of the different scenarios is designed based on the value ranges of the characteristic parameters of geological carbon dioxide sequestration in the oil reservoir; Use the prediction dataset for the geological carbon dioxide sequestration effect in the oil reservoir to train a variational autoencoder-deep belief neural network model to obtain a prediction proxy model for the geological carbon dioxide sequestration effect in the oil reservoir; Take the safety risk coefficient, effective sequestration coefficient, and sequestration benefit coefficient as safety, potential, and economic indicators, and construct a multi-objective optimization mathematical model for the process parameters of geological carbon dioxide sequestration in the oil reservoir; Combine the prediction proxy model for the geological carbon dioxide sequestration effect in the oil reservoir, and use the non-dominated sorting genetic algorithm to solve the multi-objective optimization mathematical model for the process parameters of geological carbon dioxide sequestration in the oil reservoir.

[0006] Optionally, the characteristic parameters include: caprock geological parameters, reservoir geological parameters, and injection well parameters; the caprock geological parameters include: caprock porosity, caprock permeability, caprock thickness, caprock lithology, caprock distribution continuity, and caprock gas-sealing index; the reservoir geological parameters include: reservoir porosity, reservoir permeability, reservoir pressure, reservoir temperature, reservoir formation thickness, reservoir interlayer heterogeneity, reservoir permeability anisotropy, reservoir slope, geothermal gradient, and reservoir area; the injection well parameters include: carbon dioxide injection volume, carbon dioxide injection rate, carbon dioxide injection time, and carbon dioxide injection pressure.

[0007] Optionally, establishing a geological carbon dioxide sequestration model for the oil reservoir, conducting numerical simulations of different scenarios and extracting the results to generate a prediction dataset for the geological carbon dioxide sequestration effect in the oil reservoir includes: Use the Petrel platform to construct a geological model for carbon dioxide sequestration in the oil reservoir; According to the value ranges of the characteristic parameters of geological carbon dioxide sequestration in the oil reservoir, use the Latin hypercube sampling method to generate n simulation sample scenarios for geological carbon dioxide sequestration in the oil reservoir; According to the n simulation sample schemes for geological sequestration of carbon dioxide in reservoirs, the reservoir carbon dioxide geological sequestration simulation model is called n times, and numerical simulation is carried out on each simulation sample scheme for geological sequestration of carbon dioxide in reservoirs through the reservoir carbon dioxide geological sequestration simulation method, and a simulation parameter group under different simulation sample schemes for geological sequestration of carbon dioxide in reservoirs is obtained; the simulation parameter group includes the caprock fracture coefficient, the fault slip trend, the carbon dioxide sequestration volume, the cumulative carbon dioxide injection volume, and the cumulative oil production; Based on the simulation sample scheme for geological sequestration of carbon dioxide in reservoirs and the corresponding simulation parameter group of the simulation sample scheme for geological sequestration of carbon dioxide in reservoirs, a prediction data set for the geological sequestration effect of carbon dioxide in reservoirs is constructed.

[0008] Optionally, using the Petrel platform, a geological model for carbon dioxide sequestration in reservoirs is constructed, including: In the Petrel platform, a multi-layer structure is set as a to-be-determined geological model for carbon dioxide sequestration in reservoirs according to a preset size; the multi-layer structure is set from top to bottom; the preset size is 200×200×10; Determine the first preset number of layers at the top as the caprock, and assign the caprock geological properties to the caprock; Determine all the layers outside the caprock in the to-be-determined geological model for carbon dioxide sequestration in reservoirs as the reservoir, and assign the reservoir geological properties to the reservoir; Add a fault to the to-be-determined geological model for carbon dioxide sequestration in reservoirs to obtain a geological model for carbon dioxide sequestration in reservoirs.

[0009] Optionally, the caprock fracture coefficient is: ; Wherein, is the caprock fracture coefficient; S v is the maximum principal stress; S h is the minimum principal stress; μ is the friction coefficient; The fault slip trend is: ; Wherein, is the fault slip trend; is the fault normal stress; is the minimum principal stress; The carbon dioxide sequestration volume, the cumulative carbon dioxide injection volume, and the cumulative oil production are the numerical simulation results of the geological model for carbon dioxide sequestration in reservoirs; The sequestration benefit coefficient is: ; b is the discount rate; is the cumulative oil production; is the crude oil price; C is the production cost; is the cost of carbon dioxide; is the separation cost of carbon dioxide after being produced with crude oil; is the cumulative carbon dioxide injection volume.

[0010] Optionally, the variational auto - encoder deep belief neural network model includes: a variational auto - encoder neural network and a deep belief neural network connected in sequence; The variational auto - encoder neural network is used to reduce the dimension, sample and reconstruct the input data to form a feature vector; The deep belief neural network is used to predict the geological carbon dioxide sequestration effect in the reservoir; The deep belief neural network includes: n restricted Boltzmann machines connected in sequence; The restricted Boltzmann machine includes: a visible layer and a hidden layer connected in sequence; The visible layer of the first restricted Boltzmann machine is connected to the variational auto - encoder neural network; The hidden layer of the i - th restricted Boltzmann machine is connected to the visible layer of the (i + 1) - th restricted Boltzmann machine; i = 1, 2,..., n - 1; The hidden layer of the n - th restricted Boltzmann machine is used to output and predict the geological carbon dioxide sequestration effect in the reservoir.

[0011] Optionally, the method for reconstructing data in the variational auto - encoder neural network includes: Obtain a second preset number of characteristic parameters of geological carbon dioxide sequestration in the reservoir to construct a characteristic parameter matrix; Input the characteristic parameter matrix into the encoder in the variational auto - encoder neural network to obtain the Gaussian distribution in the latent variable space; the encoding layer in the encoder is used to map the characteristic parameter matrix into a low - dimensional latent variable space through non - linear transformation and output the Gaussian distribution in the latent variable space; Randomly sample the Gaussian distribution in the latent variable space to generate a latent variable Z; Input the latent variable Z into the decoder to generate a feature vector; the decoder is used to perform non - linear transformation on the latent variable Z to realize the data decoding process, and then generate a feature vector F.

[0012] Optionally, the training process of the deep belief neural network is: Use the feature vector F and the numerical simulation results as a training sample to construct a training sample set; Initialize the parameters of all restricted Boltzmann machines in the deep belief neural network; Adopt the method of hyperparameter intelligent optimization algorithm to optimize the hyperparameters of the deep belief neural network, and save the deep belief neural network after hyperparameter optimization; Train the deep belief neural network with optimized hyperparameters using the training sample set, and save the trained deep belief neural network.

[0013] Optionally, the multi-objective optimization mathematical model includes an objective function and constraint conditions; The objective function is: ; Where is the objective function for optimizing the three objective functions simultaneously; is the safety risk coefficient; is the effective sealing coefficient; is the sealing benefit coefficient; and are both weight coefficients, and sum to 1; is the caprock fracture coefficient; is the fault slip trend; is the geological carbon dioxide sequestration volume in the reservoir; is the pore volume of the reservoir; is the carbon dioxide density under reservoir conditions; b is the discount rate; is the cumulative oil production; is the crude oil price; C is the production cost; is the carbon dioxide cost; is the separation cost of carbon dioxide after being produced with crude oil; is the cumulative carbon dioxide injection volume; The constraint conditions are: ; Where is the lower bound of the value range of the geological carbon dioxide sequestration process parameters in the reservoir; is the geological carbon dioxide sequestration process parameter in the reservoir; is the upper bound of the value range of the geological carbon dioxide sequestration process parameters in the reservoir.

[0014] Optionally, combined with the proxy model for predicting the geological carbon dioxide sequestration effect in the reservoir, use the non-dominated sorting genetic algorithm to solve the multi-objective optimization mathematical model of the geological carbon dioxide sequestration process parameters in the reservoir, including: Initialize the parameters of the non-dominated sorting genetic algorithm and the geological sequestration process parameters; the parameters of the non-dominated sorting genetic algorithm include population size, crossover probability, mutation probability, non-dominated sorting, crowding distance, and elitist strategy; the geological sequestration process parameters are injection well parameters; Calculate the value of the objective function based on the proxy model for predicting the geological carbon dioxide sequestration effect in the reservoir, and then calculate the Pareto front and crowding distance, and sort the population using non-dominated sorting; Judge whether the convergence condition is satisfied according to the population sorting result to obtain a judgment result; If the judgment result is negative, generate a new population, update the geological storage process parameters after crossover and mutation, and return to the step of "calculating the value of the objective function based on the proxy model for predicting the geological storage effect of reservoir carbon dioxide, and then calculating the Pareto front and crowding distance, and using non-dominated sorting to sort the population"; If the judgment result is positive, output the geological storage process parameters when the safety risk coefficient, effective storage coefficient, and benefit optimization coefficient respectively take the maximum values.

[0015] According to the specific embodiments provided in this application, the following technical effects are disclosed in this application: This application provides a multi-objective optimization method for geological storage process parameters of reservoir carbon dioxide. Using a dataset for predicting the geological storage effect of reservoir carbon dioxide, a variational autoencoder-deep belief neural network model is trained to obtain a proxy model for predicting the geological storage effect of reservoir carbon dioxide. Taking the safety risk coefficient, effective storage coefficient, and storage benefit coefficient as safety, potential, and economic indicators, a multi-objective optimization mathematical model for geological storage process parameters of reservoir carbon dioxide is constructed. Combining with the proxy model for predicting the geological storage effect of reservoir carbon dioxide, a non-dominated sorting genetic algorithm is used to solve the multi-objective optimization mathematical model for geological storage process parameters of reservoir carbon dioxide. Compared with conventional optimization methods for reservoir carbon dioxide storage, the technology of this application can not only optimize the safety, potential, and economic problems involved in the process of carbon dioxide storage simultaneously, comprehensively optimize the carbon dioxide storage effect, but also has a shorter optimization time, greatly reducing the time cost and repetitive work. Therefore, this application has important guiding significance for the optimization of geological storage of reservoir carbon dioxide. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a flowchart of a multi-objective optimization method for geological storage process parameters of reservoir carbon dioxide in an embodiment of this application.

[0018] Figure 2 It is a flowchart of a multi-objective optimization method for geological storage process parameters of reservoir carbon dioxide in another embodiment of this application.

[0019] Figure 3 It is a flowchart of a method for analyzing main control factors by the grey correlation method in an embodiment of this application.

[0020] Figure 4 Schematic diagram of the main factor analysis results in an embodiment of the present application.

[0021] Figure 5 Schematic diagram of the reservoir grid in an embodiment of the present application.

[0022] Figure 6 Schematic diagram of the VAS-DBN prediction model structure in an embodiment of the present application.

[0023] Figure 7 Schematic diagram of the NSGA-II optimization process in an embodiment of the present application.

[0024] Figure 8 Graph showing the variation of the storage safety factor with time corresponding to the maximum value of the safety risk factor in an embodiment of the present application.

[0025] Figure 9 Graph showing the variation of the storage potential factor with time corresponding to the maximum value of the effective storage factor in an embodiment of the present application.

[0026] Figure 10 Graph showing the variation of the storage economic factor with time corresponding to the maximum value of the benefit optimization factor in an embodiment of the present application.

[0027] Figure 11 Iteration graph of the optimization process in an embodiment of the present application. Detailed implementation manners

[0028] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0029] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0030] In an exemplary embodiment, as Figure 1 shown, a multi-objective optimization method for reservoir carbon dioxide geological storage process parameters is provided, including: Step 101: Screen the characteristic parameters and determine the value ranges of the characteristic parameters for geological CO₂ sequestration in the reservoir. The characteristic parameters include: caprock geological parameters, reservoir geological parameters, and injection well parameters. The caprock geological parameters include: caprock porosity, caprock permeability, caprock thickness, caprock lithology, caprock distribution continuity, and caprock gas sealing index. The reservoir geological parameters include: reservoir porosity, reservoir permeability, reservoir pressure, reservoir temperature, reservoir formation thickness, reservoir interlayer heterogeneity, reservoir permeability anisotropy, reservoir slope, geothermal gradient, and reservoir area. The injection well parameters include: CO₂ injection volume, CO₂ injection rate, CO₂ injection time, and CO₂ injection pressure. Use the grey relational analysis method to analyze the main control factors. The grey relational analysis method is as Figure 3 shown; the analysis results of the main control factors are as Figure 4 .

[0031] Step 102: Establish a geological CO₂ sequestration model for the reservoir, conduct numerical simulations of different scenarios and extract the results to generate a prediction dataset for the geological CO₂ sequestration effect in the reservoir. Different scenarios are designed based on the value ranges of the characteristic parameters for geological CO₂ sequestration in the reservoir.

[0032] Step 102 includes: Step 102-1: Use the Petrel platform to construct a geological model for CO₂ sequestration in the reservoir.

[0033] Step 102-1 includes: Step 102-1-2: In the Petrel platform, set a multi-layer structure as a to-be-determined geological model for CO₂ sequestration in the reservoir according to the preset size. The multi-layer structure is set from top to bottom. The preset size is 200×200×10.

[0034] Step 102-1-3: Determine the first preset number of layers at the top as the caprock, and assign the caprock geological attributes to the caprock.

[0035] Step 102-1-4: Determine all the layers outside the caprock in the to-be-determined geological model for CO₂ sequestration in the reservoir as the reservoir, and assign the reservoir geological attributes to the reservoir.

[0036] Step 102-1-5: Add a fault in the to-be-determined geological model for CO₂ sequestration in the reservoir to obtain a geological model for CO₂ sequestration in the reservoir.

[0037] Step 102-2: According to the value ranges of the characteristic parameters for geological CO₂ sequestration in the reservoir, use the Latin hypercube sampling method to generate n simulation sample scenarios for geological CO₂ sequestration in the reservoir.

[0038] Step 102-3: According to the n simulation sample schemes for geological sequestration of carbon dioxide in reservoirs, call the n simulation models for geological sequestration of carbon dioxide in reservoirs, and conduct numerical simulations on each simulation sample scheme for geological sequestration of carbon dioxide in reservoirs through the simulation method for geological sequestration of carbon dioxide in reservoirs, so as to obtain the simulation parameter groups under different simulation sample schemes for geological sequestration of carbon dioxide in reservoirs. The simulation parameter groups include the caprock fracture coefficient, fault slip tendency, carbon dioxide sequestration volume, cumulative carbon dioxide injection volume, and cumulative oil production.

[0039] The caprock fracture coefficient is: .

[0040] Where is the caprock fracture coefficient, representing the stability of the caprock during the geological sequestration of carbon dioxide in the reservoir. S v is the maximum principal stress. S h is the minimum principal stress. μ is the friction coefficient.

[0041] The fault slip tendency is: .

[0042] Where is the fault slip tendency, representing the stability of the fault during the geological sequestration of carbon dioxide in the reservoir. is the normal stress of the fault. is the minimum principal stress.

[0043] The carbon dioxide sequestration volume, cumulative carbon dioxide injection volume, and cumulative oil production are the numerical simulation results of the geological model for carbon dioxide sequestration in the reservoir.

[0044] The sequestration benefit coefficient is: .

[0045] b is the discount rate; is the cumulative oil production; is the crude oil price; C is the production cost; is the carbon dioxide cost; is the separation cost of carbon dioxide after being produced with the crude oil; is the cumulative carbon dioxide injection volume.

[0046] Establish a geological model for carbon dioxide sequestration in the reservoir, design by considering characteristic parameters, conduct numerical simulations for different schemes and extract the results to generate a prediction data set for the effect of carbon dioxide sequestration in the reservoir: Use the Petrel platform to construct a geological model for carbon dioxide sequestration in the reservoir. The size of the geological model for carbon dioxide sequestration in the reservoir is 200×200×10, and the reservoir grid is as Figure 5As shown in the figure. The longitudinal section of the geological model for CO₂ sequestration in the reservoir is divided into 10 layers. Among them, layers 1 and 2 are defined as the caprock, and the geological properties of the caprock are assigned to layers 1 and 2. Layers 3 - 10 are defined as the reservoir, and the geological properties of the reservoir are assigned to layers 3 - 10. Then, a fault is added to the geological model for CO₂ sequestration in the reservoir to obtain a typical geological model for CO₂ sequestration in the reservoir. According to the value range of the characteristic parameters of geological CO₂ sequestration in the reservoir, the Latin hypercube sampling method is used to generate n simulation sample schemes for geological CO₂ sequestration in the reservoir. According to the obtained simulation sample schemes for geological CO₂ sequestration in the reservoir, n simulation models for geological CO₂ sequestration in the reservoir are established. Through the numerical simulation of each sample scheme by the simulation method of geological CO₂ sequestration in the reservoir, the caprock fracture coefficient, fault slip trend, CO₂ sequestration volume, cumulative CO₂ injection volume, and cumulative oil production under different schemes are obtained.

[0047] Step 102 - 4: Based on the simulation sample schemes for geological CO₂ sequestration in the reservoir and the corresponding simulation parameter groups of the simulation sample schemes for geological CO₂ sequestration in the reservoir, construct a prediction data set for the effect of geological CO₂ sequestration in the reservoir.

[0048] Step 103: Use the prediction data set for the effect of geological CO₂ sequestration in the reservoir to train the variational autoencoder - deep belief neural network model (VAE - DBN) to obtain a prediction proxy model for the effect of geological CO₂ sequestration in the reservoir.

[0049] As Figure 6 , the variational autoencoder - deep belief neural network model includes: a variational autoencoder neural network (VAE) and a deep belief neural network (DBN) connected in sequence.

[0050] The variational autoencoder neural network is used to reduce the dimension, sample, and reconstruct the input data to form a feature vector. The data is processed by the VAE in the VAE - DBN model. The VAE has a built - in encoder and decoder, and can reduce the dimension, sample, and reconstruct the input data to form a feature vector F.

[0051] Among them, the method for reconstructing data in the variational autoencoder neural network includes: Obtain a second preset number of characteristic parameters of geological CO₂ sequestration in the reservoir to construct a characteristic parameter matrix.

[0052] Input the characteristic parameter matrix into the encoder in the variational autoencoder neural network to obtain the Gaussian distribution of the latent variable space. The encoding layer in the encoder is used to map the characteristic parameter matrix into a low - dimensional latent variable space through non - linear transformation and output the Gaussian distribution of the latent variable space.

[0053] Randomly sample from the Gaussian distribution of the latent variable space to generate a latent variable Z.

[0054] Input the latent variable Z into the decoder to generate a feature vector. The decoder is used to perform a non-linear transformation on the latent variable Z to implement the data decoding process, and then generate the feature vector F.

[0055] Specifically, the steps for the VAE model to reconstruct data are as follows: (1) Construct the input data. Determine 10 feature parameters to form the input data.

[0056] Each feature parameter includes 2 capping layer geological parameters, 5 reservoir geological parameters, and 3 injection well parameters, forming a feature parameter matrix X with the shape of n×10. The input data is composed of the feature parameter matrix X.

[0057] (2) Data dimensionality reduction. Substitute the input data into the encoding layer. Inside the encoding layer, the input data is mapped to a low-dimensional latent variable space through non-linear transformation. The encoder outputs the Gaussian distribution in the latent variable space, which is represented by the mean and variance in the latent variable space.

[0058] (3) Sampling. Randomly sample to generate a latent variable Z according to the Gaussian distribution in the latent variable space output by the encoder.

[0059] (4) Decoding process. Perform a non-linear transformation on the latent variable Z to implement the data decoding process, and then generate the feature vector F.

[0060] The deep belief neural network is used to predict the geological carbon dioxide sequestration effect in the reservoir. The deep belief neural network includes: n restricted Boltzmann machines connected in sequence. The restricted Boltzmann machine includes: a visible layer and a hidden layer connected in sequence. The visible layer of the first restricted Boltzmann machine is connected to the variational autoencoder neural network. The hidden layer of the i-th restricted Boltzmann machine is connected to the visible layer of the (i + 1)-th restricted Boltzmann machine. i = 1, 2,..., n - 1. The hidden layer of the n-th restricted Boltzmann machine is used to output and predict the geological carbon dioxide sequestration effect in the reservoir.

[0061] The training process of the deep belief neural network is as follows: Use the feature vector F and the numerical simulation results as a training sample to construct a training sample set.

[0062] Initialize the parameters of all restricted Boltzmann machines in the deep belief neural network.

[0063] Use the method of the hyperparameter intelligent optimization algorithm to optimize the hyperparameters of the deep belief neural network, and save the deep belief neural network after hyperparameter optimization.

[0064] Use the training sample set to train the deep belief neural network after hyperparameter optimization, and save the trained deep belief neural network.

[0065] Using the deep belief neural network (DBN) in the VAE-DBN model as the prediction model, the DBN is composed of multiple restricted Boltzmann machines (RBMs) stacked together. The RBM has a visible layer and a hidden layer built-in. The output of the hidden layer of the previous RBM will be used as the input of the visible layer of the next RBM. The sampling result of the hidden layer of the last RBM is used as the input of the BP layer, and the output of the BP layer is the prediction result.

[0066] The steps for training the DBN model are as follows: (1) Reorganization of training data. The feature vector F generated by VAE and the numerical simulation result Y are combined to form a training sample.

[0067] The numerical simulation results of each sample include the caprock fracture coefficient, fault slip trend, carbon dioxide storage capacity, cumulative carbon dioxide injection volume, and cumulative oil production. Construct a training sample to form a target variable matrix Y with a shape of n×5; each of the above training samples is composed of the feature vector F generated by VAE and the numerical simulation result Y. After reorganizing the training data, the training sample format is (F, Y), which is used to train the DBN model.

[0068] (2) Initialization of hyperparameters. Initialize the hyperparameters in each layer of the RBM.

[0069] (3) Hyperparameter optimization. Use the method of hyperparameter intelligent optimization algorithm to optimize the model hyperparameters, and save the DBN model after hyperparameter optimization.

[0070] (4) Model training. Use the DBN model with optimized hyperparameters of the training data for training, and save the trained DBN model.

[0071] Step 104: Using the safety risk coefficient, effective storage coefficient, and storage benefit coefficient as safety, potential, and economic indicators, construct a multi-objective optimization mathematical model for the technological parameters of reservoir carbon dioxide geological storage.

[0072] The multi-objective optimization mathematical model includes an objective function and constraints.

[0073] Taking the safety risk coefficient N s , the effective storage coefficient N p , and the storage benefit coefficient N e as the optimization objectives, and taking the perfusion well parameters as the optimization variables, establish a multi-objective optimization mathematical model for the geological storage process parameters. The multi-objective optimization mathematical model includes an objective function and constraints.

[0074] The objective function is: .

[0075] Among them, is the objective function for simultaneously optimizing the three objective functions. is the safety risk coefficient. is the effective sealing coefficient. is the sealing benefit coefficient. and are both weight coefficients, and The sum of them is 1. is the caprock fracture coefficient. is the fault slip trend. is the geological carbon dioxide storage volume in the reservoir. is the pore volume of the reservoir. is the density of carbon dioxide under reservoir conditions. b is the discount rate. is the cumulative oil production. is the crude oil price. C is the production cost. is the carbon dioxide cost. is the separation cost of carbon dioxide after being produced with the crude oil. is the cumulative carbon dioxide injection volume.

[0076] The constraint condition is: .

[0077] Among them, is the lower bound of the value range of the geological carbon dioxide storage process parameters in the reservoir. is the geological carbon dioxide storage process parameter in the reservoir. is the upper bound of the value range of the geological carbon dioxide storage process parameters in the reservoir.

[0078] Step 105: Combine the proxy model for predicting the geological carbon dioxide storage effect in the reservoir, and use the non-dominated sorting genetic algorithm to solve the multi-objective optimization mathematical model of the geological carbon dioxide storage process parameters in the reservoir.

[0079] Step 105 includes: Step 105-1: Initialize the parameters of the dominated sorting genetic algorithm and the geological storage process parameters. The parameters of the dominated sorting genetic algorithm include population size, crossover probability, mutation probability, non-dominated sorting, crowding degree, and elite strategy. The geological storage process parameters are the parameters of the injection well.

[0080] Step 105-2: Calculate the value of the objective function based on the proxy model for predicting the geological carbon dioxide storage effect in the reservoir, and then calculate the Pareto front and crowding distance, and sort the population using non-dominated sorting.

[0081] Step 105-3: Judge whether the convergence condition is satisfied according to the population sorting result, and obtain the judgment result.

[0082] Step 105-4: If the judgment result is negative, generate a new population, update the geological sequestration process parameters after crossover and mutation, and return to Step 105-2.

[0083] Step 105-6: If the judgment result is positive, output the geological sequestration process parameters when the safety risk coefficient, effective sequestration coefficient, and benefit optimization coefficient respectively take the maximum values.

[0084] Select the non-dominated sorting genetic algorithm (NSGA-II) to solve the multi-objective numerical model, and the proxy model for predicting the geological sequestration effect of reservoir carbon dioxide, forming a multi-objective optimization method for the geological sequestration process parameters of reservoir carbon dioxide based on storage safety, potential, and economy. For example Figure 7 , use the non-dominated sorting genetic algorithm (NSGA-II) to solve the multi-objective numerical model, initialize the NSGA-II parameters (including population size, crossover probability, mutation probability, non-dominated sorting, crowding distance, and elite strategy) and the geological sequestration process parameters, calculate the value of the objective function based on the proxy model for predicting the geological sequestration effect of reservoir carbon dioxide, calculate the Pareto front and crowding distance, sort the population using non-dominated sorting, judge whether the convergence condition is satisfied. If the judgment result is negative, generate a new population, update the geological sequestration process parameters after crossover and mutation, and continue the above steps based on the proxy model for predicting the geological sequestration effect of reservoir carbon dioxide until the convergence condition is satisfied and then end the optimization, output the optimal solution set and the geological sequestration process parameters when the safety risk coefficient, effective sequestration coefficient, and benefit optimization coefficient respectively take the maximum values.

[0085] Next, taking the value range shown in the table as an example, the multi-objective optimization method for the geological sequestration process parameters of reservoir carbon dioxide provided in this embodiment will be specifically described. For example Figure 2 As shown, the following steps are included.

[0086] Step 1: Determine the characteristic parameters.

[0087] The characteristic parameters and their value ranges are shown in Table 1.

[0088] Table 1 Value range table of characteristic parameters in specific embodiments

[0089] Step 2: Establish a reservoir geological sequestration model, design considering the characteristic parameters, conduct numerical simulations of different schemes and extract the results, and generate a dataset for predicting the geological sequestration effect of reservoir carbon dioxide.

[0090] In this embodiment, the Petrel platform is used to construct a geological model for carbon dioxide sequestration in the reservoir. The basic parameters of the model are shown in Table 2. The geological model for carbon dioxide sequestration in the reservoir is longitudinally divided into 10 layers. Among them, layers 1 and 2 are defined as caprock, and the geological properties of the caprock are assigned to layers 1 and 2. Layers 3 - 10 are defined as reservoir, and the geological properties of the reservoir are assigned to layers 3 - 10. Then, a fault is added to the geological model for carbon dioxide sequestration in the reservoir to obtain a typical geological model for carbon dioxide sequestration in the reservoir, as Figure 3 shown.

[0091] Table 2 Basic parameter table of the model in the specific embodiment

[0092] In this embodiment, according to the value range of the characteristic parameters of carbon dioxide geological sequestration in the reservoir, the Latin hypercube sampling method is adopted to generate 1000 simulation sample schemes for carbon dioxide geological sequestration in the reservoir.

[0093] For the above 1000 simulation sample schemes for carbon dioxide geological sequestration in the reservoir, numerical simulations are carried out according to the established typical geological model for carbon dioxide sequestration in the reservoir to obtain the corresponding caprock fracture coefficient, fault slip trend, carbon dioxide sequestration volume, cumulative carbon dioxide injection volume, and cumulative oil production under different schemes. The characteristic parameters and corresponding results of each scheme constitute a prediction data set for the effect of carbon dioxide geological sequestration in the reservoir.

[0094] Step 3: Use the data set to train the variational autoencoder - deep belief neural network model (VAE - DBN) to obtain a trained proxy model for predicting the effect of carbon dioxide geological sequestration in the reservoir.

[0095] The data is processed by the VAE in the VAE - DBN model. The characteristic parameters of each model are combined into a one - dimensional array to form a data matrix X with a shape of 1000×20. Among them, 1000 is the number of samples, and 20 is the number of input features. The data matrix X is input into the VAE model, and finally the feature vector F is output.

[0096] The feature vector F generated by the VAE and the numerical simulation result Y are combined into a training sample. The numerical simulation results of each above - mentioned sample include the caprock fracture coefficient, fault slip trend, carbon dioxide sequestration volume, cumulative carbon dioxide injection volume, and cumulative oil production to construct a training sample, forming a target variable matrix Y with a shape of 1000×5; each above - mentioned training sample is composed of the feature vector F generated by the VAE and the numerical simulation result Y. After reorganizing the training data, the training sample format is (F, Y), which is used to train the DBN model. The trained VAE - DBN model is saved to obtain a proxy model for predicting the effect of carbon dioxide geological sequestration in the reservoir.

[0097] Step 4: With the safety risk coefficient Ns 、Effective storage coefficient N p and benefit optimization coefficient N e , a multi-objective optimization mathematical model of reservoir carbon dioxide geological storage process parameters is constructed.

[0098] Step 5: Select the non-dominated sorting genetic algorithm (NSGA-II) to solve the multi-objective numerical model, and further combine with the reservoir carbon dioxide geological storage effect prediction surrogate model formed in Step 2 to obtain three solutions that achieve the maximum values under the safety risk coefficient N s 、Effective storage coefficient N p and benefit optimization coefficient N e .

[0099] Based on Block M1, optimize the reservoir carbon dioxide geological storage. The geological parameters of this block are shown in Table 3.

[0100] Table 3 Geological parameter table of Block M1

[0101] Solve the multi-objective numerical model using the non-dominated sorting genetic algorithm (NSGA-II). First, initialize the NSGA-II parameters (including population size, crossover probability, mutation probability, non-dominated sorting, crowding degree, and elitist strategy), generate the initial solution u1 and the ideal parameter combination solution u0 in the design space u, obtain the objective function values based on the reservoir carbon dioxide geological storage effect prediction surrogate model, calculate the Pareto front and crowding distance, use non-dominated sorting to sort the population, and judge whether the convergence condition is satisfied. If the judgment result is no, generate a new population, update the geological storage process parameters after crossover and mutation, and continue the above steps based on the reservoir carbon dioxide geological storage effect prediction surrogate model until the convergence condition is satisfied and then end the optimization. Use the Hypervolume index to evaluate the convergence of the multi-objective algorithm. Finally, the convergence condition is satisfied after 72 iterations, and the optimization ends. Finally, the Pareto optimal solution set is obtained. Compare the three solutions that achieve the maximum values under the safety risk coefficient Ns, effective storage coefficient Np, and benefit optimization coefficient Ne in the Pareto optimal solution set, and conduct actual numerical simulation to obtain the optimization effect. All objectives of the multi-objective optimization cannot reach the optimal value simultaneously, and the most suitable process parameter design solution should be selected according to the specific situation. As shown in Table 4. The storage safety coefficient corresponding to the maximum value of the safety risk coefficient changes with time as Figure 8 , the storage potential coefficient corresponding to the maximum value of the effective storage coefficient changes with time as Figure 9 , the storage economic coefficient corresponding to the maximum value of the benefit optimization coefficient changes with time as Figure 10 ; the optimization process iterates as Figure 11 .

[0102] Table 4 Sealing Process Parameter Table for Different Schemes

[0103] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a multi-objective optimization method for reservoir carbon dioxide geological storage process parameters.

[0104] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, it implements the steps in the above method embodiments.

[0105] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, it implements the steps in the above method embodiments.

[0106] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0107] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0108] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0109] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0110] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A multi-objective optimization method for technological parameters of carbon dioxide geological storage in oil reservoirs, characterized in that, Including: Screening characteristic parameters to determine the value range of the characteristic parameters for geological carbon dioxide sequestration in the reservoir; Establishing a geological carbon dioxide sequestration model for the reservoir, conducting numerical simulations for different scenarios and extracting the results to generate a prediction dataset for the geological carbon dioxide sequestration effect in the reservoir; each of the different scenarios is designed based on the value range of the characteristic parameters for geological carbon dioxide sequestration in the reservoir; Using the prediction dataset for the geological carbon dioxide sequestration effect in the reservoir to train a variational autoencoder-deep belief neural network model to obtain a prediction proxy model for the geological carbon dioxide sequestration effect in the reservoir; Taking the safety risk coefficient, effective sequestration coefficient, and sequestration benefit coefficient as safety, potential, and economic indicators, constructing a multi-objective optimization mathematical model for the technological parameters of geological carbon dioxide sequestration in the reservoir; Combining the prediction proxy model for the geological carbon dioxide sequestration effect in the reservoir and using the non-dominated sorting genetic algorithm to solve the multi-objective optimization mathematical model for the technological parameters of geological carbon dioxide sequestration in the reservoir.

2. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 1, characterized in that The characteristic parameters include: caprock geological parameters, reservoir geological parameters, and injection well parameters; the caprock geological parameters include: caprock porosity, caprock permeability, caprock thickness, caprock lithology, caprock distribution continuity, and caprock gas-sealing index; the reservoir geological parameters include: reservoir porosity, reservoir permeability, reservoir pressure, reservoir temperature, reservoir formation thickness, reservoir interlayer heterogeneity, reservoir permeability anisotropy, reservoir slope, geothermal gradient, and reservoir area; the injection well parameters include: carbon dioxide injection volume, carbon dioxide injection rate, carbon dioxide injection time, and carbon dioxide injection pressure.

3. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 1, wherein Establishing a geological carbon dioxide sequestration model for the reservoir, conducting numerical simulations for different scenarios and extracting the results to generate a prediction dataset for the geological carbon dioxide sequestration effect in the reservoir, including: Using the Petrel platform to construct a geological model for carbon dioxide sequestration in the reservoir; According to the value range of the characteristic parameters for geological carbon dioxide sequestration in the reservoir, adopting the Latin hypercube sampling method to generate n simulation sample scenarios for geological carbon dioxide sequestration in the reservoir; According to the n simulation sample scenarios for geological carbon dioxide sequestration in the reservoir, calling the geological carbon dioxide sequestration simulation model n times, and conducting numerical simulations for each simulation sample scenario for geological carbon dioxide sequestration in the reservoir through the geological carbon dioxide sequestration simulation method to obtain a set of simulation parameters under different simulation sample scenarios for geological carbon dioxide sequestration in the reservoir; the set of simulation parameters includes the caprock fracture coefficient, fault slip trend, carbon dioxide sequestration volume, cumulative carbon dioxide injection volume, and cumulative oil production; Based on the simulation sample scenarios for geological carbon dioxide sequestration in the reservoir and the set of simulation parameters corresponding to the simulation sample scenarios for geological carbon dioxide sequestration in the reservoir, constructing a prediction dataset for the geological carbon dioxide sequestration effect in the reservoir.

4. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 3, characterized in that, Using the Petrel platform to construct a geological model for carbon dioxide sequestration in the reservoir, including: In the Petrel platform, setting a multi-layer structure as a to-be-determined geological model for carbon dioxide sequestration in the reservoir according to a preset size; the multi-layer structure is set from top to bottom; the preset size is 200×200×10; Determining the first preset number of layers at the top as the caprock and assigning the caprock geological attributes to the caprock; All layers outside the caprock in the geological undetermined model for carbon dioxide storage in the reservoir are determined as the reservoir, and the geological properties of the reservoir are assigned to the reservoir; A fault is added to the geological undetermined model for carbon dioxide storage in the reservoir to obtain a geological model for carbon dioxide storage in the reservoir.

5. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 3, characterized in that, The fracture coefficient of the caprock is: ； Among them, is the caprock fracture coefficient; S v is the maximum principal stress; S h is the minimum principal stress; μ is the friction coefficient; The fault slip trend is: ； Among them, is the fault slip trend; is the normal stress of the fault; is the minimum principal stress; The carbon dioxide storage volume, the cumulative carbon dioxide injection volume, and the cumulative oil production are the numerical simulation results of the geological model for carbon dioxide storage in the reservoir; The storage benefit coefficient is: ； b is the discount rate; is the cumulative oil production; is the crude oil price; C is the production cost; is the carbon dioxide cost; is the separation cost of carbon dioxide after being produced with the crude oil; is the cumulative carbon dioxide injection volume.

6. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 1, characterized in that The variational autoencoder-deep belief neural network model includes: a variational autoencoder neural network and a deep belief neural network connected in sequence; The variational autoencoder neural network is used to reduce the dimension, sample, and reconstruct the input data to form a feature vector; The deep belief neural network is used to predict the geological carbon dioxide storage effect in the reservoir; The deep belief neural network includes: n restricted Boltzmann machines connected in sequence; The restricted Boltzmann machine includes: a visible layer and a hidden layer connected in sequence; The visible layer of the first restricted Boltzmann machine is connected to the variational autoencoder neural network; The hidden layer of the i-th restricted Boltzmann machine is connected to the visible layer of the i + 1-th restricted Boltzmann machine; i = 1, 2,..., n - 1; The hidden layer of the n-th restricted Boltzmann machine is used to output and predict the geological carbon dioxide storage effect in the reservoir.

7. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 6, characterized in that, The method for reconstructing data in the variational autoencoder neural network includes: Obtain a second preset number of geological carbon dioxide storage characteristic parameters in the reservoir to construct a characteristic parameter matrix; Input the characteristic parameter matrix into the encoder in the variational autoencoder neural network to obtain the Gaussian distribution in the latent variable space; the encoding layer in the encoder is used to map the characteristic parameter matrix into a low-dimensional latent variable space through nonlinear transformation and output the Gaussian distribution in the latent variable space; Randomly sample the Gaussian distribution in the latent variable space to generate a latent variable Z; Input the latent variable Z into the decoder to generate a feature vector; the decoder is used to perform nonlinear transformation on the latent variable Z to realize the data decoding process, and then generate a feature vector F.

8. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 7, characterized in that The training process of the deep belief neural network is: Use the feature vector F and the numerical simulation results as a training sample to construct a training sample set; Initialize the parameters of all restricted Boltzmann machines in the deep belief neural network; Optimize the hyperparameters of the deep belief neural network by using the method of hyperparameter intelligent optimization algorithm, and save the deep belief neural network after hyperparameter optimization; Use the training sample set to train the deep belief neural network after hyperparameter optimization, and save the trained deep belief neural network.

9. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 1, characterized in that, The multi-objective optimization mathematical model includes an objective function and constraint conditions; The objective function is: ； Among them, is the objective function for simultaneous optimization of three objective functions; is the safety risk coefficient; is the effective sealing coefficient; is the sealing benefit coefficient; and are both weight coefficients, and sum to 1; is the caprock fracture coefficient; is the fault slip trend; is the geological carbon dioxide storage volume in the reservoir; is the pore volume of the reservoir; is the carbon dioxide density under reservoir conditions; b is the discount rate; is the cumulative oil production; is the crude oil price; C is the production cost; is the carbon dioxide cost; is the separation cost of carbon dioxide after being produced with the crude oil; is the cumulative carbon dioxide injection volume; The constraint conditions are as follows: ; Among them, is the lower bound of the value range of the process parameters for geological carbon dioxide sequestration in the reservoir; is the process parameter for geological carbon dioxide sequestration in the reservoir; is the upper bound of the value range of the process parameters for geological carbon dioxide sequestration in the reservoir.

10. The multi-objective optimization method for reservoir carbon dioxide geological storage process parameters according to claim 1, wherein Combined with the proxy model for predicting the geological carbon dioxide storage effect in the reservoir, use the non-dominated sorting genetic algorithm to solve the multi-objective optimization mathematical model of the geological carbon dioxide storage process parameters in the reservoir, including: Initialize the parameters of the dominance sorting genetic algorithm and the geological sequestration process parameters; the parameters of the dominance sorting genetic algorithm include population size, crossover probability, mutation probability, non-dominated sorting, crowding degree, and elite strategy; the geological sequestration process parameter is the injection well parameter; Calculate the value of the objective function based on the proxy model for predicting the geological sequestration effect of reservoir carbon dioxide, and then calculate the Pareto front and crowding distance, and sort the population using non-dominated sorting; Judge whether the convergence condition is satisfied according to the population sorting result to obtain a judgment result; If the judgment result is no, generate a new population, update the geological sequestration process parameters after crossover and mutation, and return to the step "Calculate the value of the objective function based on the proxy model for predicting the geological sequestration effect of reservoir carbon dioxide, and then calculate the Pareto front and crowding distance, and sort the population using non-dominated sorting"; If the judgment result is yes, output the optimal solution set of the safety risk coefficient, effective sequestration coefficient, and benefit optimization coefficient and their geological sequestration process parameters.

Images