Miniature three-dimensional multi-element thermal fluid injection-production testing method

Abstract

The invention provides a micro three-dimensional multi-element thermal fluid injection and production testing method. The testing method comprises the steps that S1, an experimental testing method is adopted; step S2, performing parameter optimization; and S3, the optimal injection proportion of the multi-element heat flow fluid is obtained. Engineering parameters can be conveniently and rapidly optimized on site according to reservoir physical property conditions, and reference is provided for construction decision making for improving the recovery efficiency of the multi-element thermal fluid.

Description

Technical Field

[0001] The present invention relates to the technical field of heavy oil reservoir development, and particularly to a micro three-dimensional multi-component thermal fluid injection-production test method. Background Art

[0002] According to the distribution of world oil and gas resources, heavy oil and bitumen resources are several times more than conventional oil and gas resources. Heavy oil is widely distributed and has been discovered in almost all oil-producing countries. The onshore heavy oil areas in China mainly include Liaohe, Shengli, Karamay and Henan oilfields, and the offshore heavy oil areas are mainly distributed in the Bohai oilfield. Thermal recovery is generally used for heavy oil reservoirs. At present, the main technologies for heavy oil reservoir development include steam stimulation, steam flooding and in-situ combustion, etc. Compared with onshore heavy oil oilfields, offshore heavy oil oilfields have larger well spacing and relatively deeper oil layer burial. In addition, restricted by offshore development costs and platform life, it is difficult to place steam injection equipment, the thermal recovery cost is high, and the economic factors have strong constraints, which cannot meet the requirements of high-speed and efficient development of offshore oilfields. Multi-component thermal fluid stimulation is an innovative heavy oil development technology that has been widely applied in Chinese offshore oilfields and has been proven to be a method for enhancing oil recovery with broad application prospects. The thermal recovery equipment required for multi-component thermal fluid technology has characteristics such as small volume and light weight, and is suitable for installation on offshore platforms. Compared with steam stimulation widely used in onshore oilfields, the composition of the carried thermal fluid is different, and its action mechanism is more complex.

[0003] The components of multi-component thermal fluid are diverse, mainly including CO2, N2, water vapor, and a viscosity reducer is occasionally used for ultra-heavy oil. Under high temperature and high pressure conditions, multi-component thermal fluid will undergo complex physical and chemical reactions with reservoir rocks and fluids. For example, the injection of high-temperature and high-pressure steam will heat the crude oil, reduce the viscosity of the crude oil and change the relative permeability of oil and water; the dissolution of CO2 in the crude oil can reduce the viscosity of heavy oil and improve fluidity; after N2 enters the formation, due to gravity segregation, it is distributed in the upper part of the oil layer, forming a heat insulation layer, reducing the heat transfer rate of steam to the overlying rock formation, improving the utilization efficiency of the injected steam, and driving the heated crude oil downward. Moreover, N2 has a large expansion coefficient, which can expand the steam heating radius and increase the swept volume of steam. At the same time, considering the different densities of each component, the spatial synergy effect of multi-component fluid is also involved. Therefore, the mechanism of enhancing oil production by multi-component thermal fluid must also take into account the three-dimensional multi-component fluid distribution characteristics.

[0004] Conventional multi-component thermal fluid simulation and parameter optimization methods include numerical simulation methods, PVT fluid testing methods, one-dimensional core experimental methods, three-dimensional physical simulation experimental methods, etc., but all of the above methods have obvious disadvantages. Multi-component thermal fluids contain multiple complex components. When a mixture of multiple fluids in different proportions is injected into the reservoir under high temperature and high pressure conditions, complex physical and chemical changes occur during the seepage process. Although numerical simulation methods and PVT fluid testing methods are relatively simple, they rely on mathematical calculation methods such as state equations and assume that conditions are relatively ideal, which is quite different from the actual complex reactions of multi-component thermal fluids, oil, water, and rocks. Physical simulation tests are more accurate research methods for multi-component thermal fluids, but multi-component thermal fluid parameter optimization experiments based on one-dimensional cores cannot consider the spatial synergistic effect of multi-component thermal fluids, which is inconsistent with the key mechanism of multi-component thermal fluid development in reality. The three-dimensional physical simulation experimental method is closest to the field application conditions, but the experimental method is complex to operate and has a long experimental cycle, making it difficult to carry out systematic research in batches.

[0005] At present, the existing technical methods all test multi-directional multi-component thermal fluid simulation by drilling cores in different directions. The test device and the displacement direction are one-dimensional. However, the three-dimensional flow of the fluid in the core is not equal to the simple superposition of three one-dimensional flows. Conventional multi-component thermal fluid simulation devices cannot increase or change the displacement direction, so the spatial synergy effect of multi-component thermal fluids cannot be considered. In the present invention, a micro three-dimensional multi-component thermal fluid injection and production model based on square cores is used to consider the characteristics of multi-directional injection of multi-component thermal fluids, which can truly simulate the migration pattern of multi-component thermal fluids in the reservoir, which is closer to the real thing than the traditional one-dimensional model. In addition, in order to guide the application of multi-component thermal fluid technology in the field, the present invention proposes a method of inputting the test data into the BP neural network model, and optimizing it with the NSGA-Ⅱ algorithm, and then calculating the optimal mining parameters of the multi-component thermal fluid technology under a specific reservoir.

[0006] With the continuous improvement of computer performance, intelligent algorithms have developed rapidly. Among them, BP neural network is widely used as an optimization proxy model because it can map the relationship between complex optimization parameters and optimization targets. NSGA-Ⅱ algorithm belongs to genetic algorithm, which is an intelligent algorithm that imitates biological reproduction and evolution. It is widely used in optimization design because of its good global optimization characteristics. At present, traditional multivariate thermal fluid parameter optimization experiments only focus on the impact of each parameter modification on the experiment, ignoring the comprehensive impact of different parameter modifications on the performance of the device.

[0007] For the parameter optimization of the multi-component thermal fluid scheme, there is an urgent need for a solution that can not only simulate the complex physical and chemical reaction processes of multi-component thermal fluid, oil-water, and rock to the greatest extent, but also reflect the spatial synergy effect and efficiently conduct a large number of tests, so as to obtain benchmark parameters closer to the formation conditions and provide sufficient sample data support for parameter optimization. At the same time, a comprehensive parameter optimization algorithm is needed to analyze parameter mapping and correlation, facilitate the rapid optimization of engineering parameters according to reservoir physical property conditions on site, and provide reference for the construction decision-making of enhancing oil recovery by multi-component thermal fluid. Summary of the Invention

[0008] In view of the above problems, the present invention is proposed to provide a micro three-dimensional multi-component thermal fluid injection-production test method that can overcome the above problems or at least partially solve the above problems.

[0009] According to one aspect of the present invention, there is provided a micro three-dimensional multi-component thermal fluid injection-production test method, and the test method includes:

[0010] Step S1: Adopt the method of experimental testing;

[0011] Step S2: Conduct parameter optimization;

[0012] Step S3: Obtain the optimal injection ratio of the multi-component thermal fluid.

[0013] Optionally, the step S1: Adopting the method of experimental testing specifically includes:

[0014] Obtain multiple cubic rock core samples in each heavy oil production block, and conduct limited simulation production experiments on the multiple cubic rock core samples based on the combination of multiple experimental parameters by using the small-scale three-dimensional physical simulation experiment method;

[0015] Monitor and obtain limited data of production indexes in the limited simulation production experiment.

[0016] Optionally, the step S2: Conducting parameter optimization specifically includes:

[0017] Use a BP neural network to construct a correlation model between crude oil parameters and recovery rate based on the limited combination of multi-component thermal fluid and the limited data of production indexes.

[0018] Optionally, the crude oil parameters specifically include: reservoir permeability, crude oil viscosity, and multi-component thermal fluid ratio.

[0019] Optionally, the step S3: Obtaining the optimal injection ratio of the multi-component thermal fluid specifically includes:

[0020] Obtain the geological characteristics of each heavy oil production block;

[0021] The NSGA-II algorithm is used to optimize the constructed BP neural network model, study the correlation mapping between the optimal ratio of multi-component thermal fluids and the recovery rate under specific geological characteristics in the study area, and obtain the optimal injection ratio of the multi-component thermal fluid combination and the heavy oil production plot.

[0022] Optionally, the cubic rock core sample is in a cubic structure, and the construction method includes:

[0023] Set a horizontal well drilled to the reservoir depth, obtain rock samples from the heavy oil production plot, and collect heavy oil samples from the heavy oil production plot;

[0024] Use wire cutting technology to cut the outcrop rock sample into cubic cores of the target size;

[0025] Clean the impurities attached to the cubic core, and dry and weigh the cubic core.

[0026] Optionally, the impurities specifically include: oil, water, salt, soil.

[0027] Optionally, the specific steps of the limited simulation experiment of using the small-scale three-dimensional physical simulation experiment method for simulation production include:

[0028] Place the cubic rock core sample in the high-pressure vessel of the experimental system, and evacuate the model;

[0029] Saturate the cubic core with water, establish the initial temperature field, and monitor the temperature of each temperature measurement point inside the cubic rock core sample through the measurement and control system;

[0030] Saturate the cubic core with oil, establish the initial saturation field, and calculate the initial oil saturation and irreducible water saturation;

[0031] Start the steam generator in the injection system, set the temperature, control the pressure, start the temperature and pressure data acquisition system, monitor the temperature and pressure inside the model in real time, and observe the change of the temperature field;

[0032] Simulate multi-component thermal fluid flooding, open the steady flow and steady pressure valve, and continuously inject steam and mixed gas through the six-way valve at the designed injection speed. Produce from another horizontal well, and the measuring device collects the produced fluid in stages and makes records.

[0033] Optionally, the test parameters of the limited production experiment include: the flow rate, ratio, pressure, temperature and recovery rate of the multi-component thermal fluid.

[0034] Optionally, the geological characteristics include: reservoir permeability, crude oil viscosity data.

[0035] Optionally, the specific steps of constructing the correlation model between crude oil parameters and recovery rate by using the BP neural network based on the limited combination of multi-component thermal fluids and the limited data of production indexes include:

[0036] Take the reservoir permeability, crude oil viscosity, and the proportion of multi-component thermal fluid in all measured experimental results as input data and import them into the BP neural network. The output layer is the recovery factor;

[0037] Set the transfer function and training function of the output layer of the BP neural network, and use part of the training data to train the model.

[0038] A micro three-dimensional multi-component thermal fluid injection-production test method provided by the present invention, the test method includes: step S1, using the experimental test method; step S2, performing parameter optimization; step S3, obtaining the optimal injection ratio of the multi-component thermal fluid. It is convenient to quickly optimize engineering parameters on-site according to reservoir physical property conditions and provide a reference for the construction decision of increasing the recovery factor of multi-component thermal fluid.

[0039] The above description is only an overview of the technical solution of the present invention. In order to be able to more clearly understand the technical means of the present invention, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention are specifically exemplified below. Brief Description of the Drawings

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

[0041] Figure 1 For the embodiments provided by the present invention Figure 1 Flowchart of the heavy oil simulation production method provided by the embodiments of the present invention;

[0042] Figure 2 Curve graph showing the change of the production performance of multi-component thermal fluid flooding with the injected PV number provided by the embodiments of the present invention. Detailed Description of the Embodiments

[0043] The following will describe the exemplary embodiments of the present disclosure in more detail with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0044] The terms "including" and "having" and any variations thereof in the embodiments of the specification, claims, and drawings of the present invention are intended to cover non-exclusive inclusion. For example, including a series of steps or units.

[0045] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0046] As Figure 1 shown, the introduction of the experimental system is as follows:

[0047] This experimental system mainly consists of four parts, specifically including an injection system, a physical model, a production system, and a data acquisition system. The model uses a cubic rock core (model materials and characteristics), a high-temperature and high-pressure micro three-dimensional displacement device; the injection system includes a constant-pressure and constant-speed pump, a high-temperature constant-temperature box, high-temperature and high-pressure oil, gas, and water intermediate containers, and related pipe and valve components. The high-temperature and high-pressure oil, gas, and water intermediate containers are composed of gas flow meters and gas cylinders to simulate the gas components in the multi-component thermal fluid and control their flow rates. The measurement system includes pressure measurements such as injection pressure and back pressure, oil, gas, and water separation, and flow measurement. The production system mainly includes a gas-liquid separator and a collection bottle. The limited simulation experiment of the present invention is realized relying on this experimental system.

[0048] Before conducting the limited simulation experiment of the production, it is necessary to prepare the cubic rock core sample and the experimental environment, specifically as follows:

[0049] The cubic rock core sample has a cubic structure, and the construction method of the cubic rock core sample:

[0050] Set a horizontal well drilled to the reservoir depth, obtain the core of the geology in the heavy oil production research area and its permeability, and collect the heavy oil crude oil and its viscosity in the heavy oil production plot.

[0051] The test process of the small-scale three-dimensional physical simulation experiment method includes:

[0052] Step 1: Use wire cutting technology to cut the outcrop rock sample into

[0053] a 50mm×50mm×50mm cubic core;

[0054] Step 2: Clean the impurities such as oil, water, salt, and soil attached to the cubic core, and dry and weigh the cubic core.

[0055] Step 3: Place the cubic core into the high-temperature and high-pressure micro three-dimensional displacement device 12.

[0056] Step 4: Connect the model to the vacuum process, soak the cubic core with formation water in three directions to saturate it with water, and measure the porosity and permeability of the cubic core from the x, y, and z directions through the water absorption volume.

[0057] Step 5: Establish an initial temperature field. Use the constant-temperature box 10 to make the model reach the initial temperature designed for the experiment. The temperature field should be uniform, and the temperature difference between each temperature measurement point is less than 2°C.

[0058] Step 6: Establish the initial saturation field. Open the conventional valve 2 and use a constant-pressure and constant-speed pump to inject experimental oils with different viscosities from the intermediate container into the model at a speed of 1 mL / min to establish irreducible water until no water flows out at the outlet and the pressure difference is stable. Measure the total amount of water flowing out and calculate the initial oil saturation and irreducible water saturation.

[0059] Step 7: Start the injection system. Start the constant-pressure and constant-speed pump and set the flow rate; start the steam generator, set the temperature, and debug the dryness; control the pressure through the hand pump and backpressure valve. Start the temperature and pressure data acquisition system to monitor the temperature and pressure inside the model in real time and observe the change of the temperature field.

[0060] Step 8: Simulate multi-component thermal fluid flooding. Open the flow-stabilizing and pressure-stabilizing valve and continuously inject steam and N2+CO2 mixture through the six-way valve at the designed injection speed. Produce from another horizontal well, and the measuring device collects the produced fluid in stages and makes records.

[0061] The construction method of the correlation model between reservoir permeability, crude oil viscosity, multi-component thermal fluid ratio and recovery factor constructed by using the BP neural network based on the limited combination of multi-component thermal fluids and the limited data of production indexes includes:

[0062] Take the reservoir permeability, crude oil viscosity and multi-component thermal fluid ratio in all measured experimental results as input data and import them into the BP neural network, and the output layer is the recovery factor.

[0063] Table 1 Schematic table of partial experimental parameter schemes

[0064]

[0065]

[0066] The transfer function of the output layer of the BP neural network adopts the purelin function, the training function adopts the Levenberg_Marquardt algorithm, 75% of the training data is taken to train the model, and the accuracy of the model is verified by the latter 25% group of data.

[0067] Among them, the model construction method is as follows:

[0068] The first step: Set variables and parameters

[0069] Suppose there are N training samples, and each sample has M input feature variables, that is, the input vector X = [x1, x2,..., xM]T. In this BP neural network model, the reservoir permeability, crude oil viscosity and multi-component thermal fluid ratio are used as input data to predict the recovery factor.

[0070] For the hidden layer, assume there are H neurons. For the output layer, since the recovery factor needs to be predicted, there is only 1 neuron in the output layer.

[0071] Set the following variables and parameters:

[0072] Xk = [k1, k2, k3] is the training sample, and N is the number of training samples

[0073] WJI is the weight matrix between the first layer and the second layer, where Wji represents the weight from the i-th neuron in the first layer to the j-th neuron in the second layer. The dimension of WJI is [M, H]

[0074] Wkj is the weight matrix between the second layer and the third layer, where Wkj represents the weight from the j-th neuron in the second layer to the neuron in the output layer. The dimension of Wkj is [H, 1]

[0075] Yk is the output result after forward propagation of the k-th training sample, and its dimension is [1, 1]

[0076] dk is the expected output, that is, the true recovery factor of the k-th training sample, and its dimension is [1, 1]

[0077] η is the learning rate

[0078] g is the number of iterations, that is, the current training round

[0079] Step 2: Initialize the weight matrix and bias term

[0080] Initializing the weight matrix and bias term is an important step, which will determine the final convergence effect of the model. In the BP neural network model, both the weights and bias terms are given random values and their values are continuously adjusted during the training process to improve the fitting ability of the model.

[0081] Assume that the weights and bias terms are initialized as random numbers, expressed as:

[0082] WJI(0) is the weight matrix between the first layer and the second layer, where Wji(0) represents the weight from the i-th neuron in the first layer to the j-th neuron in the second layer. The dimension of WJI(0) is [M, H]

[0083] Wkj(0) is the weight matrix between the second layer and the third layer, where Wkj(0) represents the weight from the j-th neuron in the second layer to the neuron in the output layer. The dimension of Wkj(0) is [H, 1]

[0084] b(0) is the bias term, and its dimension is [H, 1]

[0085] Step 3: Randomly input the sample Xk and set g = 0

[0086] Randomly select a sample from N training samples, use it as the input, and initialize g (representing the number of iterations) to 0.

[0087] Step 4: Forward propagation to calculate the input and output signals of neurons in each layer

[0088] Calculate the input and output signals of neurons in each layer through forward propagation and store the results in a matrix. Assume there are M neurons in the first layer, H neurons in the second layer, and 1 neuron in the third layer. The input signal is Xk, and the output signal is Yk(g). Then the calculation process of forward propagation is as follows:

[0089] Calculation from the first layer to the second layer (weighted sum + activation function):

[0090] Zj(g) = ∑[i = 1, M](Xki * Wji(g)) + bj(g)

[0091] Hj(g) = f(Zj(g))

[0092] Calculation from the second layer to the third layer (weighted sum + activation function):

[0093] Z(g) = ∑[j = 1, H](Hj(g) * Wkj(g))

[0094] Yk(g) = f(Z(g))

[0095] Among them, f is the activation function. Commonly used activation functions include the sigmoid function and the ReLU function.

[0096] Step 5: Calculate the error and determine whether the requirements are met

[0097] Calculate the error of each training sample and determine whether the error meets the requirements. Among them, the error is represented by the mean squared error (MSE):

[0098] E(g) = 1 / 2 * ∑[k = 1, N](dk - Yk(g))^2

[0099] It is necessary to compare this error with a preset threshold to determine whether the current model meets the convergence requirements (that is, whether the error is less than the threshold).

[0100] Step 6: Determine whether the number of iterations has reached the maximum value

[0101] If the number of iterations exceeds the preset maximum number of iterations, jump to Step 8; otherwise, execute Step 7.

[0102] Step 7: Backward propagation to calculate the local gradients of neurons in each layer

[0103] Calculate the local gradients of neurons in each layer through backpropagation, and update the values of weights and bias terms. Assume that the weight matrix from the first layer to the second layer is WJI, and the weight matrix from the second layer to the third layer is Wkj. Then the calculation process of backpropagation is as follows:

[0104] Error transfer from the third layer to the second layer

[0105] δ(g) = (Yk(g) - dk) * f'(Z(g))

[0106] where f' represents the derivative of f

[0107] Error transfer from the second layer to the first layer

[0108] δj(g) = δ(g) * Wkj(g) * f'(Zj(g))

[0109] Update of weights and bias terms

[0110] Wji(g + 1) = Wji(g) - η * δj(g) * Xki

[0111] Wkj(g + 1) = Wkj(g) - η * δ(g) * Hj(g)

[0112] bj(g + 1) = bj(g) - η * δj(g)

[0113] Then, set g + 1 and go to the fourth step to continue training.

[0114] Eighth step: Determine whether all training samples have been learned

[0115] If N training samples have been iterated, end the training and complete the establishment of the BP neural network model. Otherwise, jump to the third step and continue to randomly input samples for training.

[0116] Obtain the geological characteristics of each heavy oil production plot, and use the NSGA-II algorithm to optimize the constructed BP neural network model. The method for studying the correlation mapping between specific geological characteristics and the optimal combination of multiple thermal fluids includes:

[0117] Use the trained BP neural network as the fitness function of the NSGA-II algorithm, set the population size to 80, the number of iterations to 3000 times, the crossover probability to 0.4, and the mutation probability to 0.4. Select the parameter combination when the maximum recovery rate is achieved under specific reservoir permeability and crude oil viscosity as the final optimized parameter combination, and obtain the corresponding proportion of multiple thermal fluids.

[0118] Step 1: Initialize population P with size O. In this example, each individual in population P represents a set of parameter combinations, including the selected specific reservoir permeability and crude oil viscosity, as well as other parameters such as the proportion of thermal fluid. Each individual consists of an unordered set of real values, with each real value corresponding to a parameter.

[0119] Step 2: Calculate the non-dominated rank value, crowding distance, and improved ranking fitness value for each individual. First, calculate the objective function value for each individual, that is, for the selected specific reservoir permeability and crude oil viscosity, solve for the proportion of multi-component thermal fluid required to achieve the maximum recovery rate. Then, use the non-dominated sorting method of the NSGA-II algorithm to divide the individuals in the population into different ranks and calculate the non-dominated rank value for each individual. Next, a selection operation based on the crowding distance is used to maintain the diversity of the solutions. Finally, use the improved ranking fitness value to determine the relative fitness of each individual.

[0120] Step 3: Enter the loop iteration i = 2. In this example, the main objective of each iteration is to generate a new generation of individuals and update the population.

[0121] Step 4: Perform a threshold selection operation using the roulette wheel method for each sub-population based on the non-dominated rank value, crowding distance, and improved ranking fitness value of each individual. First, divide population Q into several sub-populations. Then, for each individual in each sub-population, calculate its selection probability. Finally, randomly select a certain number of individuals according to the selection probability as the parents of this sub-population to participate in the mutation operation.

[0122] Step 5: Use the arithmetic crossover operator to perform the mutation operation to obtain O offspring. In this example, arithmetic crossover means taking the weighted average of the real values at the corresponding positions in the parent individuals to generate new offspring individuals. To increase the diversity of the solutions, two different crossover operators are used in this example: single-point crossover and multi-point crossover.

[0123] Step 6: Calculate the fitness value for each individual after the mutation operation. Specifically, the parent individuals selected in Step 4, after crossover and mutation operations, generate a new set of offspring individuals. Then, for each offspring individual, calculate its objective function value, non-dominated rank value, crowding distance, and improved ranking fitness value.

[0124] Step 7: Collect all individuals in the i-th and (i + 1)-th generations to obtain population Q with size 2O. In this example, a new set of offspring individuals is generated in each iteration. Then, the parent and offspring individuals are combined into population Q for the next loop iteration.

[0125] Step 8: Calculate the non-dominated level value, crowding distance, and improved ranking fitness value of each individual in population Q, and select the better O individuals as the optimal population P using the on-demand stratification strategy. Specifically, first divide the population into several layers according to non-dominated sorting, then select the number of non-dominated solutions in each layer according to the set goals of stratification, and finally select individuals from each layer according to the improved ranking fitness value to form the optimal population P.

[0126] Step 9: If the stopping condition is met, stop; otherwise, i = i + 1, and go to Step 4. In this example, set some stopping conditions, such as the maximum number of iterations, the maximum number of evaluations, the convergence accuracy, etc. If any of the stopping conditions is met, stop the algorithm and output the current optimal individual. Otherwise, continue the loop iteration.

[0127] Step 10: Output the results. After the algorithm stops, output the optimal parameter combination and its corresponding multi-component thermal fluid ratio as the optimal strategy for the specific reservoir to achieve the maximum recovery rate under the selected permeability and crude oil viscosity conditions.

[0128] Compare the multi-component thermal fluid parameters of the obtained optimal combination with the recovery rate after applying the parameters before optimization, including:

[0129] Step 11: Use the final optimized parameter combination to perform Steps 1 to 8 to obtain the recovery rate situation after using the optimized parameters, and conduct a comparative analysis of the production dynamics of multi-component thermal fluid flooding on the change in the recovery rate before and after optimization.

[0130] Table 3 Optimal solution set

[0131]

[0132]

[0133] The results show that:

[0134] As can be seen from Table 2, the MAE in the performance indicators of the BP neural network is 0.1623, and the R-Square is 0.9985, indicating a high credibility of the model.

[0135] Table 2 Performance indicators of the BP neural network

[0136]

[0137] From Figure 2It can be seen that the production performance of the multi-component thermal fluid drive is represented by the change of the recovery factor with the increase of the PV number. As the PV number increases, the increase rate of the recovery factor is first fast and then slow, and finally levels off. The final recovery factor of the parameter combination that achieves the best oil displacement effect before optimization is 66.1%. The final recovery factor of the parameter combination optimized by the BP neural network + NSGA-II algorithm is 87.8%. The difference between the two is 21.7%, indicating that the optimization of the BP neural network + NSGA-II algorithm is of great significance for improving the recovery factor.

[0138] Advantageous effects: Based on the experimental test data of multi-component thermal fluid in cubic cores, the present invention realizes using a computational model to replace the traditional method dominated by one-by-one experiments or numerical simulations. Only a certain number of numerical simulation results are used for training. Utilizing the strong non-linear analysis ability, learning ability and regression ability of the computational model, the parameters of interest are predicted. Combining the BP neural network model and the NSGA-II algorithm, a prediction model between the input parameters (reservoir permeability, crude oil viscosity) and the output parameter (optimal multi-component thermal fluid ratio) is constructed. When the input parameters are filled into the model on site, the optimal injection ratio of the multi-component thermal fluid is calculated to achieve the highest recovery factor.

[0139] The above specific implementation manners have further detailed the purpose, technical solution and advantageous effects of the present invention. It should be understood that the above are only specific implementation manners of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for testing injection and production of a micro three-dimensional multi-component thermal fluid, characterized in that, the testing method includes: Step S1: Adopt the method of experimental testing; Step S2: Conduct parameter optimization; Step S3: Obtain the optimal injection ratio of the multi-component thermal fluid.

2. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 1, characterized in that, the specific implementation of Step S1: Adopt the method of experimental testing includes: Obtain multiple cubic rock core samples in each heavy oil production area, and conduct limited simulation production experiments on the multiple cubic rock core samples based on the combination of multiple experimental parameters using a small-scale three-dimensional physical simulation experiment method; Monitor and obtain limited data of production indicators in the limited simulation production experiment.

3. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 1, characterized in that, the specific implementation of Step S2: Conduct parameter optimization includes: Use a BP neural network to construct a correlation model between crude oil parameters and recovery rate based on the limited combination of multi-component thermal fluids and the limited data of production indicators.

4. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 3, characterized in that, the crude oil parameters specifically include: reservoir permeability, crude oil viscosity, and multi-component thermal fluid ratio.

5. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 1, characterized in that, the specific implementation of Step S3: Obtain the optimal injection ratio of the multi-component thermal fluid includes: Obtain the geological characteristics of each heavy oil production area; Use the NSGA-II algorithm to optimize the constructed BP neural network model, study the correlation mapping between the optimal ratio of multi-component thermal fluids and the recovery rate under the specific geological characteristics of the study area, and obtain the optimal injection ratio of the multi-component thermal fluid combination and the heavy oil production area.

6. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 2, characterized in that, the cubic rock core sample is of a cubic structure, and the construction method includes: Set a horizontal well drilled to the reservoir depth, obtain rock samples in the heavy oil production area, and collect heavy oil samples in the heavy oil production area; Use wire cutting technology to cut the outcrop rock sample into cubic rock cores of the target size; Clean the impurities attached to the cubic rock core, and dry and weigh the cubic rock core.

7. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 6, characterized in that, the impurities specifically include: oil, water, salt, and soil.

8. The method for testing injection and production of a micro three-dimensional multi-component thermal fluid according to claim 2, characterized in that, the specific implementation of using a small-scale three-dimensional physical simulation experiment method to conduct limited simulation production experiments includes: Place the cubic rock core sample in the high-pressure vessel of the experimental system, and evacuate the model; Saturate the cubic rock core with water, establish the initial temperature field, and monitor the temperature of each temperature measurement point inside the cubic rock core sample through the measurement and control system; Saturate the cubic rock core with oil, establish the initial saturation field, and calculate the initial oil saturation and irreducible water saturation; Start the steam generator in the injection system, set the temperature, control the pressure, start the temperature and pressure data acquisition system, monitor the temperature and pressure inside the model in real time, and observe the change of the temperature field; Simulate multi-component thermal fluid flooding. Open the flow and pressure stabilizing valve and continuously inject steam and mixed gas through the six-way valve at the designed injection rate. Another horizontal well is used for production. The measuring device collects the produced fluid in stages and makes records.

9. A micro three-dimensional multi-component thermal fluid injection-production test method according to claim 2, characterized in that the test parameters of the limited production experiment include: the flow rate, ratio, pressure, temperature and recovery rate of the multi-component thermal fluid.

10. A micro three-dimensional multi-component thermal fluid injection-production test method according to claim 5, characterized in that the geological features include: reservoir permeability and crude oil viscosity data.

11. A micro three-dimensional multi-component thermal fluid injection-production test method according to claim 2, characterized in that the specific steps of constructing the correlation model between crude oil parameters and recovery rate by using the BP neural network based on the limited combination of multi-component thermal fluid and the limited data of production indexes include: Taking the reservoir permeability, crude oil viscosity and multi-component thermal fluid ratio in all the measured experimental results as input data and importing them into the BP neural network, and the output layer is the recovery rate; Set the transfer function and training function of the output layer of the BP neural network, and use part of the training data to train the model.

Images