CO2 dynamic miscibility field intelligent prediction method based on multi-physics field coupling
By combining multiphysics coupling with intelligent models, the problem of dynamic full-field characterization of CO2-crude oil miscibility prediction was solved, achieving accurate miscibility prediction and visualization, and supporting the optimization of CO2-driven reservoir development.
Patent Information
- Application Number
- CN202511059995.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-30
- Publication Date
- 2025-11-14
AI Technical Summary
Existing technologies lack dynamic full-field characterization and multi-physics coupling in predicting the miscibility of CO2-crude oil, resulting in large prediction biases and making it difficult to achieve real-time calculation and visualization at the mine scale. There is also a lack of targeted development solutions for optimization.
By combining multiphysics field coupling simulation with intelligent models, data on multi-scale influencing factors are collected, a CO2-crude oil miscibility prediction model is established, and a dynamic miscibility field of CO2-crude oil is calculated and formed in real time. Machine learning and numerical simulation methods are used for accurate prediction.
It achieves a leap from static single-point prediction to dynamic full-field distribution, capturing in real time the differences in mixing capacity in different development stages and regions, providing accurate basis for CO2 drive development schemes, and providing dynamic visualization tools.
Smart Images
Figure CN120954547A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of CO2 flooding reservoir development technology, and in particular to an intelligent prediction method for dynamic miscibility of CO2 based on multi-physics field coupling. Background Technology
[0002] The miscibility of CO2 with crude oil is a core indicator that determines the efficiency of oil displacement. Its strength directly affects the range of the miscible zone, the recovery rate, and the CO2 storage efficiency.
[0003] However, existing technologies still have some limitations in predicting miscibility. Firstly, current methods primarily rely on static, single-point predictions, lacking dynamic, full-field characterization. Traditional methods measure single-point parameters such as minimum miscibility pressure (MMP) through capillary experiments or calculate miscibility at specific locations based on empirical formulas. These methods fail to reflect the impact of dynamic changes in pressure, composition, and temperature fields on miscibility during reservoir development, making it difficult to capture miscibility differences across different development stages and regions. Secondly, the characterization of multi-physics coupling effects is insufficient. CO2-crude oil miscibility is influenced by multiple factors, including temperature and crude oil composition. Existing numerical simulations often simplify the coupling relationships between physical fields, leading to significant prediction errors. Furthermore, the lack of intelligent prediction tools at the field scale makes it difficult to achieve real-time calculation and visualization of miscibility at the field scale, resulting in a lack of targeted optimization of injection parameters.
[0004] Therefore, there is an urgent need for a method that combines multi-physics coupling with intelligent prediction to achieve dynamic, full-field, and accurate characterization of CO2-crude oil miscibility, supporting the optimization of development schemes at the mine scale. Summary of the Invention
[0005] The purpose of this invention is to provide an intelligent prediction method for the dynamic miscibility of CO2 based on multi-physics coupling. By combining multi-physics coupling simulation with intelligent models, the dynamic prediction and visualization of CO2-crude oil miscibility at the reservoir scale can be achieved, providing a precise basis for optimizing CO2 flooding development schemes.
[0006] To achieve the above objectives, the present invention provides the following solution:
[0007] A smart prediction method for the dynamic miscibility field of CO2 based on multi-physics coupling includes:
[0008] Collect multi-scale influencing factors of CO2-crude oil miscibility pressure and construct a sample set;
[0009] A CO2-crude oil miscibility prediction model was established based on the aforementioned sample set.
[0010] Collect reservoir geological data and fluid data to establish a three-dimensional reservoir numerical model;
[0011] Based on the three-dimensional reservoir numerical model and the CO2-crude oil miscibility prediction model, the dynamic miscibility field of CO2-crude oil is calculated and formed in real time.
[0012] Optionally, data on multi-scale influencing factors of CO2-crude oil miscibility pressure are collected to construct a sample set including:
[0013] Identify the multi-scale influencing factors affecting the miscibility pressure of CO2-crude oil, collect relevant experimental and simulation data, and construct a multi-dimensional sample database;
[0014] Statistical characteristic analysis was performed on the multi-dimensional sample database. The Z-score method was used to identify outliers and missing values, and the gradient boosting tree regression imputation method was used to process the outliers and missing values.
[0015] An improved Min-Max method is used to normalize the processed multi-dimensional sample database, and the normalized multi-dimensional sample database is divided into a training sample set and a test sample set. The improved Min-Max method introduces a dynamic scaling factor, which is dynamically adjusted according to the sensitivity of the variable to the mixing pressure.
[0016] Optionally, the improved Min-Max method is:
[0017]
[0018] Where X is the value of the variable, X min Let X be the minimum value of the variable. max w is the maximum value of the variable. k Let be the sensitivity coefficient of the k-th variable.
[0019] Optionally, establishing a CO2-crude oil miscibility prediction model based on the sample set includes:
[0020] Select several machine learning models as benchmark models;
[0021] The sample set is input into the benchmark model for hyperparameter fitting, and the benchmark model is optimized for hyperparameters using a grid search strategy to obtain the trained benchmark model.
[0022] Several evaluation metrics were used to evaluate the performance of the trained benchmark model, and the benchmark model with the best performance was selected as the CO2-crude oil miscibility prediction model.
[0023] Optionally, the grid search strategy includes:
[0024] Define a set of candidate values for the hyperparameters of the baseline model and arrange and combine them.
[0025] For each combination, the model is trained using the sample set, and the generalization ability is evaluated through cross-validation;
[0026] The combination with the highest cross-validation score is selected as the optimal hyperparameter of the baseline model.
[0027] Optionally, collecting reservoir geological data and fluid data to establish a three-dimensional reservoir numerical model includes:
[0028] Fluid property data were obtained through high-temperature and high-pressure visualization experiments;
[0029] Based on the fluid property data, the phase parameters of crude oil and the minimum miscibility pressure are fitted. When the fitting accuracy reaches the preset accuracy, the numerical model of fluid composition is output.
[0030] Based on the fluid composition numerical model, the three-dimensional reservoir numerical model is constructed by combining reservoir geological parameters.
[0031] Optionally, the fluid property data includes crude oil composition, minimum miscibility pressure of CO2 and crude oil, crude oil density, crude oil viscosity, and gas-oil ratio;
[0032] The reservoir geological parameters include reservoir porosity, reservoir permeability, reservoir temperature, formation pressure, reservoir thickness, reservoir depth, well location information, relative permeability curve, reservoir oil saturation, pressure coefficient, and geothermal gradient.
[0033] Optionally, based on the three-dimensional reservoir numerical model and the CO2-crude oil miscibility prediction model, the real-time calculation and formation of the CO2-crude oil dynamic miscibility field includes:
[0034] Based on the three-dimensional reservoir numerical model, a multiphysics time series result file of the entire reservoir grid is run in real time and outputs. The multiphysics field includes component fields, temperature fields, and pressure fields.
[0035] The data in the multiphysics time series results file are converted into a three-dimensional matrix array in batches, and the three-dimensional matrix array is processed by spatial coordinate hash mapping;
[0036] The processed three-dimensional matrix array is combined with the CO2-crude oil miscibility prediction model for calculation. The minimum miscibility pressure of CO2-crude oil is predicted grid by grid and time step by time to obtain the dynamic CO2-crude oil minimum miscibility pressure data matrix. The dynamic CO2-crude oil minimum miscibility pressure data matrix is then converted into a spatiotemporal grid heat map.
[0037] The beneficial effects of this invention are as follows:
[0038] This invention provides an intelligent prediction method for the dynamic miscibility field of CO2 based on multi-physics coupling. By combining multi-physics coupling with an intelligent prediction model, it achieves a leap from static single-point prediction to dynamic full-field distribution. It can capture the differences in miscibility of different development stages and different grids in real time, and provides a dynamic visualization tool for understanding the mechanism of CO2-driven miscibility. Attached Figure Description
[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0040] Figure 1 This is a flowchart of an intelligent prediction method for dynamic miscibility field of CO2 based on multi-physics coupling, according to an embodiment of the present invention.
[0041] Figure 2 This is a data sample distribution diagram of an embodiment of the present invention;
[0042] Figure 3 This is a comparison chart of the prediction errors of the baseline model in an embodiment of the present invention;
[0043] Figure 4 This is a graph showing the prediction results of the optimal baseline model in an embodiment of the present invention.
[0044] Figure 5 This is a figure showing the fitting results of the numerical model of fluid components in an embodiment of the present invention;
[0045] Figure 6 This is a schematic diagram of a three-dimensional reservoir numerical model according to an embodiment of the present invention;
[0046] Figure 7 This is a diagram showing the prediction results of the dynamic miscibility field in an embodiment of the present invention. Detailed Implementation
[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0049] This embodiment provides a method for intelligent prediction of the dynamic miscibility field of CO2 based on multi-physics coupling, including:
[0050] Collect multi-scale influencing factors of CO2-crude oil miscibility pressure and construct a sample set;
[0051] A CO2-crude oil miscibility prediction model was established based on the aforementioned sample set.
[0052] Collect reservoir geological data and fluid data to establish a three-dimensional reservoir numerical model;
[0053] Based on the three-dimensional reservoir numerical model and the CO2-crude oil miscibility prediction model, the dynamic miscibility field of CO2-crude oil is calculated and formed in real time.
[0054] Specifically, this embodiment combines multiphysics coupling with an intelligent prediction model to achieve a leap from static single-point prediction to dynamic full-field distribution. It can capture in real time the differences in mixing capabilities at different development stages and in different grids, providing a dynamic visualization tool for understanding the mechanism of CO2-driven mixing. For example... Figure 1 As shown, the specific steps include:
[0055] Step S1: Obtain the sample set for model training and testing, and perform missing value and outlier value processing and analysis;
[0056] Step S2: Establish a CO2-crude oil miscibility prediction model based on the processed sample set;
[0057] Step S3: Collect reservoir geological data and fluid data to establish a three-dimensional reservoir numerical model;
[0058] Step S4: Based on the multiphysics simulation results of the three-dimensional reservoir numerical model, the miscibility prediction model is linked with the real-time data interface to realize the real-time calculation and dynamic update of miscibility at the grid scale of the entire reservoir, forming a spatiotemporally continuous dynamic miscibility field of CO2-crude oil.
[0059] Step S1 involves obtaining, processing, and analyzing the sample sets used for model training and testing.
[0060] Identify the multi-scale influencing factors affecting the miscibility pressure of CO2-crude oil, collect relevant experimental and simulation data, and construct a multi-dimensional sample database;
[0061] The collected data were statistically analyzed, and outliers were identified using the Z-score method. The gradient boosting tree regression imputation method was then used to handle outliers and missing values, thus solving the adaptability problem of traditional methods in heterogeneous datasets.
[0062] An improved Min-Max method is used to normalize the processed dataset, and a dynamic scaling factor is introduced to dynamically adjust the variables based on their sensitivity to miscibility pressure, thereby further eliminating the dimensional influence between variables and improving the model's prediction efficiency.
[0063] In step S1, the Z-score method is as follows:
[0064]
[0065] Where x is the value of a single data point, μ is the mean of the dataset, and σ is the standard deviation of the dataset. Data points with |Z|>3 are considered outliers.
[0066] In step S1, the regression imputation method is as follows:
[0067] Select the sample data without missing data and the actual sample data to be filled;
[0068] Using complete sample data, a regression prediction model is constructed using gradient boosting trees:
[0069]
[0070] Where β is the model parameter, F(x_i;β) is the gradient boosting tree prediction function, and L is the mean squared error loss function.
[0071] In step S1, the improved Min-Max method is as follows:
[0072]
[0073] Where X is the value of the variable, X min Let X be the minimum value of the variable. max w is the maximum value of the variable. k The sensitivity coefficient of the k-th variable is calculated using grey relational analysis.
[0074] In step S2, the specific steps for establishing the CO2-crude oil miscibility prediction model are as follows:
[0075] Multiple machine learning models were selected as benchmark models, including: Ridge, SVR, RF, GBM, XGBoost, and MLP.
[0076] The processed dataset is divided into training and test sets, input into the benchmark model for hyperparameter fitting, and a grid search strategy is used to optimize the hyperparameters of the benchmark model.
[0077] Multiple evaluation metrics were used to assess the performance of the benchmark model, and the benchmark model with the best performance was selected as the miscibility prediction model.
[0078] In step S2, the grid search strategy is as follows:
[0079] Define a set of candidate values for the hyperparameters of each baseline model and arrange and combine them.
[0080] For each combination, the model is trained with the training data and its generalization ability is evaluated through cross-validation.
[0081] The combination with the highest cross-validation score is selected as the optimal hyperparameter of the baseline model.
[0082] In step S2, various evaluation indicators include:
[0083] Root Mean Square Error (RMSE):
[0084]
[0085] Where n is the number of samples, y i Let y′ be the true value of the i-th sample. i Let be the predicted value for the i-th sample;
[0086] Mean absolute error:
[0087]
[0088] Coefficient of determination R²:
[0089]
[0090] In step S3, the specific steps for establishing the three-dimensional reservoir numerical model are as follows:
[0091] Key fluid property data such as crude oil composition, minimum miscibility pressure between CO2 and crude oil, crude oil density, crude oil viscosity, and gas-oil ratio were obtained based on high-temperature and high-pressure visualization experiments.
[0092] Based on the fluid property data, the CMG-Winprop phase software is used to fit the crude oil phase parameters and minimum miscibility pressure. When the fitting accuracy is higher than 90%, the numerical model of fluid components is output.
[0093] A three-dimensional reservoir numerical model is constructed based on a fluid composition numerical model and reservoir geological parameters. The reservoir geological parameters include: reservoir porosity, reservoir permeability, reservoir temperature, formation pressure, reservoir thickness, reservoir depth, well location information, relative permeability curve, reservoir oil saturation, pressure coefficient, and geothermal gradient.
[0094] In step S4, the specific steps for constructing the CO2-crude oil dynamic miscibility field include:
[0095] Based on a three-dimensional reservoir numerical model, the system runs in real time and outputs time-series results files of the component fields, temperature fields, and pressure fields of the entire reservoir grid.
[0096] The data processing module is written in Python. It uses the NumPy library to convert multiphysics data into three-dimensional matrix arrays in batches. Spatial coordinate hash mapping is used to ensure that the grid index and time node of each matrix are completely aligned.
[0097] Based on the PyTorch distributed computing framework, the preprocessed three-dimensional matrix is combined with the trained optimal miscibility prediction model for calculation, and the minimum miscibility pressure is predicted grid by grid and time step by time.
[0098] The predicted dynamic MMP data matrix was converted into a spatiotemporal grid heatmap using the matplotlib library. Pressure values were encoded by color gradients, which intuitively revealed the spatial differentiation characteristics and dynamic evolution of MMP in different regions and development stages of the reservoir.
[0099] The following example uses a low-permeability reservoir with a burial depth of 3550.0m to 4060.8m, a reservoir porosity of 5.7%, a reservoir permeability of 5-10 mD, a reservoir temperature of 154.2℃, a reservoir pressure of 36.1 MPa, and a CO2-crude oil miscibility pressure of 40.8 MPa. The specific steps are as follows:
[0100] (1) Sample set processing:
[0101] 1.1 Data were collected on three categories of factors: reservoir conditions, crude oil composition, and injected gas composition, including one reservoir parameter, four crude oil parameters, and five injected gas composition parameters. The specific distribution of the data is shown in Table 1.
[0102] Table 1
[0103]
[0104] 1.2 After normalization and outlier removal, 147 valid samples were retained, with 100 samples in the training set and 47 samples in the test set. The distribution is shown in the figure below. Figure 2 As shown.
[0105] (2) Model training and testing:
[0106] 2.1 The baseline models constructed include: Ridge, SVR, RF, GBM, XGBoost, and MLP. The grid search was performed 88 times, 5-fold cross-validation was conducted, and a total of 124 training iterations were performed.
[0107] 2.2 Define the hyperparameter search space for each model, including:
[0108] ①Ridge:'alpha'[0.01,0.1,1,10],'solver'['auto','svd','cholesky'];
[0109] ②SVR: 'C'[0.1,1,10],'gamma'['scale','auto',0.01,0.1];
[0110] ③RandomForest:'n_estimators'[100,200],'max_depth'[None,10,20],'min_samples_split'[2,5];
[0111] ④GBM: 'n_estimators'[100,200], 'learning_rate'[0.01,0.1], 'max_depth'[3,5];
[0112] ⑤XGBoost: 'n_estimators'[100,200],'learning_rate'[0.01,0.05,0.1],'max_depth'[3,5,7],'subsample'[0.8,0.9];
[0113] ⑥MLP: 'hidden_layer_sizes'[(50,),(50,25)], 'alpha'[0.0001,0.001], 'learning_rate_init'[0.001,0.01].
[0114] 2.3 After grid search, the optimal hyperparameters for each model are:
[0115] ①Ridge:alpha=1,solver='auto';
[0116] ②SVR:C=10,gamma=0.1;
[0117] ③RandomForest:n_estimators=200, max_depth=10, min_samples_split=2;
[0118] ④GBM:n_estimators=200, learning_rate=0.01, max_depth=3;
[0119] ⑤XGBoost:n_estimators=200, learning_rate=0.05, max_depth=5, subsample=0.8;
[0120] ⑥MLP: hidden_layer_sizes=(50,25), alpha=0.0001, learning_rate_init=0.001.
[0121] 2.4 Model error comparison results are as follows Figure 3 As shown;
[0122] 2.5 The prediction results of the optimal model are as follows Figure 4 As shown.
[0123] (3) Construction of three-dimensional reservoir model:
[0124] 3.1 Based on indoor experiments, key fluid property data such as crude oil composition, minimum miscibility pressure of CO2 and crude oil, crude oil density, crude oil viscosity, and gas-oil ratio were obtained;
[0125] 3.2 Based on the aforementioned fluid property data, the CMG-Winprop phase model software was used to fit the crude oil phase parameters and minimum miscibility pressure. When the fitting accuracy was higher than 90%, the numerical model of fluid components was output, and the results are as follows. Figure 5 As shown;
[0126] 3.3 Based on the fitted fluid model and reservoir geological data (as shown in Table 2), a three-dimensional reservoir numerical model was established using the CMG-GEM simulator, as follows: Figure 6 As shown.
[0127] Table 2
[0128]
[0129] (4) Generation of dynamic miscibility field:
[0130] Based on a three-dimensional reservoir numerical model, the CO2 component field, temperature field, and pressure field result files were run and output. The influencing factor was the content of impurity gases, which was set to 100% CO2.
[0131] A data extraction and transformation program was written in Python to extract CO2 component field, temperature field, pressure field, and influencing factor data into an input matrix array format;
[0132] Based on the obtained matrix array, the CO2-crude oil miscibility pressure in different grids is predicted by combining the intelligent prediction model of miscibility. The matrix array is uniformly extracted and converted into a matrix array form that can be directly input into the intelligent model (the dimensions correspond one-to-one with the reservoir grid), ensuring the consistency of data spatial coordinates, time nodes and physical dimensions.
[0133] Based on the matrix array obtained from preprocessing, the trained intelligent prediction model for miscibility is called to predict the minimum miscibility pressure (MMP) of CO2-crude oil grid by grid and time node by time node, generating a dynamic MMP data matrix covering the entire reservoir space and simulation cycle.
[0134] The predicted dynamic MMP data matrix was converted into a spatiotemporal grid heatmap using the matplotlib library. Pressure values were encoded using color gradients (e.g., warm colors for high-value areas and cool colors for low-value areas), visually revealing the spatial differentiation characteristics and dynamic evolution patterns of MMPs in different reservoir regions and development stages. The prediction results are as follows: Figure 7 As shown.
[0135] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for intelligent prediction of the dynamic miscibility field of CO2 based on multi-physics coupling, characterized in that, include: Collect multi-scale influencing factors of CO2-crude oil miscibility pressure and construct a sample set; A CO2-crude oil miscibility prediction model was established based on the aforementioned sample set. Collect reservoir geological data and fluid data to establish a three-dimensional reservoir numerical model; Based on the three-dimensional reservoir numerical model and the CO2-crude oil miscibility prediction model, the dynamic miscibility field of CO2-crude oil is calculated and formed in real time.
2. The intelligent prediction method for dynamic miscibility of CO2 based on multiphysics coupling according to claim 1, characterized in that, Data on multi-scale influencing factors of CO2-crude oil miscibility pressure were collected, and a sample set was constructed including: Identify the multi-scale influencing factors affecting the miscibility pressure of CO2-crude oil, collect relevant experimental and simulation data, and construct a multi-dimensional sample database; Statistical characteristic analysis was performed on the multi-dimensional sample database. The Z-score method was used to identify outliers and missing values, and the gradient boosting tree regression imputation method was used to process the outliers and missing values. An improved Min-Max method is used to normalize the processed multi-dimensional sample database, and the normalized multi-dimensional sample database is divided into a training sample set and a test sample set. The improved Min-Max method introduces a dynamic scaling factor, which is dynamically adjusted according to the sensitivity of the variable to the mixing pressure.
3. The intelligent prediction method for dynamic miscibility field of CO2 based on multi-physics coupling according to claim 2, characterized in that, The improved Min-Max method is as follows: Where X is the value of the variable, X min Let X be the minimum value of the variable. max w is the maximum value of the variable. k Let be the sensitivity coefficient of the k-th variable.
4. The intelligent prediction method for dynamic miscibility of CO2 based on multi-physics coupling according to claim 1, characterized in that, The CO2-crude oil miscibility prediction model based on the aforementioned sample set includes: Select several machine learning models as benchmark models; The sample set is input into the benchmark model for hyperparameter fitting, and the benchmark model is optimized for hyperparameters using a grid search strategy to obtain the trained benchmark model. Several evaluation metrics were used to evaluate the performance of the trained benchmark model, and the benchmark model with the best performance was selected as the CO2-crude oil miscibility prediction model.
5. The intelligent prediction method for dynamic miscibility field of CO2 based on multiphysics coupling according to claim 4, characterized in that, The grid search strategy includes: Define a set of candidate values for the hyperparameters of the baseline model and arrange and combine them. For each combination, the model is trained using the sample set, and the generalization ability is evaluated through cross-validation; The combination with the highest cross-validation score is selected as the optimal hyperparameter of the baseline model.
6. The intelligent prediction method for dynamic miscibility field of CO2 based on multi-physics coupling according to claim 1, characterized in that, The collection of reservoir geological and fluid data to establish a three-dimensional reservoir numerical model includes: Fluid property data were obtained through high-temperature and high-pressure visualization experiments; Based on the fluid property data, the phase parameters of crude oil and the minimum miscibility pressure are fitted. When the fitting accuracy reaches the preset accuracy, the numerical model of fluid composition is output. Based on the fluid composition numerical model, the three-dimensional reservoir numerical model is constructed by combining reservoir geological parameters.
7. The intelligent prediction method for dynamic miscibility field of CO2 based on multi-physics coupling according to claim 6, characterized in that, The fluid property data includes crude oil composition, minimum miscibility pressure of CO2 and crude oil, crude oil density, crude oil viscosity, and gas-oil ratio; The reservoir geological parameters include reservoir porosity, reservoir permeability, reservoir temperature, formation pressure, reservoir thickness, reservoir depth, well location information, relative permeability curve, reservoir oil saturation, pressure coefficient, and geothermal gradient.
8. The intelligent prediction method for dynamic miscibility field of CO2 based on multiphysics coupling according to claim 1, characterized in that, Based on the aforementioned three-dimensional reservoir numerical model and CO2-crude oil miscibility prediction model, the dynamic miscibility field of CO2-crude oil is calculated and formed in real time, including: Based on the three-dimensional reservoir numerical model, a multiphysics time series result file of the entire reservoir grid is run in real time and outputs. The multiphysics field includes component fields, temperature fields, and pressure fields. The data in the multiphysics time series results file are converted into a three-dimensional matrix array in batches, and the three-dimensional matrix array is processed by spatial coordinate hash mapping; The processed three-dimensional matrix array is combined with the CO2-crude oil miscibility prediction model for calculation. The minimum miscibility pressure of CO2-crude oil is predicted grid by grid and time step by time to obtain the dynamic CO2-crude oil minimum miscibility pressure data matrix. The dynamic CO2-crude oil minimum miscibility pressure data matrix is then converted into a spatiotemporal grid heat map.