Construction method and device for digital oil reservoir twinning

Abstract

The invention relates to the field of digital twinning, in particular to a digital oil reservoir twinning construction method and device which comprises the following steps that geological data, physical data, real-time production dynamic data and historical data of an oil reservoir are collected through a data collection module; the data preprocessing module is used for cleaning, sorting and verifying the collected original data to ensure the quality and consistency of the collected data; according to the method, firstly, a sensor collects oil reservoir data and caches the data, data preprocessing cleans and formats the collected original data, the obtained formatted data is stored in a database, oil reservoir model building and data processing conduct machine learning optimization on a time sequence analysis result by using a double-layer dynamic optimization method, and the time sequence analysis result is obtained. And the obtained result is used as the input of the oil reservoir model, and the oil reservoir model assists a user in making oil reservoir decision display through a visualization and decision support interface, so that the prediction precision of the oil reservoir model is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of digital twins, and in particular to a method and device for constructing a digital reservoir twin. Background Art

[0002] Reservoir management is a key branch of petroleum engineering, involving reservoir development and production optimization. With advances in computing technology, digital twin technology has become a crucial tool for improving the efficiency and accuracy of reservoir management. By creating virtual replicas of physical entities, digital twin technology allows engineers to simulate and predict reservoir behavior without affecting the actual reservoir. However, existing reservoir models are complex to build, computationally expensive, and difficult to update in real time, limiting their application in reservoir management. Furthermore, in the later stages of oilfield development, traditional reservoir management strategies and technologies often struggle to cope with complex geological conditions and changing production environments. This is particularly true for unconventional reservoirs, whose unique reservoir characteristics and development challenges make traditional reservoir characterization and development methods ineffective. Therefore, oilfield development urgently requires a new technology that enables highly adaptive and intelligent optimization.

[0003] Digital twins, as an emerging technology, focus on building a virtual space that closely mirrors the physical world. This virtual space enables real-time monitoring, analysis, and management of reservoirs. However, the current application of digital twin technology in reservoir management still faces some limitations. For example, existing digital twin models often focus on macroscopic descriptions, failing to capture the complexities of the microscopic scale. This results in a limited understanding of the multi-scale flow characteristics of reservoirs.

[0004] To understand the microscopic characteristics of oil reservoirs, such as pore structure, fracture distribution, and fluid flow behavior at the microscopic scale, more sophisticated models and algorithms are needed. Currently, the market lacks a high-precision method for building digital twin models of oil reservoirs, which limits the accuracy and efficiency of reservoir management.

[0005] Furthermore, reservoir digital twin modeling methods must account for a variety of uncertainties, such as drilling and completion operations, geological parameter uncertainties, and oil and gas price fluctuations. Existing technologies rarely effectively incorporate these uncertainties into digital twin models, limiting their predictive capabilities.

[0006] Existing technologies for supporting reservoir development decisions rely heavily on offline analysis and post-processing, lacking a mechanism that can respond to reservoir status changes in real time and provide decision support. This results in delayed oilfield development decisions and a failure to maximize the economic benefits of the field.

[0007] Therefore, developing a modeling method for a reservoir digital twin system that can achieve detailed characterization of reservoir microscopic characteristics, integration of multiple uncertainty factors, and real-time decision support is of great significance for improving reservoir management level, oilfield development efficiency and economic benefits. Summary of the Invention

[0008] In order to solve the above problems, the present invention provides a method and device for constructing a digital reservoir twin, which can achieve fine characterization of reservoir microscopic characteristics, integration of multiple uncertainty factors, and real-time decision support, and is of great significance for improving reservoir management level, increasing oilfield development efficiency and economic benefits.

[0009] The present invention provides a method for constructing a digital reservoir twin, comprising the following steps: Step 1: Collect reservoir geological data, physical data, real-time production dynamic data and historical data through the data collection module; Step 2: Use the data preprocessing module to clean, organize, and verify the collected raw data to ensure the quality and consistency of the collected data, and perform format conversion; Step 3: The formatted data after the data preprocessing module is stored in the database module, and the historical reservoir data and the parameters of the user's decision are recorded; Step 4: Construct an initial reservoir model through the reservoir model building and data processing module. Use the reservoir's geological, geophysical, and engineering parameters to establish an initial static model of the reservoir. Input the real-time data processed by the two-layer dynamic optimization algorithm into the static basic model to update the model to reflect the dynamic characteristics of the reservoir, improve the accuracy of the model prediction, and complete the dynamic adjustment of the model to realize the reservoir digital twin model. Step 5: Use the visualization and decision support module to display the optimized digital twin model results in a graphical interface to assist users in making efficient decisions.

[0010] Preferably, a two-layer dynamic optimization algorithm includes a time series analysis layer and a machine learning optimization layer, wherein the machine learning optimization layer is based on the time series analysis results; At the time series analysis layer, the two-layer dynamic optimization algorithm first processes the data in the reservoir database through the autoregressive moving average (ARMA) model. The ARMA model is expressed as: ; in, is the value of the reservoir monitoring data at a point in time; is a constant term, is a white noise sequence, representing the model error; is the order of the autoregressive term, which indicates the amount of historical data; is the order of the moving average term, which represents the historical amount of error; is the autoregressive coefficient, which indicates the strength of the relationship with the previous i reservoir monitoring data values; θ i is the moving average coefficient, which indicates the strength of the relationship with the previous j errors.

[0011] Preferably, the machine learning optimization layer in the two-layer dynamic optimization algorithm uses SVM, or support vector machine, for pattern recognition and prediction. The decision function of SVM is: ; In the above formula, w is the weight vector, which indicates the importance of the feature; is the reservoir characteristic vector; T is the transpose; The objective function of SVM is expressed as: ; ; Where w is the weight vector; is the bias term; is the regularization parameter, which controls the complexity of the model; is a slack variable, which means that some data points are allowed to be on the wrong side of the decision boundary; It is the kernel function that maps the original input space to a high-dimensional feature space; x i and y i are training samples and labels; T is the transpose.

[0012] A digital reservoir twin device is obtained using a digital reservoir twin construction method, the device comprising a data acquisition module, a data preprocessing module, a database module, a reservoir model establishment and data processing module, and a visualization and decision support module, which are sequentially communicatively connected; The data acquisition module includes a geological oil pressure production unit, a sensor acquisition unit, and a cache unit connected in sequence; The geological oil pressure production unit is used to collect statistics of physical entity data to be monitored; The sensor acquisition unit is used to acquire the physical data of the oil reservoir to be monitored using sensors; The cache unit is used to cache data collected by the sensor; The data preprocessing module includes a data cleaning unit, a data screening unit, and a data formatting unit connected in sequence; the data cleaning unit is communicatively connected to the cache unit; The data cleaning unit is used to clean the originally collected reservoir data to improve analysis efficiency and data quality; The data screening unit is used to remove irrelevant data and reduce the complexity of the data set; The data formatting unit is used to convert reservoir data into a unified format to facilitate subsequent analysis, processing, storage and sharing; The database module includes the reservoir data storage unit, The reservoir data storage unit is used for formatted data storage, information integration and management, storage of historical data, and recording of real-time production data; The reservoir model establishment and data processing module includes a two-layer dynamic optimization unit, a static reservoir model unit, and a dynamic reservoir model unit which are sequentially connected in communication; The two-layer dynamic optimization unit includes a time series analysis layer and a machine learning optimization layer. In the time series analysis layer, the ARMA model is used to process the collected reservoir monitoring data to identify reservoir patterns and trends, extract key information, and reflect the dynamic behavior of the reservoir and predict the changing trend of the reservoir data. Based on the results obtained in the time series analysis layer, the support vector machine (SVM) model is used to process the data in the machine learning optimization layer to adjust and optimize the reservoir model using the identified patterns and trends, thereby predicting the future state, behavior, and production of the reservoir. The static reservoir model unit is used to establish an initial static reservoir model of the reservoir using data such as the reservoir's geology, geophysical features, and engineering parameters; The dynamic reservoir model unit is used to feed back the dynamic changes of the reservoir to the visualization decision module, so that the user can adjust the parameters according to the prediction results of the reservoir model; The data formatting unit is in communication with the double-layer dynamic optimization unit via the reservoir data storage unit; The visualization and decision support module includes a reservoir visualization unit and a user decision unit that are communicatively connected to each other; Reservoir visualization unit, used to convert complex reservoir data into intuitive graphics and images to improve data comprehensibility; A user decision unit that analyzes data from the reservoir database and visualization unit to help users evaluate different reservoir development and management options; The dynamic reservoir model unit is communicatively connected with the reservoir visualization unit.

[0013] Compared with existing technologies, the beneficial effects brought by this application are: The digital reservoir twin method and device provided by the present invention first collect reservoir data by sensors and cache them, and then perform data preprocessing to clean and format the collected raw data, and store the obtained formatted data in a database. The reservoir model is established and the data is processed by using a two-layer dynamic optimization method, and machine learning optimization is performed on the results of time series analysis. The obtained results are used as the input of the reservoir model, and the reservoir model is displayed through a visualization and decision support interface to assist users in making reservoir decisions, thereby significantly improving the prediction accuracy of the reservoir model. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The figure shows a flowchart of a method for constructing a digital reservoir twin.

[0015] Figure 2 Schematic diagram of a digital reservoir twin device. DETAILED DESCRIPTION

[0016] The present invention provides a method for constructing a digital reservoir twin, comprising the following steps: Step 1: Collect reservoir geological data, physical data, real-time production dynamic data and historical data through the data collection module; Step 2: Use the data preprocessing module to clean, organize, and verify the collected raw data to ensure the quality and consistency of the collected data, and perform format conversion; Step 3: The formatted data after the data preprocessing module is stored in the database module, and the historical reservoir data and the parameters of the user's decision are recorded; Step 4: Construct an initial reservoir model through the reservoir model building and data processing module. Use the reservoir's geological, geophysical, and engineering parameters to establish an initial static model of the reservoir. Input the real-time data processed by the two-layer dynamic optimization algorithm into the static basic model to update the model to reflect the dynamic characteristics of the reservoir, improve the accuracy of the model prediction, and complete the dynamic adjustment of the model to realize the reservoir digital twin model. Step 5: Use the visualization and decision support module to display the optimized digital twin model results in a graphical interface to assist users in making efficient decisions.

[0017] In one embodiment, a two-layer dynamic optimization algorithm includes a time series analysis layer and a machine learning optimization layer, wherein the machine learning optimization layer is based on the time series analysis results; At the time series analysis layer, the two-layer dynamic optimization algorithm first processes the data in the reservoir database through the autoregressive moving average (ARMA) model. The ARMA model is expressed as: ; in, is the value of the reservoir monitoring data at a point in time; is a constant term, is a white noise sequence, representing the model error; is the order of the autoregressive term, which indicates the amount of historical data; is the order of the moving average term, which represents the historical amount of error; is the autoregressive coefficient, which indicates the strength of the relationship with the previous i reservoir monitoring data values; θ iis the moving average coefficient, which indicates the strength of the relationship with the previous j errors.

[0018] In one embodiment, the machine learning optimization layer in the two-layer dynamic optimization algorithm uses SVM for pattern recognition and prediction. The decision function of SVM is: ; In the above formula, w is the weight vector, which indicates the importance of the feature; is the reservoir characteristic vector, T is the transpose; The objective function of SVM is expressed as: ; ; Where w is the weight vector; is the bias term; is the regularization parameter, which controls the complexity of the model; is a slack variable, which means that some data points are allowed to be on the wrong side of the decision boundary; It is the kernel function that maps the original input space to a high-dimensional feature space; x i and y i are training samples and labels; T is the transpose.

[0019] A digital reservoir twin device is obtained using a digital reservoir twin construction method, the device comprising a data acquisition module, a data preprocessing module, a database module, a reservoir model establishment and data processing module, and a visualization and decision support module, which are sequentially communicatively connected; The data acquisition module includes a geological oil pressure production unit, a sensor acquisition unit, and a cache unit connected in sequence; The geological oil pressure production unit is used to collect statistics of physical entity data to be monitored; The sensor acquisition unit is used to acquire the physical data of the oil reservoir to be monitored using sensors; The cache unit is used to cache data collected by the sensor; The data preprocessing module includes a data cleaning unit, a data screening unit, and a data formatting unit connected in sequence; the data cleaning unit is communicatively connected to the cache unit; The data cleaning unit is used to clean the originally collected reservoir data to improve analysis efficiency and data quality; The data screening unit is used to remove irrelevant data and reduce the complexity of the data set; The data formatting unit is used to convert reservoir data into a unified format to facilitate subsequent analysis, processing, storage and sharing; The database module includes the reservoir data storage unit, The reservoir data storage unit is used for formatted data storage, information integration and management, storage of historical data, and recording of real-time production data; The reservoir model establishment and data processing module includes a two-layer dynamic optimization unit, a static reservoir model unit, and a dynamic reservoir model unit which are sequentially connected in communication; The two-layer dynamic optimization unit includes a time series analysis layer and a machine learning optimization layer. In the time series analysis layer, the ARMA model is used to process the collected reservoir monitoring data to identify reservoir patterns and trends, extract key information, and reflect the dynamic behavior of the reservoir and predict the changing trend of the reservoir data. Based on the results obtained in the time series analysis layer, the support vector machine (SVM) model is used to process the data in the machine learning optimization layer to adjust and optimize the reservoir model using the identified patterns and trends, thereby predicting the future state, behavior, and production of the reservoir. The static reservoir model unit is used to establish an initial static reservoir model of the reservoir using data such as the reservoir's geology, geophysical features, and engineering parameters; The dynamic reservoir model unit is used to feed back the dynamic changes of the reservoir to the visualization decision module, so that the user can adjust the parameters according to the prediction results of the reservoir model; The data formatting unit is in communication with the double-layer dynamic optimization unit via the reservoir data storage unit; The visualization and decision support module includes a reservoir visualization unit and a user decision unit that are communicatively connected to each other; Reservoir visualization unit, used to convert complex reservoir data into intuitive graphics and images to improve data comprehensibility; A user decision unit that analyzes data from the reservoir database and visualization unit to help users evaluate different reservoir development and management options; The dynamic reservoir model unit is communicatively connected with the reservoir visualization unit.

[0020] Example 1 See also Figure 1 ,The reservoir digital twin construction method includes the following steps: S100. First, in the data acquisition module, in the geological, oil pressure and production unit, determine the collected data, such as geology, oil pressure, production, historical data, temperature, etc. In the sensor acquisition unit, collect the oil pressure, production, geology and historical data, and store the collected data in the cache unit. The data format includes a timestamp, the sensor corresponding to the data source, and the measured value. The measured value includes oil pressure, production, geology, historical data, etc.

[0021] S200. Secondly, in the data preprocessing module, data is taken out from the cache unit for cleaning, and duplicate rows in the data are deleted to prevent excessive weighting of certain information. Missing data is processed by deleting records with missing values, filling missing values ​​with the mean, median, mode or using a prediction model, or ignoring missing values, and detecting and correcting errors and outliers in the data set.

[0022] The cleaned data is then formatted uniformly into timestamp, oil pressure field, production field, geological field, and historical data field formats, and the unified field data is stored in the database.

[0023] S300, please refer to Figure 2 , formatted data is retrieved from the database and analyzed using a two-layer dynamic algorithm. Taking oil production as an example, in the first layer of time series analysis, the ARMA (p, q) model is used to process historical reservoir production data to predict future reservoir production. The ARMA model can be expressed as: ; in, is the value of the reservoir production data at the prediction timestamp; is a constant term; is a white noise sequence; is the order of the autoregressive term, which determines how far back in time the historical oil production data should be looked; is the order of the moving average term, which determines how many previous reservoir production forecast errors will be used to predict the current reservoir production value; is the autoregressive coefficient, which represents the relationship between the current reservoir production value and its previous i reservoir production historical values ​​in the time series; θ i is the coefficient of the moving average term, which represents the relationship between the error in predicted reservoir production and its previous j historical values ​​of reservoir production.

[0024] Taking oil reservoir production as an example, if we set p=2, q=2 in the ARMA(p,q) model and process the oil reservoir production data, we can get the predicted value of oil reservoir production at time stamp t. , ; is the autoregressive part, is the moving average part, and the final result is , thus we get the predicted value of reservoir production at time t when p=2, q=2. In this application, the maximum likelihood estimation method is used to estimate the ARMA model. The value of and fitting of historical reservoir production data, the oil pressure, geology, water injection volume and other data can be predicted using an ARMA model similar to the reservoir production.

[0025] In the second layer of the two-layer dynamic optimization algorithm, the machine learning optimization layer, based on the reservoir production results predicted by the time series analysis layer, predicts continuous changes in the reservoir, such as production volume or remaining oil volume. That is, regression analysis is performed and the SVM model is used for pattern recognition and prediction. The decision function of the SVM is: ; Where w is the weight vector; T is the transpose; is the reservoir eigenvector, which is composed of the eigenvalues ​​predicted by ARMA, i.e., the reservoir data consists of , is the ARMA predicted reservoir production, where are other features related to reservoir production, including formation pressure, oil pressure, water injection volume, and temperature, which together constitute the feature vector in the SVM decision function , is the bias term, and the objective function of SVM can be expressed as: ; ; Among them, w is the weight vector, which determines the decision boundary of the model; is a bias term that affects the location of the decision boundary, is a regularization parameter used to control the complexity of the model to avoid overfitting; is a slack variable that allows some reservoir production data points to be on the wrong side of the decision boundary; It is the kernel function that maps the original input space to a high-dimensional feature space; x i and y i is the feature vector and label of the training sample; label y i Usually the value is +1 or -1, representing two different categories, y i It is a parameter given in the training data set, which is pre-labeled and indicates the category to which each training sample belongs; T is the transpose.

[0026] The SVM model is trained by inputting reservoir feature vectors to determine the optimal kernel function. In this application, the radial basis function (RBF) kernel function is used, i.e. ; Among them, σ is the width parameter of RBF, which controls the radial range of the function.

[0027] S400, complete the prediction through SVM and get the result ,Will Applied to digital twin reservoir devices, the geological and fluid core dynamic characteristic models of the reservoir are updated to achieve adjustment and optimization of the reservoir development process.

[0028] S500, the optimized reservoir model is converted into a graphical interface, which displays reservoir production, oil pressure, temperature, historical data, etc. to assist decision makers in making efficient decisions.

[0029] Based on the abnormal oil pressure value warning, the decision parameters are input in time to make a decision plan for oil reservoir exploitation. The decision parameters are transmitted to the oil reservoir digital twin model and the oil reservoir digital twin device is updated.

[0030] Compared with existing technologies, the beneficial effects brought by this application are: The present invention provides a method and device for constructing a digital reservoir twin. First, sensors collect reservoir data and cache it. Data preprocessing cleans and formats the collected raw data, and the obtained formatted data is stored in a database. The reservoir model is established and data processing is carried out by using a two-layer dynamic optimization method. Machine learning optimization is performed on the results of time series analysis. The obtained results are used as the input of the reservoir model. The reservoir model is displayed through a visualization and decision support interface to assist users in making reservoir decisions, thereby significantly improving the prediction accuracy of the reservoir model.

[0031] The following is an implementation of a digital reservoir twin device, which is used to execute the various execution steps and corresponding technical effects of the reservoir digital twin construction method shown in the above embodiments and possible implementations. Figure 2 As shown, the device includes a data acquisition module, a data preprocessing module, a database module, a reservoir model establishment and data processing module, and a visualization and decision support module. The data acquisition module includes a geological oil pressure production unit, a sensor acquisition unit, and a cache unit. The data preprocessing module includes a data cleaning unit, a data screening unit, and a data formatting unit. The database module includes a reservoir data storage unit. The reservoir model establishment and data processing module includes a two-layer dynamic optimization unit, a static reservoir model unit, and a dynamic reservoir model unit. The visualization and decision support module includes a reservoir visualization unit and a user decision unit. The geological oil pressure production unit is the physical entity data to be monitored in the real world, which is collected by sensors; The sensor acquisition unit obtains reservoir data, such as oil pressure, oil production, temperature, and water injection volume, from the geological oil pressure and production unit, and sends the obtained data to the cache unit for caching; The reservoir data is taken out from the cache unit and sent to the data preprocessing module. First, it is cleaned by the data cleaning unit. The cleaned data is sent to the data screening unit for screening. The screened data is then sent to the data formatting unit for data format unification. The unified data is then stored in the reservoir data storage unit in the database module; Then, the formatted data is taken out from the reservoir data storage unit and sent to the double-layer dynamic optimization unit for data processing. The processed results are sent to the initial static reservoir model unit to update the dynamic reservoir model. The dynamic reservoir model unit feeds back the predicted dynamic changes of future reservoir data to the visualization decision module, allowing users to adjust parameters based on the reservoir model prediction results; The reservoir visualization unit converts the complex reservoir data of the reservoir dynamic model unit into intuitive graphics and images, improving the comprehensibility of the data and delivering the graphical interface to the user decision-making unit; In the user decision unit, the user observes the changes in reservoir data displayed on the visualization interface, makes reasonable decisions, inputs parameters, and feeds them back to the reservoir visualization unit, which then transmits them to the dynamic reservoir model for real-time updating of the reservoir model.

Claims

1. A method for constructing a digital reservoir twin, characterized in that: The following steps are involved: Step 1: Collect reservoir geological data, physical data, real-time production dynamic data and historical data through the data collection module; Step 2: Use the data preprocessing module to clean, organize, and verify the collected raw data to ensure the quality and consistency of the collected data, and perform format conversion; Step 3: The formatted data after the data preprocessing module is stored in the database module, and the historical reservoir data and the parameters of the user's decision are recorded; Step 4: Construct an initial reservoir model through the reservoir model building and data processing module. Use the reservoir's geological, geophysical, and engineering parameters to establish an initial static model of the reservoir. Input the real-time data processed by the two-layer dynamic optimization algorithm into the static basic model to update the model to reflect the dynamic characteristics of the reservoir, improve the accuracy of the model prediction, and complete the dynamic adjustment of the model to realize the reservoir digital twin model. Step 5: Use the visualization and decision support module to display the optimized digital twin model results in a graphical interface to assist users in making efficient decisions.

2. The method for constructing a digital reservoir twin according to claim 1, wherein: The two-layer dynamic optimization algorithm includes a time series analysis layer and a machine learning optimization layer, and the machine learning optimization layer is based on the time series analysis results; At the time series analysis layer, the two-layer dynamic optimization algorithm first processes the data in the reservoir database through the autoregressive moving average (ARMA) model. The ARMA model is expressed as: ； in, is the value of the reservoir monitoring data at a point in time, is a constant term, is a white noise sequence, representing the model error; is the order of the autoregressive term, which indicates the amount of historical data; is the order of the moving average term, which represents the historical amount of error; is the autoregressive coefficient, which indicates the strength of the relationship with the previous i reservoir monitoring data values; θ i is the moving average coefficient, which indicates the strength of the relationship with the previous j errors.

3. The method for constructing a digital reservoir twin according to claim 2, wherein: The machine learning optimization layer in the two-layer dynamic optimization algorithm uses SVM for pattern recognition and prediction. The decision function of SVM is: ； In the above formula, w is a weight vector, indicating the importance of the feature; is the reservoir characteristic vector; T is transpose; The objective function of SVM is expressed as: ； ； Among them, w is the weight vector; b is the bias term; is the regularization parameter, which controls the complexity of the model; is a slack variable, which means that some data points are allowed to be on the wrong side of the decision boundary; It is the kernel function that maps the original input space to a high-dimensional feature space; x i and y i are training samples and labels; T is the transpose.

4. A digital reservoir twin device, obtained using the method for constructing a digital reservoir twin according to any one of claims 1 to 3, the device comprising a data acquisition module, a data preprocessing module, a database module, a reservoir model establishment and data processing module, and a visualization and decision support module, all of which are communicatively connected in sequence; in, The data acquisition module includes a geological oil pressure production unit, a sensor acquisition unit, and a cache unit which are connected in sequence; The geological oil pressure production unit is used to collect statistics of physical entity data to be monitored; The sensor acquisition unit is used to acquire the physical data of the oil reservoir to be monitored using sensors; The cache unit is used to cache data collected by the sensor; The data preprocessing module includes a data cleaning unit, a data screening unit, and a data formatting unit connected in sequence; the data cleaning unit is communicatively connected to the cache unit; The data cleaning unit is used to clean the originally collected reservoir data to improve analysis efficiency and data quality; The data screening unit is used to remove irrelevant data and reduce the complexity of the data set; The data formatting unit is used to convert reservoir data into a unified format to facilitate subsequent analysis, processing, storage and sharing; The database module includes the reservoir data storage unit, The reservoir data storage unit is used for formatted data storage, information integration and management, storage of historical data, and recording of real-time production data; The reservoir model establishment and data processing module includes a two-layer dynamic optimization unit, a static reservoir model unit, and a dynamic reservoir model unit which are sequentially connected in communication; The dual-layer dynamic optimization unit includes a time series analysis layer and a machine learning optimization layer. In the time series analysis layer, the ARMA model is used to process the collected reservoir monitoring data to identify reservoir patterns and trends, extract key information, and reflect the dynamic behavior of the reservoir and predict the changing trend of the reservoir data. Based on the results from the time series analysis layer, the machine learning optimization layer processes the data using the SVM model to adjust and optimize the reservoir model using the identified patterns and trends, thereby predicting the future state, behavior, and production of the reservoir. The static reservoir model unit is used to establish an initial static reservoir model of the reservoir using data such as the reservoir's geology, geophysical features, and engineering parameters; The dynamic reservoir model unit is used to feed back the dynamic changes of the reservoir to the visualization decision module, so that the user can adjust the parameters according to the prediction results of the reservoir model; The data formatting unit is in communication with the double-layer dynamic optimization unit via the reservoir data storage unit; The visualization and decision support module includes a reservoir visualization unit and a user decision unit that are communicatively connected to each other; Reservoir visualization unit, used to convert complex reservoir data into intuitive graphics and images to improve data comprehensibility; A user decision unit that analyzes data from the reservoir database and visualization unit to help users evaluate different reservoir development and management options; The dynamic reservoir model unit is communicatively connected with the reservoir visualization unit.

Images