Oil reservoir crack intelligent monitoring system and method based on microseismic image

Abstract

The invention relates to the technical field of oil reservoir crack detection systems, in particular to an oil reservoir crack intelligent monitoring system and method based on a micro-seismic image, and the system comprises a micro-seismic monitoring and transmission module, a micro-seismic image processing and model fusion module which is in communication connection with the micro-seismic monitoring and transmission module; the micro-seismic intelligent decision-making and auxiliary system module is in communication connection with the micro-seismic image processing and model fusion module; the micro-seismic data and model intelligent optimization module is in communication connection with the micro-seismic intelligent decision-making and auxiliary system module; correcting the microseismic data by using a transfer learning method; optimizing the oil reservoir geomechanical model by using the optimization function; and the data platform module is in communication connection with the micro-seismic data and model intelligent optimization module. According to the invention, a complete oil reservoir crack intelligent monitoring ecological system is constructed. Seamless connection and collaborative optimization of all links are achieved, and the operation efficiency and reliability of the system are greatly improved through the integral design.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir fracture detection systems, in particular to an intelligent monitoring system for reservoir fractures based on microseismic images and a method thereof. Background Art

[0002] As oil and gas exploration and development expands to complex geological conditions and unconventional oil and gas reservoirs, it becomes increasingly important to accurately monitor and predict the development of reservoir fractures. Traditional reservoir fracture monitoring methods mainly rely on logging, seismic and other technologies, but these methods often have problems such as low resolution, poor real-time performance and high cost. In recent years, microseismic monitoring technology has been widely used in the field of reservoir fracture monitoring due to its high sensitivity and real-time performance.

[0003] However, existing microseismic monitoring technology still faces many challenges. First, microseismic signals are usually weak and easily interfered by noise, resulting in low positioning accuracy. Second, traditional microseismic data processing methods are difficult to effectively process large-scale, high-dimensional data, which restricts the accurate characterization of complex fracture networks. Furthermore, existing fracture prediction models are mostly based on simplified geomechanical models, which are difficult to accurately reflect the true dynamic changes of oil reservoirs. In addition, the geological conditions between different oil reservoirs vary greatly, and how to transfer the empirical knowledge of existing oil reservoirs to new oil reservoirs is also a problem that needs to be solved urgently.

[0004] The closest existing technologies usually use independent microseismic monitoring systems and data processing platforms, lacking the organic integration of microseismic data, geomechanical models, and historical development experience. This fragmented processing method leads to insufficient information utilization, making it difficult to achieve intelligent and refined monitoring and prediction of reservoir fractures. At the same time, existing technologies also have deficiencies in real-time and scalability, making it difficult to meet the needs of large-scale oilfield development. Summary of the invention

[0005] The present invention aims to solve the above technical problems and provide an intelligent monitoring system for oil reservoir fractures based on microseismic images and its method. The system realizes high-precision, real-time monitoring and prediction of oil reservoir fractures by innovatively integrating advanced technologies such as microseismic monitoring, graph neural networks, and deep reinforcement learning.

[0006] The present invention proposes an intelligent monitoring system for oil reservoir fractures based on microseismic images, comprising:

[0007] Microseismic monitoring and transmission module for:

[0008] Collect microseismic wave data through underground microseismic sensor networks and ground microseismic sensor networks;

[0009] Transmit and integrate the collected microseismic wave data;

[0010] The microseismic image processing and model fusion module is connected to the microseismic monitoring and transmission module for:

[0011] Receiving microseismic wave data sent by the microseismic monitoring and transmission module;

[0012] Based on the microseismic wave data, a fracture extension prediction model is constructed using a graph neural network;

[0013] The microseismic intelligent decision-making and auxiliary system module is connected to the microseismic image processing and model fusion module for:

[0014] Receiving the fracture extension prediction model data sent by the microseismic image processing and model fusion module;

[0015] Based on the fracture propagation prediction model data and in combination with the reservoir geomechanics model, a deep reinforcement learning algorithm is used to perform microseismic data preprocessing, reservoir geomechanics evolution analysis, and fracture propagation prediction;

[0016] The microseismic data and model intelligent optimization module is connected to the microseismic intelligent decision-making and auxiliary system module for:

[0017] Receiving microseismic data and reservoir geomechanical model data sent by the microseismic intelligent decision-making and auxiliary system module;

[0018] Correction of microseismic data using transfer learning methods;

[0019] Use optimization functions to optimize reservoir geomechanical models;

[0020] The data platform module is connected to the microseismic data and model intelligent optimization module for:

[0021] Receiving and storing the microseismic data and optimized data sent by the model intelligent optimization module;

[0022] Perform data fusion to form atlas data;

[0023] Data is synchronized locally and in the cloud.

[0024] Preferably, the microseismic monitoring and transmission module comprises:

[0025] Downhole microseismic sensor network, used to deploy optical fiber sensor network underground to collect microseismic waves;

[0026] A ground microseismic sensor network is used to deploy microseismic sensors on the ground to transmit microseismic waves;

[0027] The gateway is connected to the underground microseismic sensor network and the ground microseismic sensor network for integrating the underground and ground microseismic data and transmitting the integrated data to the microseismic image processing and model fusion module.

[0028] Preferably, the microseismic image processing and model fusion module comprises:

[0029] A three-dimensional microseismic positioning module is used to locate the microseismic points in the well based on the microseismic waves detected by the downhole optical fiber sensor network and the ground microseismic sensors;

[0030] A microseismic imaging module, which is in communication connection with the three-dimensional microseismic positioning module, and is used to perform microseismic imaging through multiple three-dimensional microseismic positioning and to perform crack identification on the microseismic data map;

[0031] The fracture extension prediction module based on graph neural network is communicatively connected with the microseismic imaging module and is used to construct a graph neural network through microseismic images to predict fracture extension.

[0032] Preferably, the microseismic intelligent decision-making and auxiliary system module includes:

[0033] Microseismic data intelligent processing module, used for denoising, preprocessing and standardization of microseismic data;

[0034] A reservoir geomechanical evolution analysis module, which is in communication connection with the microseismic data intelligent processing module and is used to input microseismic data, fracture images and reservoir geomechanical models into a deep reinforcement learning algorithm to analyze the reservoir geomechanical evolution process;

[0035] The fracture extension prediction module is communicatively connected with the reservoir geomechanical evolution analysis module and is used for predicting fracture extension based on graph neural network.

[0036] Preferably, the microseismic data and model intelligent optimization module comprises:

[0037] Microseismic data correction module, used to correct microseismic data through transfer learning method;

[0038] The reservoir geomechanics model optimization module is in communication with the microseismic data correction module and is used for performing model optimization using an optimization function.

[0039] Preferably, the data platform module includes:

[0040] Data storage module, used to store data from microseismic monitoring and transmission, microseismic image processing and model fusion, microseismic intelligent decision-making and auxiliary systems in a relational database format;

[0041] A data fusion module, which is in communication with the data storage module and is used to read data from the data storage system and fuse the microseismic data, the reservoir geomechanics model and the graph neural network algorithm to form graph data;

[0042] The algorithm server is connected to the data fusion module for performing atlas data calculation, combining the microseismic data with the reservoir geomechanical model, and transmitting the calculated data back to the data storage system.

[0043] Preferably, the data storage module and the data fusion module are composed of two parts, local and cloud, and the data are kept consistent through synchronization; wherein the local data can be updated in real time, and the cloud data can be synchronized in real time.

[0044] Preferably, a knowledge graph module is further included, which is communicatively connected with the microseismic intelligent decision-making and auxiliary system module and is used for:

[0045] Construct a reservoir management knowledge database based on the distribution of historical development wells;

[0046] Combined with the earthquake information obtained by the microseismic monitoring and transmission module, the knowledge graph technology is used to generate the best development plan recommendation;

[0047] Assist microseismic intelligent decision-making and assist system modules in making decisions.

[0048] Preferably, the knowledge graph module includes:

[0049] A historical data extraction unit, used for extracting the distribution of development wells from historical data;

[0050] A geomechanical parameter analysis unit, which is in communication with the historical data extraction unit and is used to select corresponding geomechanical parameters from the historical data to analyze the rationality of the historical development well location;

[0051] A high permeability zone judgment unit is connected to the geomechanical parameter analysis unit for calculating a rock mechanics parameter index. If the index is greater than a set value, the corresponding position is judged to be a high permeability zone.

[0052] The development plan recommendation unit is communicatively connected to the high permeability area judgment unit, and is used to generate an optimal development plan recommendation based on the high permeability area judgment result.

[0053] An intelligent monitoring method for reservoir fractures based on microseismic images based on the system,

[0054] S1. The microseismic sensor collects and transmits microseismic waves, and uploads the microseismic wave data to the data platform for storage;

[0055] S2. Preprocess the microseismic wave data on the data platform and construct a graph neural network for microseismic images to predict crack extension;

[0056] S3. Input microseismic data, fracture images, and reservoir geomechanical models into deep reinforcement learning to analyze the reservoir geomechanical evolution process and predict fracture expansion to adjust the mining plan;

[0057] Wherein, step S1 specifically includes:

[0058] S1.1. Arrange a fiber optic sensor network and deploy a microseismic detection array underground;

[0059] S1.2. Collect raw microseismic wave data and ground microseismic wave data;

[0060] S1.3. Upload the collected microseismic wave data to the data platform for storage;

[0061] Step S2 specifically includes:

[0062] S2.1. On the data platform, the uploaded raw microseismic wave data is denoised and microseismic positioning is performed;

[0063] S2.2. Perform three-dimensional imaging on the microseismic wave data to obtain a microseismic image;

[0064] S2.3. Identify cracks in microseismic images, establish a graph neural network for microseismic images, and predict crack extension;

[0065] Step S3 specifically includes:

[0066] S3.1. Input microseismic data, fracture images and reservoir geomechanical models into the reinforcement learning algorithm module for calculation;

[0067] S3.2. Use transfer learning algorithms to calibrate microseismic data and optimize reservoir geomechanical models;

[0068] S3.3. Use the optimized microseismic data and reservoir geomechanical model to construct a graph neural network to predict the fracture propagation path;

[0069] S3.4. Predict the trend of fracture expansion and adjust the mining plan based on changes in microseismic data and reservoir geomechanical models.

[0070] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0071] First, from a macroscopic perspective, the present invention builds a complete intelligent monitoring ecosystem for reservoir fractures. The system covers the entire process from data collection, processing, analysis to decision support, achieving seamless connection and coordinated optimization of each link. This holistic design greatly improves the operating efficiency and reliability of the system, and provides strong technical support for the refined management of oil and gas fields.

[0072] Secondly, the present invention cleverly solves multiple technical problems at the microscopic level. For example, by combining the downhole optical fiber sensor network and the ground microseismic sensor network, the acquisition quality and coverage of microseismic signals are significantly improved. For another example, the use of graph neural networks to process microseismic data effectively captures the topological structure and evolution characteristics of complex fracture networks, overcoming the limitations of traditional methods in processing high-dimensional non-Euclidean data.

[0073] Furthermore, the various modules of the present invention achieve a high degree of synergy and complementarity. The microseismic image processing and model fusion module provides high-quality input data for the microseismic intelligent decision-making and auxiliary system module, and the latter further optimizes the output of the former through the deep reinforcement learning algorithm. This virtuous cycle continuously improves the performance of the entire system.

[0074] In addition, the present invention cleverly solves some technical contradictions in reservoir monitoring. For example, by introducing the transfer learning method, the generalization ability of the model is guaranteed, and the high cost of retraining the model in new reservoirs is avoided. For another example, the dual data storage mechanism of local and cloud data storage is adopted, which not only ensures the real-time data processing, but also improves the reliability and scalability of the system.

[0075] Finally, the effect of the present invention is not only reflected in the independent performance improvement of each module, but also in the synergistic effect of the overall system. For example, the introduction of the knowledge graph module not only improves the scientific nature of decision-making, but also injects historical experience and expert knowledge into the entire system, realizing the organic combination of data-driven and knowledge-driven. This multi-dimensional and multi-level synergistic effect enables the present invention to achieve a qualitative leap in the accuracy, efficiency and intelligence level of reservoir fracture monitoring.

[0076] In summary, the present invention has comprehensively improved the accuracy, efficiency and intelligence level of reservoir fracture monitoring through innovative system design and ingenious application of advanced algorithms, providing a new technical path for the efficient development of oil and gas resources, which has important theoretical significance and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a logic block diagram of the whole system of the present invention.

[0078] Figure 2 It is a logic block diagram of the microseismic monitoring and transmission module of the present invention.

[0079] Figure 3 It is a logic block diagram of the microseismic image processing and model fusion module of the present invention.

[0080] Figure 4 It is a logic block diagram of the microseismic intelligent decision-making and auxiliary system module of the present invention.

[0081] Figure 5 This is a logical block diagram of the data platform module of the present invention.

[0082] Figure 6 This is a logical block diagram of the knowledge graph module of the present invention. DETAILED DESCRIPTION

[0083] See also Figure 1-6 The present invention provides an intelligent monitoring system for oil reservoir fractures based on microseismic images and a method thereof. The system realizes high-precision monitoring and prediction of oil reservoir fractures by innovatively combining microseismic monitoring technology, graph neural networks, deep reinforcement learning and other advanced algorithms.

[0084] First, the system of the present invention includes a microseismic monitoring and transmission module 1. The module 1 is used to collect microseismic wave data through an underground microseismic sensor network 11 and a ground microseismic sensor network 12, and to transmit and integrate the collected microseismic wave data. Preferably, the underground microseismic sensor network 11 uses optical fiber sensors to form a dense monitoring network underground. This arrangement can capture more weak seismic signals and improve the sensitivity of monitoring. The ground microseismic sensor network 12 can use conventional seismic detectors, which are usually arranged in an array to obtain better signal coverage. The gateway is communicatively connected to the underground microseismic sensor network 11 and the ground microseismic sensor network 12, and is used to integrate the microseismic data underground and on the ground, and transmit the integrated data to the microseismic image processing and model fusion module 2.

[0085] The microseismic monitoring and transmission module 1 is connected to the microseismic image processing and model fusion module 2. Module 2 receives the microseismic wave data sent by module 1, and based on these data, constructs a fracture extension prediction model using a graph neural network. Specifically, the microseismic image processing and model fusion module 2 includes a three-dimensional microseismic positioning module 21, a microseismic imaging module 22, and a fracture extension prediction module 23 based on a graph neural network.

[0086] The three-dimensional microseismic positioning module 21 mainly locates the microseismic points in the well based on the microseismic waves detected by the downhole optical fiber sensor network 11 and the microseismic waves detected by the ground microseismic sensor network 12. This positioning method combining downhole and ground data can significantly improve the positioning accuracy. In one embodiment, the positioning algorithm can adopt the double difference positioning method, and its basic principle can be expressed as:

[0087]

[0088] in, is the arrival time difference between events i and j recorded by the kth station, and are the theoretical arrival times of events i and j at the kth station, and are the occurrence times of events i and j respectively. By minimizing the double difference residuals of all stations, a more accurate earthquake source location can be obtained.

[0089] The microseismic imaging module 22 performs microseismic imaging through multiple three-dimensional microseismic positioning and performs crack identification on the microseismic data map. A commonly used microseismic imaging method is elastic wave inversion, and its objective function can be expressed as:

[0090]

[0091] Among them, m is the model parameter, d o bs and d c al are observed data and simulated data respectively, N s and N r are the number of sources and receivers, respectively, and T is the recording length. By iteratively optimizing this objective function, the spatial distribution of underground medium properties can be obtained, thereby achieving microseismic imaging.

[0092] The fracture extension prediction module 23 based on graph neural network is an innovation of the present invention. The module constructs a graph neural network through microseismic images to predict fracture extension. In one embodiment, a graph convolutional neural network (GCN) can be used to predict fracture extension. The core operation of GCN can be expressed as:

[0093]

[0094] Among them, H (l) is the feature matrix of the lth layer, To add a self-connected adjacency matrix, for The degree matrix, W (l) is a learnable weight matrix, and σ is an activation function. By stacking multiple layers of GCN, the topological structure and evolution characteristics of the crack can be effectively captured, thereby achieving high-precision crack propagation prediction.

[0095] The microseismic intelligent decision-making and auxiliary system module 3 is in communication connection with the microseismic image processing and model fusion module 2. Module 3 receives the fracture extension prediction model data sent by module 2, and based on these data, combined with the reservoir geomechanics model, uses a deep reinforcement learning algorithm to perform microseismic data preprocessing, reservoir geomechanics evolution analysis, and fracture extension prediction.

[0096] In practical applications, the system of the present invention can be flexibly configured according to specific reservoir conditions and monitoring requirements. For example, for shallow reservoirs, the density of ground microseismic sensors can be increased; for deep reservoirs, more downhole fiber optic sensors can be arranged at key locations. In addition, the data transmission rate between the various modules of the system also needs to be optimized according to actual conditions to ensure the real-time performance and reliability of the entire system.

[0097] Through the above-mentioned innovative system design and algorithm application, the intelligent monitoring system for reservoir fractures based on microseismic images of the present invention can realize all-round, high-precision monitoring and prediction of reservoir fractures, and provide strong technical support for the efficient development of oil and gas fields. The system of the present invention also includes a microseismic data and model intelligent optimization module 4, which is connected to the microseismic intelligent decision-making and auxiliary system module 3. The main function of module 4 is to receive the microseismic data and reservoir geomechanical model data sent by module 3, use the transfer learning method to correct the microseismic data, and use the optimization function to optimize the reservoir geomechanical model.

[0098] In a preferred embodiment of the present invention, the microseismic data and model intelligent optimization module 4 includes a microseismic data correction module 41 and a reservoir geomechanical model optimization module 42. The microseismic data correction module 41 uses a transfer learning method to correct the microseismic data. This method is particularly suitable for processing data differences between different reservoirs and can effectively improve the generalization ability of the model. For example, domain adaptation technology can be used to transfer knowledge from the source domain (reservoirs with sufficient annotated data) to the target domain (new reservoirs). Specifically, a deep domain adaptation network (Deep Adaptation Network, DAN) can be used, and its loss function can be expressed as:

[0099]

[0100] Among them, H (l) is the feature matrix of the lth layer, To add a self-connected adjacency matrix, for The degree matrix, W (l) is a learnable weight matrix, and σ is an activation function. By stacking multiple layers of GCN, the topological structure and evolution characteristics of the crack network can be effectively captured, thereby achieving high-precision crack propagation prediction.

[0101] The microseismic intelligent decision-making and auxiliary system module 3 is in communication connection with the microseismic image processing and model fusion module 2. Module 3 receives the fracture extension prediction model data sent by module 2, and based on these data, combined with the reservoir geomechanics model, uses a deep reinforcement learning algorithm to perform microseismic data preprocessing, reservoir geomechanics evolution analysis, and fracture extension prediction.

[0102] In practical applications, the system of the present invention can be flexibly configured according to specific reservoir conditions and monitoring requirements. For example, for shallow reservoirs, the density of ground microseismic sensors can be increased; for deep reservoirs, more downhole fiber optic sensors can be arranged at key locations. In addition, the data transmission rate between the various modules of the system also needs to be optimized according to actual conditions to ensure the real-time performance and reliability of the entire system.

[0103] Through the above-mentioned innovative system design and algorithm application, the microseismic image-based intelligent reservoir fracture monitoring system of the present invention can realize all-round, high-precision monitoring and prediction of reservoir fractures, providing strong technical support for the efficient development of oil and gas fields.

[0104] The system of the present invention also includes a microseismic data and model intelligent optimization module 4, which is in communication with the microseismic intelligent decision-making and auxiliary system module 3. The main function of module 4 is to receive the microseismic data and reservoir geomechanical model data sent by module 3, correct the microseismic data using the transfer learning method, and optimize the reservoir geomechanical model using the optimization function.

[0105] In a preferred embodiment of the present invention, the microseismic data and model intelligent optimization module 4 includes a microseismic data correction module 41 and a reservoir geomechanical model optimization module 42. The microseismic data correction module 41 uses a transfer learning method to correct the microseismic data. This method is particularly suitable for processing data differences between different reservoirs and can effectively improve the generalization ability of the model. For example, domain adaptation technology can be used to transfer knowledge from the source domain (reservoirs with sufficient annotated data) to the target domain (new reservoirs). Specifically, a deep domain adaptation network (DeepAdaptation Network, DAN) can be used, and its loss function can be expressed as:

[0106]

[0107] Among them, L c is the classification loss, X s and Y s are the input and label of the source domain respectively, and MMD is the maximum mean difference, which is used to measure the source domain characteristics. and target domain features The distribution difference, λ is the balance parameter. By minimizing this loss function, the model can learn feature representations that have good performance in both the source domain and the target domain. The reservoir geomechanics model optimization module 42 uses the optimization function to optimize the model. In another embodiment of the present invention, a Bayesian optimization method can be used. This method can find the global optimal solution within a small number of iterations, and is particularly suitable for reservoir model optimization with high computational costs. The core of Bayesian optimization is to construct a probability model of the objective function, usually using a Gaussian process:

[0108] f(x)~GP(m(x),k(x,x′)),

[0109] Where m(x) is the mean function and k(x,x′) is the kernel function. During the optimization process, the next evaluation point is selected by maximizing the acquisition function (such as expected improvement, E):

[0110] EI(x)=E[max(f(x)-f(x) + ),0)],

[0111] Among them, x + This method can achieve a good balance between exploration and utilization and quickly find the optimal parameters of the reservoir geomechanical model.

[0112] The system of the present invention also includes a data platform module 5, which is in communication with the microseismic data and model intelligent optimization module 4. The main function of the data platform module 5 is to receive and store the optimized data sent by the module 4, perform data fusion, form atlas data, and synchronize data through the local and cloud parts.

[0113] Preferably, the data platform module 5 includes a data storage module 51, a data fusion module 52 and an algorithm server 53. The data storage module 51 is used to store data transmitted from the microseismic monitoring and transmission module 1, the microseismic image processing and model fusion module 2, and the microseismic intelligent decision-making and auxiliary system module 3 in a relational database format. In practical applications, an open source relational database such as PostgreSQL can be selected, which has powerful geographic information processing capabilities and is particularly suitable for processing spatial data related to oil reservoirs.

[0114] The data fusion module 52 is connected to the data storage module 51 for reading data from the data storage system and fusing the microseismic data, reservoir geomechanics model and graph neural network algorithm to form graph data. This fusion process can use knowledge graph technology to integrate data from different sources and structures into a unified knowledge network. For example, RDF (Resource Description Framework) can be used to represent the relationship between data:

[0115] ```

[0116] <Microseismic event 1><Occurrence time>"2023-01-01T12:00:00".

[0117] <Microseismic Event 1><Location><Location 1>.

[0118] <Location 1><Longitude>"120.5".

[0119] <Location 1><Latitude>"30.5".

[0120] <Location 1><Depth>"3000m".

[0121] <Microseismic event 1><amplitude>"0.5".

[0122] <Microseismic event 1><Associated cracks><Crack 1>.

[0123] <Crack 1><Length>"100m".

[0124] <Crack 1><Direction>"N45E".

[0125] ```

[0126] This representation method can flexibly describe complex data relationships and facilitate subsequent query and analysis.

[0127] The algorithm server 53 is in communication with the data fusion module 52, and is used to perform graph data calculation, combine microseismic data with the reservoir geomechanical model, and transmit the calculated data back to the data storage system. In one embodiment of the present invention, the algorithm server can deploy a graph database (such as Neo4j) and a graph computing engine (such as GraphX) to efficiently process large-scale graph structure data.

[0128] A feature of the present invention is that the data storage module 51 and the data fusion module 52 are both composed of two parts, local and cloud, and the data is kept consistent through synchronization. Among them, local data can be updated in real time, and cloud data is synchronized in real time. This design not only ensures the real-time nature of data processing, but also improves the reliability and scalability of the system. In practical applications, distributed data synchronization technology, such as Apache Kafka, can be used to achieve high-throughput, low-latency data synchronization.

[0129] Through the above design, the system of the present invention can realize the intelligent optimization of microseismic data, the deep fusion of multi-source data, and efficient data storage and calculation. This not only improves the accuracy and efficiency of reservoir fracture monitoring, but also provides a solid data foundation for subsequent intelligent decision-making. The system of the present invention also includes an innovative knowledge graph module 6, which is communicated with the microseismic intelligent decision-making and auxiliary system module 3. The main function of the knowledge graph module 6 is to construct a reservoir management knowledge database based on the distribution of historical experience development wells, and combine the seismic information obtained by the microseismic monitoring and transmission module 1, and use the knowledge graph technology to generate the best development plan recommendation, thereby assisting the microseismic intelligent decision-making and auxiliary system module 3 in making decisions.

[0130] In a preferred embodiment of the present invention, the knowledge graph module 6 includes a historical data extraction unit 61, a geomechanical parameter analysis unit 62, a high permeability zone judgment unit 63, and a development plan recommendation unit 64. The coordinated work of these units enables the system to make full use of historical experience and improve the scientificity and reliability of decision-making.

[0131] The historical data extraction unit 61 is mainly responsible for extracting the distribution of development wells from historical data. Preferably, this unit can use natural language processing (NLP) technology to extract key information from unstructured historical reports. For example, the named entity recognition (NER) technology can be used to identify the well name, location, production and other information in the text. Specifically, a BERT-based NER model can be used, and its pre-training objective function can be expressed as:

[0132] L=L MLM +L NSP ,

[0133] Among them, L MLM is the masked language model loss, L NSP Predict the loss for the next sentence. In this way, the model is able to learn deep contextual representations to more accurately extract entity information related to reservoir development.

[0134] The geomechanical parameter analysis unit 62 is in communication with the historical data extraction unit 61, and is used to select corresponding geomechanical parameters from the historical data to analyze the rationality of the historical development well location. In one embodiment of the present invention, multivariate statistical analysis methods, such as principal component analysis (PCA) and factor analysis, can be used to reduce parameter dimensions and identify key influencing factors. The core of PCA is to solve the eigenvalue problem of the covariance matrix:

[0135] (C-λI)v=0,

[0136] Among them, C is the covariance matrix, λ is the eigenvalue, and v is the eigenvector. By selecting the eigenvectors corresponding to the largest eigenvalues, the data dimension can be greatly reduced while retaining most of the information. The high permeability area judgment unit 63 is communicated with the geomechanical parameter analysis unit 62, and is mainly used to calculate the rock mechanics parameter index, and judge whether the corresponding position is a high permeability area according to the set threshold. In practical applications, fuzzy logic or support vector machine (SVM) and other methods can be used for judgment. Taking SVM as an example, its decision function can be expressed as:

[0137]

[0138] Among them, K(x i ,x) is the kernel function, α i is the Lagrange multiplier, y i is the label, and b is the bias term. By selecting an appropriate kernel function (such as radial basis function, RBF), SVM can effectively handle high-dimensional nonlinear problems and is suitable for complex geological condition judgment.

[0139] The development plan recommendation unit 64 is in communication with the high permeability area judgment unit 63, and its function is to generate the best development plan recommendation based on the high permeability area judgment result. In a preferred embodiment of the present invention, a reinforcement learning method, such as a deep Q network (DQN), can be used to optimize the development plan. The core of DQN is to approximate the Q function through a neural network:

[0140] Q(sa;θ)≈Q * (s,a),

[0141] Where s is the state, a is the action, and θ is the network parameter. By minimizing the temporal difference error:

[0142]

[0143] The development plan can be gradually optimized. Here, r is the immediate reward, γ is the discount factor, and θ - is the target network parameter.

[0144] The present invention also provides an intelligent monitoring method for reservoir fractures based on microseismic images, which corresponds closely to the above system and includes the following steps:

[0145] S1. The microseismic sensor collects and transmits microseismic waves, and uploads the microseismic wave data to the data platform for storage;

[0146] S2. Preprocess the microseismic wave data on the data platform and construct a graph neural network for microseismic images to predict crack extension;

[0147] S3. Input microseismic data, fracture images, and reservoir geomechanical models into deep reinforcement learning to analyze the reservoir geomechanical evolution process, predict fracture expansion, and adjust the mining plan.

[0148] Wherein, step S1 specifically includes:

[0149] S1.1. Arrange a fiber optic sensor network and deploy a microseismic detection array underground;

[0150] S1.2. Collect raw microseismic wave data and ground microseismic wave data;

[0151] S1.3. Upload the collected microseismic wave data to the data platform for storage.

[0152] Preferably, in step S1.1, the density of the fiber optic sensor network can be adjusted according to the depth of the reservoir and the expected fracture distribution. For example, for a reservoir with a depth of less than 3,000 meters, a sensor point can be arranged every 50 meters; while for a shallow reservoir, the density can be increased to one sensor point every 20 meters to obtain a higher spatial resolution.

[0153] Step S2 specifically includes:

[0154] S2.1. On the data platform, the uploaded raw microseismic wave data is denoised and microseismic positioning is performed;

[0155] S2.2. Perform three-dimensional imaging on the microseismic wave data to obtain a microseismic image;

[0156] S2.3. Perform crack identification on microseismic images, establish a graph neural network for microseismic images, and predict crack extension.

[0157] In one embodiment of the present invention, the denoising process of step S2.1 may adopt a wavelet transform method. Specifically, a discrete wavelet transform (DWT) may be used to decompose the signal into different frequency bands, and then a threshold denoising is applied to each frequency band. The threshold may be determined by the following formula:

[0158]

[0159] Where σ is the noise standard deviation and N is the signal length.

[0160] Step S3 specifically includes:

[0161] S3.1. Input microseismic data, fracture images and reservoir geomechanical models into the reinforcement learning algorithm module for calculation;

[0162] S3.2. Use transfer learning algorithms to calibrate microseismic data and optimize reservoir geomechanical models;

[0163] S3.3. Use the optimized microseismic data and reservoir geomechanical model to construct a graph neural network to predict the fracture propagation path;

[0164] S3.4. Predict the trend of fracture expansion and adjust the mining plan based on changes in microseismic data and reservoir geomechanical models.

[0165] In step S3.3, the prediction of the crack extension path can be done using a graph attention network (GAT). The core of GAT is to adaptively assign weights to neighbor nodes through an attention mechanism:

[0166]

[0167] Among them, h i and h j is the node feature, W is the weight matrix, a is the attention vector, and || represents the concatenation operation. This method can better capture the local structural features in the crack network and improve the accuracy of prediction.

[0168] Through the above method, the present invention can realize comprehensive intelligent monitoring of reservoir fractures and provide strong support for oil and gas development decision-making. This method not only improves the accuracy and efficiency of fracture monitoring, but also can dynamically adjust the mining plan according to real-time data to maximize the mining benefits of oil and gas resources.

[0169] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent monitoring system for oil reservoir fractures based on microseismic images, characterized in that: include: Microseismic monitoring and transmission module, used for: Collect microseismic wave data through underground microseismic sensor networks and ground microseismic sensor networks; Transmit and integrate the collected microseismic wave data; The microseismic image processing and model fusion module is connected to the microseismic monitoring and transmission module for: Receiving microseismic wave data sent by the microseismic monitoring and transmission module; Based on the microseismic wave data, a fracture extension prediction model is constructed using a graph neural network; The microseismic intelligent decision-making and auxiliary system module is connected to the microseismic image processing and model fusion module for: Receiving the fracture extension prediction model data sent by the microseismic image processing and model fusion module; Based on the fracture propagation prediction model data and in combination with the reservoir geomechanics model, a deep reinforcement learning algorithm is used to perform microseismic data preprocessing, reservoir geomechanics evolution analysis, and fracture propagation prediction; The microseismic data and model intelligent optimization module is connected to the microseismic intelligent decision-making and auxiliary system module for: Receiving microseismic data and reservoir geomechanical model data sent by the microseismic intelligent decision-making and auxiliary system module; Correction of microseismic data using transfer learning methods; Use optimization functions to optimize reservoir geomechanical models; The data platform module is connected to the microseismic data and model intelligent optimization module for: Receiving and storing the microseismic data and optimized data sent by the model intelligent optimization module; Perform data fusion to form atlas data; Data is synchronized locally and in the cloud.

2. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: The microseismic monitoring and transmission module includes: Downhole microseismic sensor network, used to deploy optical fiber sensor network underground to collect microseismic waves; A ground microseismic sensor network is used to deploy microseismic sensors on the ground to transmit microseismic waves; The gateway is connected to the underground microseismic sensor network and the ground microseismic sensor network for integrating the underground and ground microseismic data and transmitting the integrated data to the microseismic image processing and model fusion module.

3. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: The microseismic image processing and model fusion module includes: A three-dimensional microseismic positioning module is used to locate the microseismic points in the well based on the microseismic waves detected by the downhole optical fiber sensor network and the ground microseismic sensors; A microseismic imaging module, which is in communication connection with the three-dimensional microseismic positioning module, and is used to perform microseismic imaging through multiple three-dimensional microseismic positioning and to perform crack identification on the microseismic data map; The fracture extension prediction module based on graph neural network is communicatively connected with the microseismic imaging module and is used to construct a graph neural network through microseismic images to predict fracture extension.

4. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: The microseismic intelligent decision-making and auxiliary system module includes: Microseismic data intelligent processing module, used for denoising, preprocessing and standardization of microseismic data; A reservoir geomechanical evolution analysis module, which is in communication connection with the microseismic data intelligent processing module and is used to input microseismic data, fracture images and reservoir geomechanical models into a deep reinforcement learning algorithm to analyze the reservoir geomechanical evolution process; The fracture extension prediction module is communicatively connected with the reservoir geomechanical evolution analysis module and is used for predicting fracture extension based on graph neural network.

5. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: The microseismic data and model intelligent optimization module includes: Microseismic data correction module, used to correct microseismic data through transfer learning method; The reservoir geomechanics model optimization module is in communication with the microseismic data correction module and is used for performing model optimization using an optimization function.

6. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: The data platform module includes: Data storage module, used to store data from microseismic monitoring and transmission, microseismic image processing and model fusion, microseismic intelligent decision-making and auxiliary systems in a relational database format; A data fusion module, which is in communication with the data storage module and is used to read data from the data storage system and fuse the microseismic data, the reservoir geomechanics model and the graph neural network algorithm to form graph data; The algorithm server is connected to the data fusion module for performing atlas data calculation, combining the microseismic data with the reservoir geomechanical model, and transmitting the calculated data back to the data storage system.

7. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 6 is characterized in that: The data storage module and the data fusion module are both composed of two parts, local and cloud, and the data are kept consistent through synchronization; wherein the local data can be updated in real time, and the cloud data is synchronized in real time.

8. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 1 is characterized in that: It also includes a knowledge graph module, which is in communication connection with the microseismic intelligent decision-making and auxiliary system module and is used for: Construct a reservoir management knowledge database based on the distribution of historical development wells; Combined with the earthquake information obtained by the microseismic monitoring and transmission module, the knowledge graph technology is used to generate the best development plan recommendation; Assist microseismic intelligent decision-making and assist system modules in making decisions.

9. The intelligent monitoring system for oil reservoir fractures based on microseismic images according to claim 8 is characterized in that: The knowledge graph module includes: A historical data extraction unit, used for extracting the distribution of development wells from historical data; A geomechanical parameter analysis unit, which is in communication with the historical data extraction unit and is used to select corresponding geomechanical parameters from the historical data to analyze the rationality of the historical development well location; A high permeability zone judgment unit is connected to the geomechanical parameter analysis unit for calculating a rock mechanics parameter index. If the index is greater than a set value, the corresponding position is judged to be a high permeability zone. The development plan recommendation unit is communicatively connected to the high permeability area judgment unit, and is used to generate an optimal development plan recommendation based on the high permeability area judgment result.

10. An intelligent reservoir fracture monitoring method based on microseismic images based on the system according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1. The microseismic sensor collects and transmits microseismic waves, and uploads the microseismic wave data to the data platform for storage; S2. Preprocess the microseismic wave data on the data platform and construct a graph neural network for microseismic images to predict crack extension; S3. Input microseismic data, fracture images, and reservoir geomechanical models into deep reinforcement learning to analyze the reservoir geomechanical evolution process and predict fracture expansion to adjust the mining plan; Wherein, step S1 specifically includes: S1.

1. Arrange a fiber optic sensor network and deploy a microseismic detection array underground; S1.

2. Collect raw microseismic wave data and ground microseismic wave data; S1.

3. Upload the collected microseismic wave data to the data platform for storage; Step S2 specifically includes: S2.

1. On the data platform, the uploaded raw microseismic wave data is denoised and microseismic positioning is performed; S2.

2. Perform three-dimensional imaging on the microseismic wave data to obtain a microseismic image; S2.

3. Identify cracks in microseismic images, establish a graph neural network for microseismic images, and predict crack extension; Step S3 specifically includes: S3.

1. Input microseismic data, fracture images and reservoir geomechanical models into the reinforcement learning algorithm module for calculation; S3.

2. Use transfer learning algorithms to calibrate microseismic data and optimize reservoir geomechanical models; S3.

3. Use the optimized microseismic data and reservoir geomechanical model to construct a graph neural network to predict the fracture propagation path; S3.

4. Predict the trend of fracture expansion and adjust the mining plan based on changes in microseismic data and reservoir geomechanical models.

Images