Method for early warning of gas channeling in gas injection of tight reservoirs

Abstract

This invention discloses a gas channeling early warning method for tight oil reservoirs, belonging to the field of intelligent early warning technology. The method includes acquiring static data, dynamic production data, and geological parameters of the target reservoir; identifying dominant channels and their distribution patterns in the horizontal and vertical directions based on the static data; calculating the gas channeling sensitivity index (GSI) based on the dynamic production data and geological parameters; and triggering a gas channeling early warning when the GSI exceeds a preset threshold. This invention establishes a comprehensive risk awareness foundation by integrating static geological, dynamic production, and engineering parameters. It innovatively adopts a two-level early warning logic combining "static fuzzy recognition evaluation" and "real-time early warning of the dynamic GSI index," achieving a seamless transition from potential risk screening to real-time anomaly capture. This enables precise location of high-risk layers at the fracturing cluster level, prediction of the probability of gas channeling in the next 7-15 days, and visual diagnosis of channeling directions, providing direct and quantitative decision-making basis for plugging agent selection and well location optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent early warning technology, and in particular to a method for early warning of gas channeling during gas injection in tight oil reservoirs. Background Technology

[0002] Tight oil reservoirs are an important component of global unconventional oil and gas resources, and their effective development is of great significance for ensuring energy security. Gas injection (including gas injection...) Enhancing oil recovery (ERR) is one of the key technologies for developing tight oil reservoirs. Because tight reservoirs typically require large-scale hydraulic fracturing to create complex fracture networks for economical extraction, these reservoirs often contain both artificially created fractures and natural fractures or high-permeability bands. During gas injection, injected gas is highly susceptible to channeling along these high-permeability channels, a phenomenon known as "gas channeling." Gas channeling causes injected gas to prematurely penetrate into the production well, significantly reducing sweep efficiency and limiting the effectiveness of enhanced oil recovery. It also wastes gas resources and increases recovery costs. Early and accurate warning of gas channeling, along with timely control measures, is one of the core challenges to the successful development of gas injection in tight oil reservoirs. Research and practice on gas channeling monitoring and early warning mainly focus on the following aspects: First, empirical analysis methods based on dynamic production data, such as observing a sharp increase in the gas-oil ratio (GOR) and abnormal changes in bottom hole pressure. This method is simple but has a significant lag; by the time an early warning is issued, gas channeling has often already occurred. There are three main types of early warning methods: First, gas channeling can occur through various means. Second, there are numerical simulation-based methods that predict gas channeling trends by establishing geological models and fluid motion equations. However, this method is limited by model accuracy, large computational scale, and difficulty in achieving real-time early warning. Third, there are data-driven early warning methods that have emerged in recent years. The essential difference between tight oil reservoirs and conventional clastic reservoirs is that their reservoir porosity is usually less than 10% and air permeability is less than 0.1 mD. Large-scale hydraulic fracturing is necessary to form complex fracture networks for economical exploitation. The complex coupling relationship between artificial and natural fractures makes gas channeling channels exhibit multi-scale, discontinuous, and dynamically evolving characteristics. Traditional numerical simulation methods based on continuous medium theory are difficult to describe effectively. In particular, existing numerical simulation-based research methods (such as related research at China University of Petroleum) are limited by continuous medium theory and grid scale, making it difficult to effectively characterize the microscopic channels formed by the coupling of artificial and natural fractures in tight oil reservoirs, and thus unable to achieve accurate early warning at the fracturing cluster level.

[0003] However, current common solutions have many drawbacks, including: existing methods mostly rely on a single data dimension, with insufficient integration of dynamic and static information, making it difficult to comprehensively identify the main controlling factors of gas channeling; they lack a hierarchical quantitative evaluation system from static geological risk assessment to dynamic real-time early warning, resulting in insufficient foresight in early warning; early warning indicators are usually relatively simple, with poor comprehensive response to multi-parameter coupled changes, leading to high false alarm and false alarm rates; traditional numerical simulation has low computational efficiency, and conventional data-driven models are difficult to effectively characterize the spatiotemporal correlation between wells, with limited accuracy in identifying small-scale high-risk layers; early warning results are often significantly delayed, making it difficult to achieve hourly real-time rolling prediction and effective lead time; and the early warning output format is simple, lacking visual diagnosis of channeling direction, risk layer, and probability, making it difficult to provide direct and accurate decision support for on-site control. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the problems existing in the above-mentioned gas channeling early warning methods for gas injection in tight oil reservoirs, this invention is proposed.

[0006] Therefore, the purpose of this invention is to provide a gas channeling early warning method for tight oil reservoirs. This method is applicable to solving the problems of existing methods relying on a single data dimension, insufficient integration of dynamic and static information, difficulty in comprehensively identifying the main controlling factors of gas channeling; lack of a step-by-step quantitative evaluation system from static geological risk assessment to dynamic real-time early warning, resulting in insufficient early warning foresight; early warning indicators are usually relatively simple, with poor comprehensive response to multi-parameter coupled changes, leading to high false alarm and false alarm rates; traditional numerical simulation has low computational efficiency, conventional data-driven models are difficult to effectively characterize the spatiotemporal correlation between wells, and the identification accuracy of small-scale high-risk layers is limited; early warning results are often significantly delayed, making it difficult to achieve hourly real-time rolling prediction and effective lead time.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a gas channeling early warning method for tight oil reservoirs, which includes acquiring static data, dynamic production data, and geological parameters of the target oil reservoir; based on the static data, using a fuzzy recognition method to comprehensively assess the risk of vertical channeling, identifying the dominant channels and their distribution patterns in the plane and vertical directions; based on the dynamic production data and geological parameters, calculating the gas channeling sensitivity index (GSI); when the GSI exceeds a preset threshold, triggering a gas channeling early warning, and diagnosing the channeling direction and high-risk layers based on a constructed AI early warning model.

[0008] As a preferred embodiment of the gas channeling early warning method for tight oil reservoirs described in this invention, the static data includes data on the coupling relationship between artificial fractures and natural large channels; the GSI is a multi-parameter fusion index that comprehensively reflects the risk of gas channeling.

[0009] As a preferred embodiment of the tight oil reservoir gas injection gas channeling early warning method of the present invention, the acquisition of static data, dynamic production data and geological parameters of the target oil reservoir specifically includes the following: Static data: including data reflecting the heterogeneity and connectivity of the reservoir, including at least the planar and vertical permeability distribution, porosity distribution, geometric parameters and conductivity of artificially fractured fractures, development scale and orientation of natural fractures or large channels, and coupling relationship data between artificial fractures and natural channels; Dynamic production data: including real-time or historical monitoring data at the wellhead or bottom, including at least the oil production, gas production, water production, gas-oil ratio, bottomhole flowing pressure, wellhead pressure, and injection pressure and injection volume of injection wells; Geological and engineering parameters: including parameters used to evaluate the compressibility and hydrocarbon content of the reservoir, including at least the total organic carbon content, rock brittleness index, mineral composition, oil saturation, and the direction and magnitude of geostress.

[0010] As a preferred embodiment of the tight oil reservoir gas injection gas channeling early warning method of the present invention, the method comprises: based on the static data, using a fuzzy recognition method to comprehensively evaluate the vertical channeling risk, identifying the dominant channels and their distribution patterns in the plane and vertical direction, specifically as follows: determining a set of static evaluation factors affecting the vertical channeling risk, wherein the factor set includes at least reservoir permeability difference, interlayer distribution frequency and thickness, fracture vertical extension height, and the degree of vertical communication between artificial fractures and natural channels; establishing a membership function characterizing the risk level of each static evaluation factor, and determining the weight of each factor based on expert experience or the analytic hierarchy process to form a weight vector; and assigning the target well or well... The static evaluation factor data of each sub-layer is input into the fuzzy recognition comprehensive evaluation model. Through fuzzy synthesis operation, the comprehensive evaluation value of the vertical cross-connection risk of each sub-layer or segment is calculated. Based on the comprehensive evaluation value, the risk level is divided, and high-risk segments are identified as potential advantageous channels. By comparing and analyzing the evaluation results of multiple wells, the distribution range of the advantageous channels in the plane and their development layers in the vertical direction are clarified, thereby clarifying their plane and vertical distribution patterns. The method system covers three dimensions: static, dynamic, and quantitative. Finally, it outputs a quantitative evaluation result of the water flooding degree or cross-connection risk of a single sub-layer, and classifies the channel type accordingly, providing a basis for subsequent plugging agent selection and well location optimization.

[0011] As a preferred embodiment of the tight reservoir gas channeling early warning method of the present invention, the gas channeling sensitivity index (GSI) is calculated based on the dynamic production data and geological parameters. Specifically, the dynamic production data and geological parameters include at least: dynamic response parameters: real-time or historical gas-oil ratio, bottom hole pressure change rate, and production decline rate of the production well; injection-production correlation parameters: pressure difference between the injection well and the production well, and the time trend of injected gas breakthrough; geologically sensitive parameters: total organic carbon content, brittleness index, and the ratio of permeability to porosity of the target layer; a multi-parameter fusion mathematical model is established based on reservoir engineering theory and the core calculation parameters to quantitatively characterize the gas channeling sensitivity of the current injection-production unit; real-time gas channeling risk early warning is achieved at the reservoir engineering level by inputting the real-time collected dynamic production data and static geological parameters into the GSI calculation model; when the calculated GSI value exceeds a preset threshold, the system automatically triggers a secondary diagnostic process targeting the channeling direction and high-risk layers.

[0012] As a preferred embodiment of the gas channeling early warning method for tight oil reservoirs described in this invention, the gas channeling early warning is triggered when the GSI exceeds a preset threshold, and the channeling direction and high-risk layers are diagnosed based on the constructed AI early warning model. Specifically, when the calculated value of the gas channeling sensitivity index (GSI) continuously exceeds the threshold, the system automatically issues a gas channeling early warning and initiates a secondary diagnostic process; the enhanced dataset is input into the pre-trained AI early warning model; the AI ​​early warning model outputs a visualized diagnostic report including the gas channeling probability, high-risk layers, and channeling direction.

[0013] As a preferred embodiment of the tight reservoir gas injection gas channeling early warning method of the present invention, the AI ​​early warning model is a model based on a fusion architecture of spatiotemporal graph convolutional network (ST-GCN) and deep convolutional generative adversarial network, specifically the FC-DC-GAN (Fully Connected-Deep Convolutional Generative Adversarial Network) model; the model takes the enhanced dataset as input, captures the spatiotemporal dynamic correlation features of inter-well data in the injection-production well network through the spatiotemporal graph convolutional network, and performs rolling predictions at an hourly frequency.

[0014] Secondly, to further address the aforementioned technical problems, the present invention provides a gas channeling early warning system for tight oil reservoirs, comprising: a data acquisition module for acquiring static data, dynamic production data, and geological parameters of the target oil reservoir; a pattern recognition module for comprehensively evaluating vertical channeling risk based on static data using a fuzzy recognition method, identifying dominant channels and their distribution patterns in the plane and vertical directions; an index calculation module for calculating the gas channeling sensitivity index (GSI) based on dynamic production data and geological parameters; and a direction diagnosis module for triggering a gas channeling early warning when the GSI exceeds a preset threshold, and diagnosing the channeling direction and high-risk layers based on a constructed AI early warning model.

[0015] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any step of the tight oil reservoir gas injection gas channeling early warning method as described in the first aspect of the present invention.

[0016] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the tight oil reservoir gas injection gas channeling early warning method as described in the first aspect of the present invention.

[0017] The beneficial effects of this invention are as follows: This invention constructs a comprehensive risk cognition foundation by integrating static geological, dynamic production, and engineering parameters, and innovatively adopts a two-level early warning logic combining "static fuzzy recognition evaluation" and "dynamic GSI index real-time early warning," realizing the connection from potential risk screening to real-time anomaly capture, significantly improving the foresight and reliability of early warning. Compared with existing research methods based on numerical simulation, this invention, for the first time, improves the early warning accuracy from the traditional small-layer level (5-10m) to the fracturing cluster level (0.5-1m), achieving precise positioning of cluster-scale microchannels. At the same time, with the help of an AI model based on spatiotemporal graph convolution and generative adversarial networks, it realizes rolling prediction of the probability of gas channeling in the next 7-15 days and visual diagnosis of channeling direction, providing direct and quantitative decision-making basis for plugging agent selection and well location optimization. Finally, it forms a systematic and closed-loop intelligent early warning method, effectively improving gas drive sweep efficiency and development economic benefits. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of 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. Wherein: Figure 1This is a flowchart illustrating the implementation of the present invention in Example 1.

[0019] Figure 2 This is a dynamic adjustment diagram of the present invention in Example 1. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1 Reference Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a method for early warning of gas channeling during gas injection in tight oil reservoirs, comprising the following steps: S1: Obtain static data, dynamic production data, and geological parameters of the target reservoir.

[0024] Furthermore, the static data includes data on the coupling relationship between artificial fractures and natural large channels.

[0025] Preferably, static data, dynamic production data, and geological parameters of the target reservoir are obtained, as detailed below: Static data includes data reflecting the heterogeneity and connectivity of the reservoir, including at least the planar and vertical permeability distribution, porosity distribution, geometric parameters and conductivity of artificially fractured fractures, development scale and orientation of natural fractures or large channels, and the coupling relationship between artificial fractures and natural channels.

[0026] Dynamic production data includes real-time or historical monitoring data at the wellhead or bottom, including at least the oil production, gas production, water production, gas-oil ratio, bottomhole flowing pressure, wellhead pressure, and injection pressure and injection volume of injection wells.

[0027] Geological and engineering parameters: These include parameters used to evaluate reservoir compressibility and hydrocarbon potential, including at least total organic carbon content, rock brittleness index, mineral composition, oil saturation, and the direction and magnitude of geostress.

[0028] Preferably, this step systematically integrates static geological data (including the coupling relationship between artificial fractures and natural large pores), dynamic production data (such as pressure, production, and gas-oil ratio), and key geological engineering parameters (such as total organic carbon content, brittleness index, and geostress) to construct a multi-dimensional, multi-scale comprehensive reservoir information foundation. This overcomes the limitations of traditional methods that rely on single or fragmented data sources, and achieves a unified characterization of the inherent heterogeneity of reservoirs, engineering modification characteristics, and dynamic production responses. It provides comprehensive and reliable data support for subsequent risk identification and early warning models, laying the foundation for the overall method's advanced and complete nature at the data fusion level.

[0029] For example, taking a tight oil block as an example, firstly, the planar permeability distribution map and porosity model of the target layer are extracted from the geological database. Combined with microseismic monitoring data and fracturing construction reports, the length, height, conductivity of artificial fractures and the occurrence and density of natural fractures are obtained. Then, the spatial coupling relationship between artificial fractures and natural fracture systems is quantified through three-dimensional modeling software. At the same time, the oil production, gas production, bottom hole flowing pressure, and injection pressure of each well are collected in real time from the SCADA system. The total organic carbon content obtained from laboratory core analysis, X-ray diffraction mineral composition, brittleness index obtained from rock mechanics experiments, and geostress field data are integrated to form a comprehensive dataset that combines multiple sources, multiple scales, and dynamic and static data.

[0030] S2: Based on static data, a fuzzy recognition method is used to comprehensively assess the risk of vertical collusion, and to identify the dominant channels and their distribution patterns in the plane and vertical directions.

[0031] Preferably, based on static data, a fuzzy recognition method is used to comprehensively assess the risk of vertical collusion, identify the dominant channels and their distribution patterns in the plane and vertical directions, as detailed below: Determine the set of static evaluation factors that affect the risk of vertical cross-contamination. The set of factors should include at least the reservoir permeability gradient, interlayer distribution frequency and thickness, fracture vertical extension height, and the degree of vertical communication between artificial fractures and natural channels.

[0032] For each static evaluation factor, a membership function representing its risk level is established, and the weights of each factor are determined based on expert experience or the analytic hierarchy process, forming a weight vector.

[0033] The static evaluation factor data of each sub-layer of the target well or well group are input into the fuzzy recognition comprehensive evaluation model. Through fuzzy synthesis operation, the comprehensive evaluation value of the vertical cross-connection risk of each sub-layer or segment is calculated.

[0034] Based on the comprehensive evaluation value, risk levels are classified, and high-risk sections are identified as potential advantageous channels. By comparing and analyzing the evaluation results of multiple wells, the distribution range of advantageous channels in the plane and their development strata in the vertical direction are clarified, thereby clarifying their plane and vertical distribution patterns.

[0035] The methodology covers three dimensions: static, dynamic, and quantitative. It ultimately outputs quantitative evaluation results on the degree of water flooding or the risk of cross-contamination in a single sublayer, and classifies the channel types accordingly, providing a basis for subsequent plugging agent selection and well location optimization.

[0036] Preferably, this step is based on fuzzy mathematics theory, which quantitatively and comprehensively evaluates static evaluation factors such as permeability difference, interlayer distribution, and vertical connectivity of fractures through membership functions and weight vectors. It outputs a comprehensive evaluation value of vertical channeling risk and a distribution map of the dominant channel in the plane and vertical direction, realizing early and proactive prediction of gas channeling risk. This transforms risk identification from a lagging model that relies on dynamic production anomalies to a priori model based on geological static attributes. It provides quantitative and visualized decision-making basis for development scheme optimization, risk classification management, and subsequent dynamic early warning, and serves as a priori guide for high-risk areas, improving the foresight and systematic nature of the overall early warning process.

[0037] For example, for the aforementioned block, permeability gradient, interlayer frequency, fracture longitudinal extension height, and artificial-natural fracture communication coefficient are selected as static evaluation factors. A trapezoidal membership function is used to fuzzify each factor, and weights are assigned using the analytic hierarchy process (AHP) (e.g., the communication coefficient has the highest weight). Data from each sub-layer is input into the model for fuzzy synthesis, outputting a comprehensive evaluation value for the risk of cross-contamination in each sub-layer. Based on the evaluation value, the layers are divided into risk levels I-III, identifying high-risk layers mainly distributed in the middle of a key layer, exhibiting a strip-like distribution on the plane. This clarifies the distribution pattern of dominant channels, providing a quantitative basis for selecting plugging agents in this area.

[0038] S3: Calculate the gas channeling sensitivity index (GSI) based on dynamic production data and geological parameters.

[0039] Furthermore, GSI is a multi-parameter fusion index that comprehensively reflects the risk of gas channeling.

[0040] Preferably, the gas channeling sensitivity index (GSI) is calculated based on dynamic production data and geological parameters, as detailed below: Dynamic production data and geological parameters should include at least: Dynamic response parameters: real-time or historical gas-oil ratio of production wells, bottom hole pressure change rate, and production decline rate.

[0041] Injection-production related parameters: pressure difference between injection wells and production wells, and the trend of injected gas breakthrough time.

[0042] Geologically sensitive parameters: total organic carbon content, brittleness index, and permeability to porosity ratio of the target layer.

[0043] Based on reservoir engineering theory and core calculation parameters, a multi-parameter fusion mathematical model is established to quantitatively characterize the gas channeling sensitivity of the current injection and production unit.

[0044] By inputting real-time collected dynamic production data and static geological parameters into the GSI calculation model, real-time gas channeling risk early warning at the reservoir engineering level can be achieved.

[0045] When the calculated GSI value exceeds the preset threshold, the system automatically triggers a secondary diagnostic process for the direction of collusion and high-risk layers.

[0046] It should be noted that the GSI, as a stability criterion of the injection-production system, is used to quantify the degree to which the current injection-production unit deviates from the stable displacement state.

[0047] Specifically, the GSI calculation model adopts a progressive threshold triggering mechanism: when GSI∈(0.3,0.5], the system enters a blue alert state and increases the monitoring frequency; when GSI∈(0.5,0.7], the system enters an orange alert state and automatically starts the AI ​​model for pre-diagnosis; when GSI>0.7, the system enters a red alert state, starts the secondary diagnostic process at full speed and pushes a real-time diagnostic report.

[0048] Preferably, this step innovatively constructs a gas channeling sensitivity index (GSI), which is a multi-parameter mathematical model that integrates dynamic response parameters (such as gas-oil ratio and pressure change rate), injection-production correlation parameters (such as injection-production pressure difference and breakthrough trend), and geologically sensitive parameters (such as brittleness index and porosity ratio). It condenses high-dimensional, high-frequency real-time data streams into a scalar index with clear reservoir engineering significance, realizing real-time and continuous quantitative monitoring of the gas channeling sensitivity of injection-production units. Through an automatic triggering mechanism with a preset threshold (such as 0.5), an efficient data dimensionality reduction and real-time early warning sentinel is established, balancing computational efficiency and early warning sensitivity. It provides accurate and timely triggering signals for initiating advanced intelligent diagnosis, enhancing the automation and reliability of early warning.

[0049] For example, based on the real-time production data of the block, the current gas-oil ratio, bottom hole pressure change rate, and production decline rate are calculated as dynamic response parameters. The pressure difference between injection and production wells and the gas breakthrough time trend are combined as injection-production correlation parameters, and the brittleness index and porosity ratio of the formation are introduced as geologically sensitive parameters. After normalizing these parameters, they are input into the GSI calculation model constructed based on multiple regression, and the current GSI value is 0.63. When the system detects that the value continues to exceed the preset threshold of 0.5, it automatically triggers the secondary diagnostic process, issues a gas channeling warning to the operator, and starts the AI ​​diagnostic module.

[0050] S4: When GSI exceeds the preset threshold, trigger gas channeling warning and diagnose the channeling direction and high-risk layers based on the constructed AI warning model.

[0051] Specifically, the AI ​​early warning model is a deep convolutional generative adversarial network (FC-DC-GAN) model based on a spatiotemporal graph convolutional network, which is used to capture dynamic correlations between wells and output gas channeling probability and high-risk layer information.

[0052] Preferably, when the GSI exceeds a preset threshold, a gas channeling warning is triggered, and the direction of channeling and high-risk layers are diagnosed based on the constructed AI warning model, as detailed below: When the calculated value of the Gas Channeling Sensitivity Index (GSI) continuously exceeds the threshold, the system automatically issues a gas channeling warning and initiates a secondary diagnostic process; it integrates an enhanced dataset for AI model analysis, which includes, but is not limited to: real-time and historical wellbore production data, inter-well interference test data, dynamic and geological parameters on which the GSI calculation depends, and the plane and vertical distribution information of the dominant channels obtained in step S2.

[0053] The augmented dataset is input into a pre-trained AI early warning model; wherein the AI ​​early warning model is a hybrid architecture model that integrates a spatiotemporal graph convolutional network and a deep convolutional generative adversarial network, and is configured to: Capture the spatiotemporal dynamic correlation characteristics of data between injection and production wells in the injection-production well network.

[0054] Based on features, the probability of gas leakage in the target production well within the next 7 to 15 days is predicted on an hourly basis.

[0055] It identifies and outputs the high-risk strata where gas channeling is most likely to occur, with an identification accuracy reaching the fracturing cluster level.

[0056] Infer and visualize potential cross-flow directions, i.e., the dominant flow path of injected gas from the injection well to the production well.

[0057] The AI ​​early warning model outputs a visualized diagnostic report that includes the probability of gas channeling, high-risk layers, and channeling direction. This report guides the implementation of targeted anti-channeling and control measures on-site and provides real-time, quantitative decision-making basis for subsequent plugging agent selection and well location optimization.

[0058] Specifically, the AI ​​early warning model is a model based on a fusion architecture of spatiotemporal graph convolutional network (ST-GCN) and deep convolutional generative adversarial network, specifically the FC-DC-GAN (Fully Connected-Deep Convolutional Generative Adversarial Network) model.

[0059] The model takes an augmented dataset as input, captures the spatiotemporal dynamic correlation features of data between injection and production wells through a spatiotemporal graph convolutional network, and performs rolling predictions at an hourly frequency.

[0060] The diagnostic results output by the model include: the probability of gas channeling in the target production well within the next 7 to 15 days, the high-risk layers where gas channeling will occur (identification accuracy reaches the fracturing cluster level), and the potential channeling direction between the injection well and the production well.

[0061] It should be noted that the specific implementation of the FC-DC-GAN model is as follows: Input Feature Construction: Enhance each time step in the dataset by constructing the following feature vector: Node characteristics: real-time gas production, gas-oil ratio, bottom hole flowing pressure, gas injection volume of each well, and the prior probability value of the hidden dominant channel at the location of the well obtained in step S2.

[0062] Graph structure definition: Well locations are used as nodes, and the reciprocal of the Euclidean distance between wells is used as the initial edge weights. During training, the edge weights are adaptively updated based on dynamic connectivity indicators (such as injection-production response time delay).

[0063] ST-GCN module structure: Employs 3 layers of spatiotemporal graph convolutional blocks, each layer containing: Spatial graph convolution: Using Chebyshev polynomial approximation graph Laplace to capture spatial relationships between wells.

[0064] Temporally gated convolution: Performs causal convolution along the time dimension to capture the temporal evolution of production parameters.

[0065] Generative Adversarial Training: Generator: Based on encoder-decoder structure, it takes 7 days of historical data as input and outputs the spatiotemporal distribution of air current propagation probability for the next 7-15 days.

[0066] Discriminator: A deep convolutional network is used to determine whether the generated future gas migration probability distribution is consistent with the real historical pattern.

[0067] Loss function: Combining L1 loss and adversarial loss to enhance the diversity of generated samples while ensuring prediction accuracy.

[0068] Implementation of fracturing cluster-level positioning: The output layer of the model is designed as a three-dimensional tensor: number of wells × number of fracturing clusters in the well × time step. For each cluster, the probability value of gas channeling in the next 7-15 days is output. After Softmax normalization, clusters with a probability greater than 0.6 are taken as high-risk layers for output.

[0069] It should be noted that, Figure 1The "AI early warning model" in step S4 adopts the FC-DC-GAN architecture. This model integrates a spatiotemporal graph convolutional network (ST-GCN) to capture the spatiotemporal dynamic correlation characteristics of data between injection and production wells.

[0070] It should be noted that the innovation of the FC-DC-GAN model lies in its dual-channel feature extraction architecture: Spatial Channel: The injection-production well network is constructed as a dynamic graph structure through ST-GCN. Node features include well location coordinates, fracturing cluster locations, and inter-well distances. Edge weights are adaptively updated based on dynamic connectivity.

[0071] Time channel: The attention mechanism is used to weight and fuse historical production data (7-15 days) to capture the evolution of precursor features of gas surge.

[0072] Fusion Diagnosis: Through game-theoretic training of generative adversarial networks, multi-scale gas channeling probability prediction is achieved from macro-well groups to micro-fracture clusters.

[0073] Preferably, this step utilizes an FC-DC-GAN fusion model (integrated spatiotemporal graph convolutional network ST-GCN) specifically designed for the spatiotemporal dynamics of injection-production well networks to perform rolling predictions and visual diagnoses of cross-flow direction, high-risk layers (fracturing cluster level), and cross-flow probability in the next 7-15 days. This achieves a leap from macro-level risk warning to micro-level precise positioning. By deeply mining the spatiotemporal correlation characteristics of inter-well data, it provides probabilistic, cluster-level accurate, and directional visualized diagnostic reports, directly supporting precise control decisions such as plugging agent selection, drive optimization, and well location adjustment. It transforms warning information into executable engineering instructions, forming a closed-loop management system from risk perception to intervention implementation, significantly improving the pertinence of measures and development economic benefits.

[0074] For example, after the early warning is triggered, the system integrates real-time production data, inter-well interference test data, current GSI-related parameters, and the distribution map of the dominant channel generated in step S2 to form an enhanced dataset; inputs a pre-trained FC-DC-GAN model, which learns the dynamic correlation between wells through a spatiotemporal graph convolutional network, and predicts that the probability of gas channeling in the target production well in the next 10 days is 82%, identifies the high-risk layer as being near the 5th fracturing cluster, and visualizes the channeling direction as from the gas injection well A along the northeast-oriented dominant channel to the production well B; the diagnostic report is immediately pushed to the field command system, supporting decision-makers to deploy plugging agents for the 5th cluster and adjust the injection and production regime.

[0075] In summary, this invention establishes a comprehensive risk awareness foundation by integrating static geological, dynamic production, and engineering parameters. It innovatively employs a two-tiered early warning logic combining "static fuzzy recognition evaluation" and "dynamic GSI index real-time early warning," achieving a seamless transition from potential risk screening to real-time anomaly capture. This significantly enhances the foresight and reliability of early warnings. Compared to existing numerical simulation-based research methods, this invention, for the first time, elevates early warning accuracy from the traditional small-layer level (5-10m) to the fracturing cluster level (0.5-1m), enabling precise positioning of cluster-scale microchannels. Furthermore, leveraging an AI model based on spatiotemporal graph convolution and generative adversarial networks, it achieves rolling predictions of gas channeling probability over the next 7-15 days and visualized diagnosis of channeling direction. This provides direct and quantitative decision-making support for plugging agent selection and well location optimization. Ultimately, this forms a systematic and closed-loop intelligent early warning method, effectively improving gas-driven sweep efficiency and development economic benefits.

[0076] Example 2, an embodiment of the present invention, provides a gas channeling early warning system for tight oil reservoirs, comprising: a data acquisition module for acquiring static data, dynamic production data, and geological parameters of the target oil reservoir; a pattern identification module for comprehensively evaluating the vertical channeling risk based on static data using a fuzzy recognition method, identifying the dominant channels and their distribution patterns in the plane and vertical directions; an index calculation module for calculating the gas channeling sensitivity index (GSI) based on dynamic production data and geological parameters; and a direction diagnosis module for triggering a gas channeling early warning when the GSI exceeds a preset threshold, and diagnosing the channeling direction and high-risk layers based on a constructed AI early warning model.

[0077] Example 3 is an embodiment of the present invention, which differs from the previous embodiment in that: If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0079] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0080] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0081] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for early warning of gas channeling during gas injection in tight oil reservoirs, characterized in that: include: Acquire static data, dynamic production data, and geological parameters of the target reservoir; Based on the static data, a fuzzy recognition method is used to comprehensively assess the risk of vertical collusion, and to identify the dominant channels and their distribution patterns in the plane and vertical directions. Based on the aforementioned dynamic production data and geological parameters, the gas channeling sensitivity index (GSI) is calculated. When the GSI exceeds a preset threshold, a gas channeling warning is triggered, and the direction of channeling and high-risk layers are diagnosed based on the constructed AI warning model.

2. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 1, characterized in that: The static data includes data on the coupling relationship between artificial cracks and natural large channels; The GSI is a multi-parameter fusion index that comprehensively reflects the risk of gas channeling.

3. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 1, characterized in that: The specific details of acquiring the static data, dynamic production data, and geological parameters of the target reservoir are as follows: Static data includes data reflecting the heterogeneity and connectivity of the reservoir, including at least the planar and vertical permeability distribution, porosity distribution, geometric parameters and conductivity of artificially fractured fractures, development scale and orientation of natural fractures or large channels, and the coupling relationship between artificial fractures and natural channels. Dynamic production data: including real-time or historical monitoring data at the wellhead or bottom, including at least the oil production, gas production, water production, gas-oil ratio, bottomhole flowing pressure, wellhead pressure, and injection pressure and injection volume of the gas injection well. Geological and engineering parameters: These include parameters used to evaluate reservoir compressibility and hydrocarbon potential, including at least total organic carbon content, rock brittleness index, mineral composition, oil saturation, and the direction and magnitude of geostress.

4. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 1, characterized in that: Based on the aforementioned static data, a fuzzy recognition method is used to comprehensively assess the risk of vertical collusion, identifying the dominant channels and their distribution patterns in both the horizontal and vertical directions. The specific details are as follows: Determine a set of static evaluation factors that affect the risk of vertical cross-connection. The set of factors includes at least reservoir permeability gradient, interlayer distribution frequency and thickness, fracture vertical extension height, and the degree of vertical communication between artificial fractures and natural channels. For each of the static evaluation factors, a membership function representing its risk level is established, and the weight of each factor is determined based on expert experience or the analytic hierarchy process to form a weight vector. The static evaluation factor data of each sub-layer of the target well or well group are input into the fuzzy recognition comprehensive evaluation model. Through fuzzy synthesis operation, the comprehensive evaluation value of the vertical cross-connection risk of each sub-layer or segment is calculated. Based on the comprehensive evaluation value, risk levels are classified, and high-risk sections are identified as potential advantageous channels. By comparing and analyzing the evaluation results of multiple wells, the distribution range of the advantageous channels in the plane and their development strata in the vertical direction are clarified, thereby clarifying their plane and vertical distribution patterns. The proposed methodology covers three dimensions: static, dynamic, and quantitative. It ultimately outputs a quantitative evaluation result of the degree of flooding or cross-contamination risk in a single sublayer, and classifies the channel type accordingly, providing a basis for subsequent plugging agent selection and well location optimization.

5. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 1, characterized in that: Based on the aforementioned dynamic production data and geological parameters, the Gas Channelization Sensitivity Index (GSI) is calculated, as detailed below: The dynamic production data and geological parameters include at least: Dynamic response parameters: real-time or historical gas-oil ratio of production wells, bottom hole pressure change rate, and production decline rate; Injection-production related parameters: pressure difference between injection wells and production wells, and the trend of injected gas breakthrough time; Geologically sensitive parameters: total organic carbon content, brittleness index, and permeability to porosity ratio of the target layer; Based on reservoir engineering theory and the core calculation parameters, a multi-parameter fusion mathematical model is established to quantitatively characterize the gas channeling sensitivity of the current injection and production unit. By inputting the real-time collected dynamic production data and static geological parameters into the GSI calculation model, real-time gas channeling risk early warning at the reservoir engineering level can be achieved. When the calculated GSI value exceeds the preset threshold, the system automatically triggers a secondary diagnostic process for the direction of collusion and high-risk layers.

6. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 1, characterized in that: When the GSI exceeds a preset threshold, a gas channeling warning is triggered, and the direction of channeling and high-risk layers are diagnosed based on the constructed AI warning model, as detailed below: When the calculated value of the gas channeling sensitivity index (GSI) continues to exceed the threshold, the system automatically issues a gas channeling warning and initiates a secondary diagnostic process. The augmented dataset is then input into a pre-trained AI early warning model; The AI ​​early warning model outputs a visualized diagnostic report that includes the probability of gas channeling, high-risk layers, and the direction of channeling.

7. The method for early warning of gas channeling during gas injection in tight oil reservoirs as described in claim 6, characterized in that: The AI ​​early warning model is a model based on a fusion architecture of spatiotemporal graph convolutional network (ST-GCN) and deep convolutional generative adversarial network, specifically the FC-DC-GAN (Fully Connected-Deep Convolutional Generative Adversarial Network) model. The model takes the augmented dataset as input, captures the spatiotemporal dynamic correlation features of inter-well data in the injection-production well network through the spatiotemporal graph convolutional network, and performs rolling predictions at an hourly frequency.

8. A gas channeling early warning system for tight oil reservoirs, based on the gas channeling early warning method for tight oil reservoirs according to any one of claims 1 to 7, characterized in that: include, The data acquisition module is used to acquire static data, dynamic production data, and geological parameters of the target reservoir. The pattern recognition module is used to comprehensively assess the risk of vertical collusion based on static data and using fuzzy recognition methods, and to identify the dominant channels and their distribution patterns in the plane and vertical directions. The index calculation module is used to calculate the gas channeling sensitivity index (GSI) based on dynamic production data and geological parameters. The directional diagnosis module is used to trigger a gas channeling warning when the GSI exceeds a preset threshold, and to diagnose the channeling direction and high-risk layers based on the constructed AI warning model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the tight oil reservoir gas injection gas channeling early warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the gas channeling early warning method for tight oil reservoirs according to any one of claims 1 to 7.

Images