Fractal network optimization-based injection well water injection volume improvement method

Abstract

The application discloses a fractal network optimization-based injection well water injection volume improvement method, comprising the following steps: S1, reservoir geological data acquisition and multi-source database construction; S2, adaptive fractal seepage network construction and optimization; S3, injection water pressure and flow coordinated optimization based on the optimized network; S4, dynamic adjustment of injection well water injection strategy; S5, reservoir seepage key feature extraction and regulation modeling; S6, remote water injection control strategy based on risk assessment optimization; S7, intelligent optimization of water injection parameters and remote monitoring. Through adaptive fractal seepage network optimization, combined with intelligent optimization algorithm and remote monitoring control, the application realizes accurate regulation and control of injection well water injection volume, improves reservoir seepage efficiency, optimizes water injection strategy, reduces energy consumption, and improves oilfield recovery.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method for increasing water injection volume in reinjection wells based on fractal network optimization. Background Technology

[0002] In oilfield development, water injection in reinjection wells is a crucial means of maintaining formation pressure and enhancing oil recovery. However, as oilfield development deepens, factors such as formation heterogeneity, complex seepage paths, and changes in reservoir properties make precise control difficult with traditional water injection methods. Conventional water injection control methods in reinjection wells are typically based on static geological modeling or empirical formulas. These methods rely on historical data for parameter adjustments and are ill-suited to adapting to dynamic reservoir changes, leading to uneven water injection, increased energy consumption, and reservoir blockage. Therefore, establishing efficient water injection optimization strategies to achieve intelligent, dynamic, and precise control in complex reservoir environments has become a core challenge in oilfield water injection management.

[0003] Currently, traditional water injection strategies mainly rely on single seepage models or limited numerical simulation methods. For example, the commonly used Darcy flow theory is used to simulate fluid flow in porous media, but this theory is only applicable to homogeneous media. It struggles to accurately characterize fluid transport in the heterogeneous, multi-scale seepage environments prevalent in actual reservoirs. Furthermore, traditional methods are typically based on single-objective optimization, such as minimizing pressure gradients or maximizing flow rate matching, but fail to consider dynamic reservoir response, water injection energy consumption, and long-term reservoir stability. This lack of systematic optimization often leads to water injection volumes that are difficult to adapt to reservoir changes, thus impacting the economic benefits of reservoir development.

[0004] With the rapid development of computer simulation technology and artificial intelligence, some research has begun to explore using optimization algorithms to improve water injection strategies. For example, genetic algorithms and particle swarm optimization have been used to solve water injection parameter optimization problems. However, these optimization methods still have limitations. First, traditional optimization methods often rely on static parameter inputs and cannot effectively cope with dynamic changes in reservoir conditions. Second, existing methods mainly adopt global optimization strategies, lacking adaptive control over local reservoir characteristics, which may lead to insufficient or excessive water injection in some areas. In addition, in practical applications, the formulation of water injection strategies involves multiple complex factors, such as formation permeability, fluid flow resistance, and reservoir pressure changes, making it difficult for existing technologies to achieve an effective balance between multi-objective optimization and global dynamic adjustment.

[0005] In terms of remote control and intelligent management of reinjection wells, current oilfield digital monitoring systems still rely heavily on manual intervention. For example, while traditional SCADA (Supervisory Control and Data Acquisition) systems can achieve remote data transmission and storage, they lack intelligent decision-making capabilities and still depend on manual analysis and operation for parameter adjustments. This approach not only increases labor costs but also results in slow water injection response speeds, making it difficult to adapt to real-time reservoir changes. Furthermore, traditional anomaly detection technologies typically use fixed threshold judgment methods, making it difficult to accurately identify abnormal water injection situations caused by dynamic changes in the reservoir, which may lead to reduced water injection efficiency or even induce reservoir damage.

[0006] Therefore, how to provide a method for improving water injection volume in reinjection wells based on fractal network optimization is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose a method for improving water injection volume in reinjection wells based on fractal network optimization. This invention fully utilizes fractal flow network modeling, dynamic Bayesian inference, genetic optimization algorithms, reinforcement learning optimization, and game theory dynamic equilibrium control technologies. It details an intelligent water injection optimization method based on reservoir dynamic response. By constructing an adaptive fractal flow network model, the accuracy of flow simulation in complex reservoir environments is improved. Dynamic optimization methods are used to precisely adjust water injection pressure and flow rate, and intelligent remote control technology is combined to achieve adaptive adjustment of the water injection strategy. This method possesses advantages such as high water injection efficiency, strong reservoir adaptability, high level of intelligence, and strong remote real-time optimization capabilities.

[0008] The method for increasing water injection volume in reinjection wells based on fractal network optimization according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect reservoir geological data, perform standardization processing, construct a multi-source geological database, and use fractal growth algorithm to optimize the seepage network topology to form a multi-scale seepage network model;

[0010] S2. Based on a multi-scale seepage network model, dynamic Bayesian inference is used to predict the evolution trend of reservoir permeability and pore structure, optimize the fractal network topology, and form an adaptive fractal seepage network.

[0011] S3. Based on the adaptive fractal seepage network, a genetic optimization algorithm is used to construct a synergistic optimization objective for water injection pressure and flow rate, calculate the weights of different water injection points, and generate water injection parameter schemes.

[0012] S4. By utilizing water injection parameter schemes and combining them with adaptive fractal seepage network topology changes, the pressure gradient between reinjection wells is adjusted, and the flow matching relationship is optimized to dynamically adjust the water injection strategy.

[0013] S5. Based on the adaptive fractal seepage network and the dynamically adjusted water injection strategy, the key features of reservoir seepage are extracted, the influence factors of different water injection points are calculated, and a dynamically optimized water injection control model is constructed.

[0014] S6. Utilize a dynamically optimized water injection control model, combined with risk assessment methods, to analyze the risks of heterogeneous flow and seepage blockage, and optimize the remote water injection control strategy.

[0015] S7. Based on the dynamically optimized water injection control model and risk assessment results, the water injection parameters are dynamically adjusted and optimized iteratively in real time through a remote monitoring terminal.

[0016] Optionally, S2 specifically includes:

[0017] S21. Based on the constructed multi-scale seepage network model, extract the basic permeability distribution of the reservoir. and pore structure data ,in Using reservoir spatial coordinates, combined with historical water injection data The initial seepage channel distribution is established based on fluid viscosity;

[0018] S22. Predict reservoir permeability changes using dynamic Bayesian inference methods, defining a time series. Calculate the predicted permeability of the reservoir at time t+1:

[0019] ;

[0020] in, , and Update parameters for Bayes. The Laplace operator for permeability, The historical water injection pressure at this location, Let t be the permeability of the reservoir. The predicted permeability of the reservoir at time t+1;

[0021] S23, Based on predicted penetration rate distribution Based on porosity distribution, the adaptive fractal dimension of reservoir seepage channels is calculated. :

[0022] ;

[0023] in, and Let be the effective pore volumes of the seepage channel at times t and t−1, respectively. and The feature scales at times t and t−1 are respectively. These are reservoir microfracture parameters. The weight of the crack effect, As a heterogeneity adjustment factor, is the total number of microfractures in the reservoir, and j is the index of the microfracture;

[0024] S24. Based on the computational adaptive fractal dimension, the topology of the fractal network is optimized. A dynamic fractal percolation optimization model based on heterogeneous coupling scenarios is adopted to establish the fractal network topology optimization equation:

[0025] ;

[0026] in, and These represent the topological states of the reservoir permeation network at times t and t−1, respectively. To optimize the step size for dynamic fractals, This represents the total number of reservoir grid cells. Here, represents the connection strength weight of grid cell i in the seepage network, where i is the grid cell index. Porosity For grid cell i in Location penetration, This is a factor representing localized seepage heterogeneity.

[0027] S25. Based on the optimized fractal network topology, an adaptive fractal seepage network is formed and stored in the reservoir water injection optimization database.

[0028] Optionally, S3 specifically includes:

[0029] S31. Based on the formed adaptive fractal seepage network, extract the permeability of each injection point in the reservoir. Porosity Historical water injection pressure and fluid velocity Construct a reservoir state dataset;

[0030] S32. Establish the objective optimization function, using the injection pressure of the reinjection well as the target optimization function. and traffic As optimization variables, construct a multi-objective optimization model:

[0031] ;

[0032] in, To optimize the objective function, G represents the total number of reinjection wells. For the seepage efficiency of a single reinjection well, For reservoir utilization, Energy consumption for water injection Here, is the weight parameter, max is the maximum value function, and h is the reinjection well index;

[0033] S33. Use a genetic optimization algorithm to optimize parameters and construct the fitness function F:

[0034] ;

[0035] in, These are the system's maximum seepage efficiency, maximum reservoir utilization, and maximum water injection energy consumption, respectively.

[0036] S34. A crossover and mutation strategy based on a genetic optimization algorithm is adopted, which uses a dynamic mutation mechanism based on multi-scale topological perturbation to enhance the global convergence of the optimization variables, and combines local structural dynamic perturbation to optimize the mutation probability. :

[0037] ;

[0038] in, The initial mutation probability, For maximum fitness, For topological perturbation factor, The optimized rate of pressure change;

[0039] S35. Based on the optimized reinjection well pressure and traffic Calculate the optimized water injection gradient distribution :

[0040] ;

[0041] in, This represents the reservoir pressure at the previous moment. This represents the length of the fluid permeation path within the reservoir.

[0042] S36. Based on the optimized water injection gradient distribution and optimized water injection flow rate, the fluid seepage path is calculated using a fluid dynamics model, the reservoir water injection control parameters are adjusted, and the results are stored in the optimization database.

[0043] Optionally, S4 specifically includes:

[0044] S41. Based on the optimized injection pressure and flow rate, and combined with an adaptive fractal seepage network, extract the seepage efficiency at each injection point. A dynamic optimization model based on pressure-seepage synergistic regulation was constructed to calculate the optimized dynamic injection flow rate. :

[0045] ;

[0046] in, As the weight for flow resistance compensation, G represents the total injection flow rate, G represents the total number of reinjection wells, and h represents the reinjection well index.

[0047] S42. Based on the optimized dynamic water injection flow rate and considering reservoir heterogeneity, calculate the reservoir response feedback factor. The final water injection strategy for the reinjection well is adjusted based on a two-way dynamic constraint optimization model.

[0048] ;

[0049] ;

[0050] in, The pressure gradient of the h-th reinjection well is... These are the maximum pressure gradient, maximum injection flow rate, and maximum flow resistance within the system, respectively. This is the flow resistance coefficient. For weight parameters, For dynamic adjustment of flow rate, To ultimately optimize the water injection pressure, To optimize the initial water injection pressure;

[0051] S43. Based on the optimized final reinjection well water injection pressure and optimized water injection flow rate, and combined with the adaptive fractal seepage network topology change, a reservoir adaptive regulation model based on reinforcement learning is constructed:

[0052] ;

[0053] in, and These represent the topological states of the reservoir permeation network at times t and t−1, respectively. To strengthen the weighting of learning strategies, For dynamic topology optimization step size, The weight of the influence of reinjection well h on the reservoir flow topology. Reservoir topology optimization strategies to enhance learning computation;

[0054] S44. Based on the optimized reservoir dynamic topology state and combined with the energy minimization objective of the water injection system, construct dynamic optimization constraints for the water injection strategy:

[0055] ;

[0056] in, The energy consumption for water injection in reinjection wells. and The weight parameters are used to optimize energy consumption, and min is the minimum value function;

[0057] S45. Based on the optimized energy consumption minimization constraint strategy, the reservoir pressure field changes are monitored in real time, the pressure and flow rate of the reinjection well are fine-tuned, and the data are stored in the optimization database.

[0058] Optionally, S5 specifically includes:

[0059] S51. Based on an adaptive fractal flow network, extract key flow characteristics of the reservoir, including the permeability of the injection point. Porosity reservoir pressure and flow resistance The reservoir permeability index was calculated using principal component analysis for dimensionality reduction. :

[0060] ;

[0061] in, Let G be the permeability of the area where the h-th reinjection well is located, G be the total number of reinjection wells, and h be the index of the reinjection wells.

[0062] S52, Calculation-based reservoir permeability index Combined with dynamic water injection flow The analytic hierarchy process (AHP) was used to calculate reservoir influence factors. :

[0063] ;

[0064] in, To optimize reservoir performance, adjust weight parameters. The maximum seepage capacity index within the system. This represents the maximum water injection flow rate within the system. This represents the maximum flow resistance within the system.

[0065] S53. Based on the calculated reservoir influencing factors, a dynamically optimized water injection control model is constructed, and the optimized reservoir response function is calculated using a multi-objective grey prediction method. :

[0066] ;

[0067] in, and These are the minimum and maximum reservoir pressures within the system, respectively. To regulate and adjust the coefficient;

[0068] S54. Based on the calculation of the optimized reservoir response function, the final water injection control factor of the reinjection well is calculated using the Lagrange multiplier method. :

[0069] ;

[0070] in, To regulate and match the weights, This represents the maximum water injection flow rate within the system. This represents the maximum pressure within the system.

[0071] S55. Based on the calculated final water injection control factor and combined with the reservoir topology, fuzzy adaptive control is used for dynamic adjustment calculations to optimize the final water injection pressure. and water injection flow rate :

[0072] ;

[0073] ;

[0074] in, and To adaptively adjust parameters.

[0075] Optionally, S7 specifically includes:

[0076] S71. Based on the risk assessment results and combined with the optimized dynamic water injection control model, remote monitoring data of reinjection wells are collected to construct a comprehensive monitoring dataset. :

[0077] ;

[0078] in, The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. The pressure gradient of the h-th reinjection well is... The flow resistance of the h-th reinjection well is... As the reservoir response feedback factor, These are the energy consumption parameters for reinjection wells. The injection temperature of the reinjection well. The reservoir risk index for reinjection wells;

[0079] S72. Based on the constructed remote monitoring dataset and step risk index, Bayesian risk assessment is used for adaptive anomaly detection to calculate the anomaly deviation value of the reinjection well. :

[0080] ;

[0081] in, This is the actual monitoring data for the h-th reinjection well. This is normal state data predicted based on historical data and statistical modeling. For risk assessment, the weighting parameters are... The risk anomaly contribution value calculated for the Bayesian model;

[0082] S73. Based on the adaptive anomaly detection results, a reinforcement learning method is used to adjust the remote control strategy and construct a reinforcement learning reward function:

[0083] ;

[0084] in, To strengthen the learning reward weighting parameters, This represents the maximum abnormal deviation value within the system. This represents the maximum seepage efficiency within the system. This represents the maximum energy consumption value within the system. Let h be the seepage efficiency of the reinjection well. Here are the energy consumption parameters for the h-th reinjection well. The reinforcement learning reward value for the h-th reinjection well;

[0085] S74. Based on an optimized reinforcement learning-based remote control strategy, combined with a multi-objective optimization method, the final remote control command is calculated:

[0086] ;

[0087] in, Optimize weight parameters for remote control. The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. As the reservoir response feedback factor, For the remote control command of the h-th reinjection well;

[0088] S75. Computation-based remote control commands employ game theory-based dynamic Nash equilibrium optimization for remote closed-loop control:

[0089] ;

[0090] ;

[0091] in, and For adaptive adjustment of parameters in remote control, To ultimately optimize the water injection pressure, To ultimately optimize the water injection flow rate, To adjust the water injection pressure initially, This is for the initial adjustment of the water injection flow rate. To dynamically adjust the gain parameters, For reservoir dynamics influencing factors of reinjection wells, This represents the average reservoir dynamics influence factor across all reinjection wells within the system. For mathematical constants, To optimize the water injection control factors for reinjection wells, This is the average optimized water injection control factor for all reinjection wells within the system.

[0092] The beneficial effects of this invention are:

[0093] This invention comprehensively improves the water injection strategy for reinjection wells through an adaptive fractal network optimization method. Compared with traditional water injection methods, this invention can achieve more precise and intelligent dynamic control under complex reservoir conditions. By establishing an adaptive fractal seepage network model, this invention effectively solves the problem of low accuracy in seepage simulation in heterogeneous reservoir environments using traditional methods. This allows for a more accurate depiction of the dynamic evolution of reservoir seepage paths, ensuring that the water injection strategy can adapt to reservoir changes and improve seepage efficiency. Furthermore, by combining dynamic Bayesian inference, this invention can predict reservoir permeability, pore structure, and fluid flow resistance. This allows the optimization process to not only be based on current data but also to make forward-looking adjustments based on historical trends, enhancing the long-term adaptability of the water injection strategy.

[0094] This invention employs a genetic optimization algorithm to construct a collaborative optimization model for water injection pressure and flow rate. Compared to traditional single-objective optimization methods, this invention comprehensively considers multiple factors such as reservoir pressure gradient, seepage capacity, and energy balance within a multi-objective optimization framework, thereby achieving better water injection parameter configuration and improving overall water injection efficiency. Simultaneously, by combining reinforcement learning optimization, the water injection strategy can continuously optimize itself through feedback and adjustment, improving the system's adaptability and maintaining optimal water injection performance at different reservoir development stages. Furthermore, this invention utilizes game theory-based dynamic Nash equilibrium control technology to achieve optimized collaboration between different reinjection wells, ensuring that under the premise of global optimization, local over-adjustment or resource waste is avoided, making the water injection process more balanced and efficient.

[0095] In terms of remote intelligent control, this invention combines a variational autoencoder (VAE) for anomaly detection. Compared to traditional fixed-threshold monitoring methods, this invention can learn the normal operating status based on historical data and accurately identify anomalies, improving the intelligence level of the monitoring system. Based on optimized reservoir state data, this invention employs a dynamic Nash equilibrium optimization method, enabling the injection parameters of the reinjection well to be dynamically adjusted based on global equilibrium, improving the overall system's adaptability and stability. Furthermore, it combines fuzzy adaptive control (FAC) for final remote closed-loop optimization, allowing the injection parameters to be adjusted in real time according to reservoir changes during actual operation, avoiding resource waste or insufficient injection caused by response lag in traditional methods.

[0096] The advantages of this invention are multifaceted. First, adaptive fractal flow network modeling improves the accuracy of reservoir flow simulation, enabling water injection strategies to more precisely match reservoir characteristics. Second, intelligent optimization algorithms and reinforcement learning allow the water injection strategy to continuously optimize in dynamically changing reservoir environments, enhancing long-term adaptability. Furthermore, remote intelligent control increases the system's automation level, reduces manual intervention, and makes water injection management more efficient and intelligent. Compared to existing technologies, this invention significantly improves reservoir adaptability, water injection accuracy, system intelligence, and energy consumption optimization, effectively increasing the efficiency of reinjection well water injection, extending oilfield development lifespan, and improving economic benefits. Attached Figure Description

[0097] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0098] Figure 1 The flowchart shows the method for increasing water injection volume in reinjection wells based on fractal network optimization proposed in this invention.

[0099] Figure 2 This is a schematic diagram of the topology of the adaptive fractal seepage network model for the method of increasing water injection volume in reinjection wells based on fractal network optimization proposed in this invention. Detailed Implementation

[0100] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0101] refer to Figure 1-2 The method for increasing water injection volume in reinjection wells based on fractal network optimization includes the following steps:

[0102] S1. Collect reservoir geological data, perform standardization processing, construct a multi-source geological database, and use fractal growth algorithm to optimize the seepage network topology to form a multi-scale seepage network model;

[0103] S2. Based on a multi-scale seepage network model, dynamic Bayesian inference is used to predict the evolution trend of reservoir permeability and pore structure, optimize the fractal network topology, and form an adaptive fractal seepage network.

[0104] S3. Based on the adaptive fractal seepage network, a genetic optimization algorithm is used to construct a synergistic optimization objective for water injection pressure and flow rate, calculate the weights of different water injection points, and generate water injection parameter schemes.

[0105] S4. By utilizing water injection parameter schemes and combining them with adaptive fractal seepage network topology changes, the pressure gradient between reinjection wells is adjusted, and the flow matching relationship is optimized to dynamically adjust the water injection strategy.

[0106] S5. Based on the adaptive fractal seepage network and the dynamically adjusted water injection strategy, the key features of reservoir seepage are extracted, the influence factors of different water injection points are calculated, and a dynamically optimized water injection control model is constructed.

[0107] S6. Utilize a dynamically optimized water injection control model, combined with risk assessment methods, to analyze the risks of heterogeneous flow and seepage blockage, and optimize the remote water injection control strategy.

[0108] S7. Based on the dynamically optimized water injection control model and risk assessment results, the water injection parameters are dynamically adjusted and optimized iteratively in real time through a remote monitoring terminal.

[0109] In this embodiment, S2 specifically includes:

[0110] S21. Based on the constructed multi-scale seepage network model, extract the basic permeability distribution of the reservoir. and pore structure data ,in Using reservoir spatial coordinates, combined with historical water injection data The initial seepage channel distribution is established based on fluid viscosity;

[0111] S22. Predict reservoir permeability changes using dynamic Bayesian inference methods, defining a time series. Calculate the predicted permeability of the reservoir at time t+1:

[0112] ;

[0113] in, , and Update parameters for Bayes. The Laplace operator for permeability, The historical water injection pressure at this location, Let t be the permeability of the reservoir. The predicted permeability of the reservoir at time t+1;

[0114] S23, Based on predicted penetration rate distribution Based on porosity distribution, the adaptive fractal dimension of reservoir seepage channels is calculated. :

[0115] ;

[0116] in, and Let be the effective pore volumes of the seepage channel at times t and t−1, respectively. and The feature scales at times t and t−1 are respectively. These are reservoir microfracture parameters. The weight of the crack effect, As a heterogeneity adjustment factor, is the total number of microfractures in the reservoir, and j is the index of the microfracture;

[0117] S24. Based on the computational adaptive fractal dimension, the topology of the fractal network is optimized. A dynamic fractal percolation optimization model based on heterogeneous coupling scenarios is adopted to establish the fractal network topology optimization equation:

[0118] ;

[0119] in, and These represent the topological states of the reservoir permeation network at times t and t−1, respectively. To optimize the step size for dynamic fractals, This represents the total number of reservoir grid cells. Here, represents the connection strength weight of grid cell i in the seepage network, where i is the grid cell index. Porosity For grid cell i in Location penetration, This is a factor representing localized seepage heterogeneity.

[0120] S25. Based on the optimized fractal network topology, an adaptive fractal seepage network is formed and stored in the reservoir water injection optimization database.

[0121] In this embodiment, S3 specifically includes:

[0122] S31. Based on the formed adaptive fractal seepage network, extract the permeability of each injection point in the reservoir. Porosity Historical water injection pressure and fluid velocity Construct a reservoir state dataset;

[0123] S32. Establish the objective optimization function, using the injection pressure of the reinjection well as the target optimization function. and traffic As optimization variables, construct a multi-objective optimization model:

[0124] ;

[0125] in, To optimize the objective function, G represents the total number of reinjection wells. For the seepage efficiency of a single reinjection well, For reservoir utilization, Energy consumption for water injection Here, is the weight parameter, max is the maximum value function, and h is the reinjection well index;

[0126] S33. Use a genetic optimization algorithm to optimize parameters and construct the fitness function F:

[0127] ;

[0128] in, These are the system's maximum seepage efficiency, maximum reservoir utilization, and maximum water injection energy consumption, respectively.

[0129] S34. A crossover and mutation strategy based on a genetic optimization algorithm is adopted, which uses a dynamic mutation mechanism based on multi-scale topological perturbation to enhance the global convergence of the optimization variables, and combines local structural dynamic perturbation to optimize the mutation probability. :

[0130] ;

[0131] in, The initial mutation probability, For maximum fitness, For topological perturbation factor, The optimized rate of pressure change;

[0132] S35. Based on the optimized reinjection well pressure and traffic Calculate the optimized water injection gradient distribution :

[0133] ;

[0134] in, This represents the reservoir pressure at the previous moment. This represents the length of the fluid permeation path within the reservoir.

[0135] S36. Based on the optimized water injection gradient distribution and optimized water injection flow rate, the fluid seepage path is calculated using a fluid dynamics model, the reservoir water injection control parameters are adjusted, and the results are stored in the optimization database.

[0136] In this embodiment, S4 specifically includes:

[0137] S41. Based on the optimized injection pressure and flow rate, and combined with an adaptive fractal seepage network, extract the seepage efficiency at each injection point. A dynamic optimization model based on pressure-seepage synergistic regulation was constructed to calculate the optimized dynamic injection flow rate. :

[0138] ;

[0139] in, As the weight for flow resistance compensation, G represents the total injection flow rate, G represents the total number of reinjection wells, and h represents the reinjection well index.

[0140] S42. Based on the optimized dynamic water injection flow rate and considering reservoir heterogeneity, calculate the reservoir response feedback factor. The final water injection strategy for the reinjection well is adjusted based on a two-way dynamic constraint optimization model.

[0141] ;

[0142] ;

[0143] in, The pressure gradient of the h-th reinjection well is... These are the maximum pressure gradient, maximum injection flow rate, and maximum flow resistance within the system, respectively. This is the flow resistance coefficient. For weight parameters, For dynamic adjustment of flow rate, To ultimately optimize the water injection pressure, To optimize the initial water injection pressure;

[0144] S43. Based on the optimized final reinjection well water injection pressure and optimized water injection flow rate, and combined with the adaptive fractal seepage network topology change, a reservoir adaptive regulation model based on reinforcement learning is constructed:

[0145] ;

[0146] in, and These represent the topological states of the reservoir permeation network at times t and t−1, respectively. To strengthen the weighting of learning strategies, For dynamic topology optimization step size, The weight of the influence of reinjection well h on the reservoir flow topology. Reservoir topology optimization strategies to enhance learning computation;

[0147] S44. Based on the optimized reservoir dynamic topology state and combined with the energy minimization objective of the water injection system, construct dynamic optimization constraints for the water injection strategy:

[0148] ;

[0149] in, The energy consumption for water injection in reinjection wells. and The weight parameters are used to optimize energy consumption, and min is the minimum value function;

[0150] S45. Based on the optimized energy consumption minimization constraint strategy, the reservoir pressure field changes are monitored in real time, the pressure and flow rate of the reinjection well are fine-tuned, and the data are stored in the optimization database.

[0151] In this embodiment, S5 specifically includes:

[0152] S51. Based on an adaptive fractal flow network, extract key flow characteristics of the reservoir, including the permeability of the injection point. Porosity reservoir pressure and flow resistance The reservoir permeability index was calculated using principal component analysis for dimensionality reduction. :

[0153] ;

[0154] in, Let G be the permeability of the area where the h-th reinjection well is located, G be the total number of reinjection wells, and h be the index of the reinjection wells.

[0155] S52, Calculation-based reservoir permeability index Combined with dynamic water injection flow The analytic hierarchy process (AHP) was used to calculate reservoir influence factors. :

[0156] ;

[0157] in, To optimize reservoir performance, adjust weight parameters. The maximum seepage capacity index within the system. This represents the maximum water injection flow rate within the system. This represents the maximum flow resistance within the system.

[0158] S53. Based on the calculated reservoir influencing factors, a dynamically optimized water injection control model is constructed, and the optimized reservoir response function is calculated using a multi-objective grey prediction method. :

[0159] ;

[0160] in, and These are the minimum and maximum reservoir pressures within the system, respectively. To regulate and adjust the coefficient;

[0161] S54. Based on the calculation of the optimized reservoir response function, the final water injection control factor of the reinjection well is calculated using the Lagrange multiplier method. :

[0162] ;

[0163] in, To regulate and match the weights, This represents the maximum water injection flow rate within the system. This represents the maximum pressure within the system.

[0164] S55. Based on the calculated final water injection control factor and combined with the reservoir topology, fuzzy adaptive control is used for dynamic adjustment calculations to optimize the final water injection pressure. and water injection flow rate :

[0165] ;

[0166] ;

[0167] in, and To adaptively adjust parameters.

[0168] In this embodiment, S7 specifically includes:

[0169] S71. Based on the risk assessment results and combined with the optimized dynamic water injection control model, remote monitoring data of reinjection wells are collected to construct a comprehensive monitoring dataset. :

[0170] ;

[0171] in, The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. The pressure gradient of the h-th reinjection well is... The flow resistance of the h-th reinjection well is... As the reservoir response feedback factor, These are the energy consumption parameters for reinjection wells. The injection temperature of the reinjection well. The reservoir risk index for reinjection wells;

[0172] S72. Based on the constructed remote monitoring dataset and step risk index, Bayesian risk assessment is used for adaptive anomaly detection to calculate the anomaly deviation value of the reinjection well. :

[0173] ;

[0174] in, This is the actual monitoring data for the h-th reinjection well. This is normal state data predicted based on historical data and statistical modeling. For risk assessment, the weighting parameters are... The risk anomaly contribution value calculated for the Bayesian model;

[0175] S73. Based on the adaptive anomaly detection results, a reinforcement learning method is used to adjust the remote control strategy and construct a reinforcement learning reward function:

[0176] ;

[0177] in, To strengthen the learning reward weighting parameters, This represents the maximum abnormal deviation value within the system. This represents the maximum seepage efficiency within the system. This represents the maximum energy consumption value within the system. Let h be the seepage efficiency of the reinjection well. Here are the energy consumption parameters for the h-th reinjection well. The reinforcement learning reward value for the h-th reinjection well;

[0178] S74. Based on an optimized reinforcement learning-based remote control strategy, combined with a multi-objective optimization method, the final remote control command is calculated:

[0179] ;

[0180] in, Optimize weight parameters for remote control. The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. As the reservoir response feedback factor, For the remote control command of the h-th reinjection well;

[0181] S75. Computation-based remote control commands employ game theory-based dynamic Nash equilibrium optimization for remote closed-loop control:

[0182] ;

[0183] ;

[0184] in, and For adaptive adjustment of parameters in remote control, To ultimately optimize the water injection pressure, To ultimately optimize the water injection flow rate, To adjust the water injection pressure initially, This is for the initial adjustment of the water injection flow rate. To dynamically adjust the gain parameters, For reservoir dynamics influencing factors of reinjection wells, This represents the average reservoir dynamics influence factor across all reinjection wells within the system. For mathematical constants, To optimize the water injection control factors for reinjection wells, This is the average optimized water injection control factor for all reinjection wells within the system.

[0185] Example 1:

[0186] In this oilfield block, six representative reinjection wells were selected, located in different reservoir regions with varying reservoir characteristics. Some of these wells are located in high-permeability zones, posing a risk of over-injection and reservoir damage, while others are located in low-permeability zones, experiencing chronic under-injection, making it difficult to maintain formation pressure and impacting the recovery rate of adjacent production wells. This invention first collects reservoir geological data for this area, including formation permeability, porosity, flow resistance, historical water injection volume of reinjection wells, and oil-water ratio trends. Then, an adaptive fractal flow network model for this block is constructed using a fractal flow network modeling method.

[0187] In the data processing stage, this invention employs dynamic Bayesian inference to predict the evolution trends of reservoir permeability and pore structure, establishing a dynamic evolution model for the reservoir. Prediction results show that permeability in some areas of the block has decreased by 18% over the past five years, and porosity in other areas has decreased by 12%. This indicates that traditional water injection strategies have failed to effectively adapt to reservoir changes, leading to a long-term decline in water injection efficiency. Based on these prediction data, this invention intelligently optimizes the water injection parameters of reinjection wells.

[0188] During the optimization process, a genetic optimization algorithm was used to construct a synergistic optimization objective for injection pressure and flow rate, and the weights of different injection points were calculated. Optimization results show that, compared to the original injection strategy, the optimized pressure distribution is more balanced, the average flow resistance of the reinjection wells is reduced by 15%, and the pressure gradient is decreased by 8%, effectively improving reservoir seepage conditions. Furthermore, this invention employs a reinforcement learning optimization method, enabling the injection system to continuously optimize its strategy based on historical data. After six months of operation, the optimized reinjection well water injection volume increased by 12% compared to traditional methods, with a 10% reduction in water injection volume in high-permeability zones and an 18% increase in water injection volume in low-permeability zones, ensuring balanced development of the overall reservoir.

[0189] This invention also incorporates remote intelligent monitoring technology, employing a variational autoencoder (VAE) for anomaly detection to identify abnormal conditions in reinjection wells in real time. During actual operation, the system successfully detected three instances of reservoir blockage risk and proactively optimized adjustments, preventing potential water injection anomalies. Furthermore, this invention utilizes a game-theoretic dynamic Nash equilibrium optimization method to achieve more coordinated pressure regulation between different reinjection wells, avoiding excessively high or low local pressures and improving water injection efficiency.

[0190] After implementing this invention in this oilfield block, the system's automation level was significantly improved. Compared to traditional manual adjustment methods, the remote intelligent optimization strategy reduced the need for manual intervention by 70%, improving the operating efficiency of the water injection system. Remote closed-loop optimization using fuzzy adaptive control (FAC) ensured that water injection parameters could be adjusted in real time according to reservoir changes. Ultimately, the application of this invention in this oilfield block yielded significant results, reducing overall water injection energy consumption by 14% and increasing oil well recovery by 7.5%, effectively enhancing the oilfield's economic benefits.

[0191] Table 1. Comparison of key parameters before and after optimization of reinjection wells in oilfield blocks.

[0192] Injection well number Water injection volume before optimization (m³ / d) Optimized water injection volume (m³ / d) Flow resistance before optimization (MPa) Optimized flow resistance (MPa) Formation pressure before optimization (MPa) Optimized formation pressure (MPa) Number of times of blockage risk Recovery rate before optimization (%) Optimized recovery rate (%) W-101 180.0 162.0 5.2 4.3 12.5 13.2 1.0 29.5 31.8 W-102 95.0 110.0 6.0 5.0 11.2 12.1 0.0 26.1 27.9 W-103 210.0 185.0 4.8 4.0 13.8 14.5 1.0 30.3 32.7 W-104 88.0 105.0 6.5 5.4 10.9 12.0 0.0 25.5 27.2 W-105 230.0 195.0 5.0 4.1 13.2 14.0 1.0 31.0 33.2 W-106 90.0 108.0 6.2 5.2 11.0 12.3 0.0 26.0 28.0 average value 148.83 144.17 5.62 4.67 12.1 13.0 0.5 28.1 30.1

[0193] The optimization strategy of this invention has achieved significant results in the water injection optimization project of reinjection wells in an oilfield block in Xinjiang. After optimization, the average daily water injection volume of the reinjection wells was adjusted from 148.83 m³ to 144.17 m³, effectively controlling excessive water injection in high-permeability zones while increasing the supplementary water injection volume in low-permeability zones. For example, the daily water injection volume of well W-105 was reduced by 15.2%, avoiding reservoir damage, while the daily water injection volume of well W-102 increased by 15.8%, improving the water injection effect in low-permeability zones.

[0194] Regarding the optimization of flow resistance, the average flow resistance decreased from 5.62 MPa to 4.67 MPa after optimization, a reduction of 16.9%, indicating an improvement in reservoir seepage conditions. For example, the flow resistance of well W-104 decreased by 1.1 MPa, reducing energy consumption during water flow and improving seepage capacity.

[0195] Regarding formation pressure, the optimized average pressure increased from 12.1 MPa to 13.0 MPa, an increase of 7.4%. Among them, the pressure of well W-102 increased by 0.9 MPa, which improved the reservoir energy replenishment capacity, and the pressure of well W-101 increased by 0.7 MPa, indicating that the optimized water injection strategy effectively enhanced the reservoir support capacity.

[0196] The optimized system reduced the risk of blockage. Before optimization, the monitoring system detected an average of 0.5 blockage risks, which was reduced to 0 after optimization. This avoided reservoir damage caused by excessive water injection. For example, wells W-103 and W-105 no longer showed any blockage risk after optimization.

[0197] Ultimately, the optimized strategy of this invention increased the average recovery rate from 28.1% to 30.1%, an increase of 2.0 percentage points. The recovery rate of well W-105 increased by 2.2%, and that of well W-103 by 2.4%, indicating that the optimized strategy improved reservoir recovery efficiency and enhanced water drive performance. The application of this invention effectively improved the intelligence level of the water injection system, reduced energy consumption, and increased the economic benefits of oilfield development.

[0198] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for increasing water injection volume in reinjection wells based on fractal network optimization, characterized in that, Includes the following steps: S1. Collect reservoir geological data, perform standardized processing, construct a multi-source geological database, and use fractal growth algorithm to optimize the seepage network topology to form a multi-scale seepage network model; S2. Based on a multi-scale seepage network model, dynamic Bayesian inference is used to predict the evolution trend of reservoir permeability and pore structure, optimize the fractal network topology, and form an adaptive fractal seepage network. S3. Based on the adaptive fractal seepage network, a genetic optimization algorithm is used to construct a synergistic optimization objective for water injection pressure and flow rate, calculate the weights of different water injection points, and generate water injection parameter schemes. S4. By utilizing water injection parameter schemes and combining them with adaptive fractal seepage network topology changes, the pressure gradient between reinjection wells is adjusted, and the flow matching relationship is optimized to dynamically adjust the water injection strategy. S5. Based on the adaptive fractal seepage network and the dynamically adjusted water injection strategy, the key features of reservoir seepage are extracted, the influence factors of different water injection points are calculated, and a dynamically optimized water injection control model is constructed. S6. Utilize a dynamically optimized water injection control model, combined with risk assessment methods, to analyze the risks of heterogeneous flow and seepage blockage, and optimize the remote water injection control strategy. S7. Based on the dynamically optimized water injection control model and risk assessment results, the water injection parameters are dynamically adjusted and optimized iteratively in real time through a remote monitoring terminal; The optimized fractal network topology includes: A dynamic fractal seepage optimization model based on heterogeneous coupling scenarios is used to establish the fractal network topology optimization equation: ； in, and These represent the reservoir flow network at times t and t, respectively. The topological state, To optimize the step size for dynamic fractals, This represents the total number of reservoir grid cells. Here, represents the connection strength weight of grid cell i in the seepage network, where i is the grid cell index. Porosity For grid cell i in Location penetration, This is a factor representing the heterogeneity of local seepage.

2. The method for increasing water injection volume in reinjection wells based on fractal network optimization according to claim 1, characterized in that, S2 specifically includes: S21. Based on the constructed multi-scale seepage network model, extract the basic permeability distribution of the reservoir. and pore structure data ,in Using reservoir spatial coordinates, combined with historical water injection data The initial seepage channel distribution is established based on fluid viscosity; S22. Predict reservoir permeability changes using dynamic Bayesian inference methods, defining a time series. Calculate the predicted permeability of the reservoir at time t+1: ； in, , and Update parameters for Bayes. The Laplace operator for permeability, The historical water injection pressure at this location, Let t be the permeability of the reservoir. The predicted permeability of the reservoir at time t+1; S23, Based on predicted penetration rate distribution Based on porosity distribution, the adaptive fractal dimension of reservoir seepage channels is calculated. : ； in, and They are time t and Effective pore volume of the seepage channel and They are time t and The feature scale, These are reservoir microfracture parameters. The weight of the crack effect, This is a heterogeneity adjustment factor. is the total number of microfractures in the reservoir, and j is the index of the microfracture; S24. Optimize the topology of fractal networks based on computational adaptive fractal dimension; S25. Based on the optimized fractal network topology, an adaptive fractal seepage network is formed and stored in the reservoir water injection optimization database.

3. The method for increasing water injection volume in reinjection wells based on fractal network optimization according to claim 1, characterized in that, S3 specifically includes: S31. Based on the formed adaptive fractal seepage network, extract the permeability of each injection point in the reservoir. Porosity Historical water injection pressure and fluid velocity Construct a reservoir state dataset; S32. Establish the objective optimization function, using the injection pressure of the reinjection well as the target optimization function. and traffic As optimization variables, construct a multi-objective optimization model: ； in, To optimize the objective function, G represents the total number of reinjection wells. For the seepage efficiency of a single reinjection well, For reservoir utilization, Energy consumption for water injection Here, is the weight parameter, max is the maximum value function, and h is the reinjection well index; S33. Use a genetic optimization algorithm to optimize parameters and construct the fitness function F: ； in, These are the system's maximum seepage efficiency, maximum reservoir utilization, and maximum water injection energy consumption, respectively. S34. A crossover and mutation strategy based on a genetic optimization algorithm is adopted, which uses a dynamic mutation mechanism based on multi-scale topological perturbation to enhance the global convergence of the optimization variables, and combines local structural dynamic perturbation to optimize the mutation probability. : ； in, The initial mutation probability, For maximum fitness, For topological perturbation factor, The optimized rate of pressure change; S35. Based on the optimized reinjection well pressure and traffic Calculate the optimized water injection gradient distribution : ； in, This represents the reservoir pressure at the previous moment. The length of the fluid permeation path inside the reservoir; S36. Based on the optimized water injection gradient distribution and optimized water injection flow rate, the fluid seepage path is calculated using a fluid dynamics model, the reservoir water injection control parameters are adjusted, and the results are stored in the optimization database.

4. The method for increasing water injection volume in reinjection wells based on fractal network optimization according to claim 1, characterized in that, S4 specifically includes: S41. Based on the optimized injection pressure and flow rate, and combined with an adaptive fractal seepage network, extract the seepage efficiency at each injection point. A dynamic optimization model based on pressure-seepage synergistic regulation was constructed to calculate the optimized dynamic injection flow rate. : ； in, As the weight for flow resistance compensation, G represents the total injection flow rate, G represents the total number of reinjection wells, and h represents the reinjection well index. S42. Based on the optimized dynamic water injection flow rate and considering reservoir heterogeneity, calculate the reservoir response feedback factor. The final water injection strategy for the reinjection well is adjusted based on a two-way dynamic constraint optimization model. ； ； in, The pressure gradient of the h-th reinjection well is... These are the maximum pressure gradient, maximum injection flow rate, and maximum flow resistance within the system, respectively. This is the flow resistance coefficient. For weight parameters, For dynamic adjustment of flow rate, To ultimately optimize the water injection pressure, To optimize the initial water injection pressure; S43. Based on the optimized final reinjection well water injection pressure and optimized water injection flow rate, combined with the adaptive fractal seepage network topology change, a reservoir adaptive regulation model based on reinforcement learning is constructed: ； in, and These represent the reservoir flow network at times t and t, respectively. The topological state, To strengthen the weighting of learning strategies, For dynamic topology optimization step size, The weight of the influence of reinjection well h on the reservoir flow topology. Reservoir topology optimization strategies to enhance learning computation; S44. Based on the optimized reservoir dynamic topology state and combined with the energy minimization objective of the water injection system, construct dynamic optimization constraints for the water injection strategy: ； in, The energy consumption for water injection in reinjection wells. and The weight parameters are used for energy consumption optimization, and min is the minimum value function; S45. Based on the optimized energy consumption minimization constraint strategy, the reservoir pressure field changes are monitored in real time, the pressure and flow rate of the reinjection well are fine-tuned, and the data are stored in the optimization database.

5. The method for increasing water injection volume in reinjection wells based on fractal network optimization according to claim 1, characterized in that, S5 specifically includes: S51. Based on an adaptive fractal flow network, extract key flow characteristics of the reservoir, including the permeability of the injection point. Porosity reservoir pressure and flow resistance The reservoir permeability index was calculated using principal component analysis for dimensionality reduction. : ； in, Let G be the permeability of the area where the h-th reinjection well is located, G be the total number of reinjection wells, and h be the index of the reinjection wells. S52, Calculation-based reservoir permeability index Combined with dynamic water injection flow The analytic hierarchy process (AHP) was used to calculate reservoir influence factors. : ； in, To optimize reservoir performance, adjust weight parameters. The maximum seepage capacity index within the system. This represents the maximum water injection flow rate within the system. This represents the maximum flow resistance within the system. S53. Based on the calculated reservoir influencing factors, a dynamically optimized water injection control model is constructed, and the optimized reservoir response function is calculated using a multi-objective grey prediction method. : ； in, and These are the minimum and maximum reservoir pressures within the system, respectively. To regulate and adjust the coefficient; S54. Based on the calculation of the optimized reservoir response function, the final water injection control factor of the reinjection well is calculated using the Lagrange multiplier method. : ； in, To regulate and match the weights, This represents the maximum water injection flow rate within the system. This represents the maximum pressure within the system. S55. Based on the calculated final water injection control factor and combined with the reservoir topology, fuzzy adaptive control is used for dynamic adjustment calculations to optimize the final water injection pressure. and water injection flow rate : ； ； in, and To adaptively adjust parameters.

6. The method for increasing water injection volume in reinjection wells based on fractal network optimization according to claim 1, characterized in that, Specifically, S7 includes: S71. Based on the risk assessment results and combined with the optimized dynamic water injection control model, remote monitoring data of reinjection wells are collected to construct a comprehensive monitoring dataset. : ； in, The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. The pressure gradient of the h-th reinjection well is... The flow resistance of the h-th reinjection well is... As the reservoir response feedback factor, These are the energy consumption parameters for reinjection wells. The injection temperature of the reinjection well. The reservoir risk index for reinjection wells; S72. Based on the constructed remote monitoring dataset and step risk index, Bayesian risk assessment is used for adaptive anomaly detection to calculate the anomaly deviation value of the reinjection well. : ； in, This is the actual monitoring data for the h-th reinjection well. This is normal state data predicted based on historical data and statistical modeling. For risk assessment, the weighting parameters are... The risk anomaly contribution value calculated for the Bayesian model; S73. Based on the adaptive anomaly detection results, a reinforcement learning method is used to adjust the remote control strategy and construct a reinforcement learning reward function: ； in, To strengthen the learning reward weighting parameters, This represents the maximum abnormal deviation value within the system. This represents the maximum seepage efficiency within the system. This represents the maximum energy consumption value within the system. Let h be the seepage efficiency of the reinjection well. Here are the energy consumption parameters for the h-th reinjection well. The reinforcement learning reward value for the h-th reinjection well; S74. Based on an optimized reinforcement learning-based remote control strategy, combined with a multi-objective optimization method, the final remote control command is calculated: ； in, Optimize weight parameters for remote control. The final optimized injection pressure for the h-th reinjection well. The final optimized injection flow rate for the h-th reinjection well. As the reservoir response feedback factor, For the remote control command of the h-th reinjection well; S75. Computation-based remote control commands employ game theory-based dynamic Nash equilibrium optimization for remote closed-loop control: ； ； in, and For adaptive adjustment of parameters in remote control, To ultimately optimize the water injection pressure, To ultimately optimize the water injection flow rate, To adjust the water injection pressure initially, This is for the initial adjustment of the water injection flow rate. To dynamically adjust the gain parameters, For reservoir dynamics influencing factors of reinjection wells, This represents the average reservoir dynamics influence factor across all reinjection wells within the system. For mathematical constants, To optimize the water injection control factors for reinjection wells, This is the average optimized water injection control factor for all reinjection wells within the system.

Images