Hydrothermal geothermal group well borehole trajectory optimization method and system

Abstract

The present application relates to a kind of hydrothermal geothermal group well borehole trajectory collaborative optimization method and system, to solve the problem of low heat extraction efficiency, high drilling cost, serious well interference in traditional geothermal group well development. Method includes: input group well trajectory parameters and geological information (seismic / logging data, fracture network model, drilling constraints);Three-dimensional thermal reservoir model containing fracture network is constructed and permeability and temperature field parameters are extracted;Through machine learning model, heat-flow coupling numerical simulation is called, and the heat extraction efficiency, drilling cost and well interference index of different trajectory combinations are calculated;Multi-objective function F = α · Q heat –β C drill – γI interference Quantitative evaluation results, non-dimensional and weighted generation normalized results;Based on threshold value, whether trajectory is adjusted is judged, improved multi-objective particle swarm optimization algorithm is used to optimize parameters, and iterative feedback is carried out to heat-flow coupling simulation until convergence;While drilling stage, trajectory is dynamically corrected by random forest algorithm.

Description

Technical Field

[0001] This invention relates to the field of wellbore trajectory technology for hydrothermal geothermal well clusters, and specifically to a method and system for optimizing the wellbore trajectory of hydrothermal geothermal well clusters. Background Technology

[0002] With the increasing global demand for clean energy, hydrothermal geothermal resources have become a research hotspot in the energy field due to their cleanliness and renewability. In the development of hydrothermal geothermal resources, the rational layout of well systems and optimization of well trajectories are crucial for improving heat extraction efficiency, reducing development costs, and minimizing inter-well interference. Traditional well trajectory design methods often lack comprehensive consideration of the complex geological conditions of the geothermal reservoir, making it difficult to achieve multi-objective synergistic optimization. This leads to problems such as low heat extraction efficiency, excessively high drilling costs, and severe inter-well interference, limiting the efficient development and utilization of hydrothermal geothermal resources. Summary of the Invention

[0003] This invention aims to provide a method and system for optimizing the wellbore trajectory of hydrothermal geothermal clusters. Through multi-objective collaborative optimization, dynamic feedback mechanism and improved algorithm, it realizes the scientific design and precise adjustment of the well trajectory, thereby improving the development efficiency and economic benefits of hydrothermal geothermal resources.

[0004] The main objective of this application is to provide a method for collaborative optimization of wellbore trajectories in hydrothermal geothermal well groups, comprising the following steps: Input a set of well trajectory parameters and geological information including at least seismic data, well logging data, fracture network model, and drilling engineering constraints; A three-dimensional geological model of a thermal reservoir containing a fracture network was constructed, and reservoir permeability and temperature field parameters were extracted. A machine learning model is constructed, and a thermal-fluid coupled numerical simulation model set in the machine learning model is called to calculate the heat recovery efficiency, drilling cost and inter-well interference index of different well trajectory combinations. By setting a multi-objective function in the machine learning model Quantitative analysis and evaluation results To collect heat, For drilling costs, Inter-well interference index, α ,β, γ These are the weighting coefficients; The analysis and evaluation results are dimensionless and multiplied by weights to generate normalized results. Based on the judgment benchmark value or the threshold of qualified items, determine whether trajectory adjustment is needed. If adjustment is needed, select the evaluation item with the largest normalized value first. An improved multi-objective particle swarm optimization algorithm is used to adjust trajectory parameters, and global / local searches are dynamically balanced through adaptive inertial weights. Drilling engineering constraints are satisfied through the penalty function method. The adjusted trajectory parameters are fed back into the thermal-fluid coupling simulation step for iterative optimization until the termination condition is met. During the drilling process, real-time logging data is input, and trajectory parameters are dynamically corrected using a random forest algorithm.

[0005] Furthermore, the machine learning model is also configured to: A switching module that switches between the initial planning state and the dynamic optimization state of the well trajectory parameter set; The multiple well trajectories to be modeled are configured in the machine learning model in the initial planning state; By switching the machine learning model from the initial planning state to the dynamic optimization state, the machine learning model is used to perform collaborative constraint optimization on the trajectory parameter set of multiple wells; Based on the results of thermal-fluid coupling numerical simulation, the trajectory parameters of the group wells are dynamically adjusted to maximize heat recovery, minimize inter-well interference, and minimize drilling costs. During the drilling process, the set of trajectory parameters is dynamically adjusted based on real-time logging data using a random forest algorithm.

[0006] Furthermore, the drilling engineering constraints include build-up rate limits, collision avoidance distances, and fracture penetration conditions; The optimized trajectory needs to be verified for spatial safety through three-dimensional anti-collision scanning, and to ensure maximum contact with high-permeability fracture zones.

[0007] Furthermore, the normalization result can be determined using any of the following methods: (a) Calculate the total normalized value; if it exceeds the baseline value, an adjustment is triggered. (b) Count the number of qualified items that meet the benchmark, and trigger adjustments if the number is lower than the threshold.

[0008] Furthermore, the improved PSO algorithm employs an adaptive inertia weight strategy: In the early stages of iteration, the global search capability is increased, and in the later stages, the local search accuracy is enhanced; constraint handling strategy: the penalty function method is used to force the slope rate limit and collision avoidance distance to be met.

[0009] Furthermore, the state switching of the switching module is configured as follows: Initial planning status: Configure well trajectory parameters into the 3D geological model; Dynamic optimization state: The trajectory is collaboratively optimized based on the results of thermal-fluid coupling simulation, and crack network penetration optimization is performed.

[0010] Furthermore, the random forest algorithm outputs a set of trajectory adjustment suggestions for the next drilling segment and updates them synchronously to all associated wellbores.

[0011] This invention also provides a wellbore trajectory collaborative optimization system for hydrothermal geothermal well clusters, used to implement the aforementioned wellbore trajectory collaborative optimization method for hydrothermal geothermal well clusters, comprising: The input module is used to input a set of well trajectory parameters and geological information including at least seismic data, well logging data, fracture network model, and drilling engineering constraints. The geological modeling module is connected to the input module and generates a three-dimensional thermal reservoir model containing a fracture network based on the geological information. The machine learning model includes modules for numerical simulation, analysis and evaluation, collaborative optimization, engineering constraints, dynamic adjustment, and switching. The numerical simulation module is used to perform thermal-fluid coupling simulation and output the heat collection efficiency and interference intensity. The analysis and evaluation module quantifies the evaluation results through a multi-objective function and performs weighted normalization processing. The collaborative optimization module embeds an improved multi-objective PSO algorithm to optimize the trajectory parameter set in batches. The engineering constraint module is used to verify the slope, anti-collision distance and crack penetration conditions, and generate a three-dimensional anti-collision report. The dynamic adjustment module is used to input real-time logging-while-drilling data and dynamically correct trajectory parameters using a random forest algorithm. The switching module is used to seamlessly transition between the initial planning state and the dynamic optimization state.

[0012] Furthermore, the collaborative optimization module and the dynamic adjustment module adopt a parallel computing architecture to simultaneously process the simulation, optimization and real-time correction of multi-wellbore trajectory combinations.

[0013] Furthermore, the improved PSO algorithm is integrated into the collaborative optimization module, which adjusts the search capability through adaptive inertia weights and uses the penalty function method to handle drilling engineering constraints.

[0014] The beneficial effects are as follows: This application employs multi-objective collaborative optimization to improve thermal recovery efficiency and economy. Traditional geothermal well group design often optimizes the trajectory of a single well in isolation, neglecting inter-well interference and cost constraints, resulting in low thermal recovery efficiency. This method uses a multi-objective function... Quantitative analysis and evaluation results To collect heat, For drilling costs, Inter-well interference index, α ,β, γTo accurately quantify the heat exchange efficiency of different trajectory combinations in the fracture network for the weighted coefficient thermal-fluid coupling numerical simulation, and to ensure that the wellbore maximizes contact with high-permeability fracture zones by combining the permeability field distribution, the particle swarm optimization algorithm is improved to adaptively balance global search (initial stage) and local fine-tuning optimization (later stage) during iteration, avoiding getting trapped in local optima.

[0015] This application ensures accurate trajectory execution through dynamic real-time correction, as geological uncertainties during drilling can easily cause the actual trajectory to deviate from the design. This application introduces a random forest algorithm to dynamically analyze logging-while-drilling data (such as gamma and resistivity curves), identify unpredicted high-permeability fracture zones or fault barriers in real time, generate trajectory adjustment suggestions for the next drilling segment, and simultaneously update the associated wellbore. Combined with a "state switching module," the system can seamlessly switch between initial planning (static geological model) and dynamic optimization (real-time data-driven) modes, ensuring that the trajectory conforms to both long-term geothermal reservoir development strategies and adapts to short-term geological changes.

[0016] This application ensures drilling safety through intelligent embedding of engineering constraints. Traditional optimization methods often overlook engineering feasibility, leading to excessive build-up rates or wellbore collisions. This invention embeds drilling constraints (build-up rate, anti-collision distance, and fracture penetration conditions) into the PSO algorithm using a penalty function method: when trajectory parameters violate constraints, the objective function value is penalized and reduced, forcing the algorithm to automatically search for feasible solutions. The 3D anti-collision scanning module constructs the wellbore spatial envelope based on B-spline curves, calculates the minimum distance, and generates a collision risk report. The fracture penetration condition requires the trajectory to pass through a preset high thermal conductivity fracture element; otherwise, the penalty function is triggered.

[0017] This application deeply integrates machine learning and numerical simulation. It constructs a "machine learning model" framework that integrates heat-fluid coupled numerical simulation, a PSO optimizer, and a random forest dynamic correction module. The numerical simulation solves the fluid flow-heat transfer equations in fractured media using the finite element method, inputting the permeability tensor and temperature field boundary conditions, and outputting heat recovery efficiency and disturbance intensity. The PSO algorithm uses simulation results as a fitness function to batch optimize the set of well trajectory parameters. The random forest establishes a mapping relationship between well logging data and trajectory corrections. These three components work collaboratively through a "switching module," solving the problems of low computational efficiency and disconnect between optimization and simulation in traditional methods.

[0018] This application features a parallel architecture for efficient collaborative optimization of multiple wells. Addressing the challenge of processing large datasets from multiple well systems, this application employs a parallel computing architecture: the collaborative optimization module divides the trajectory parameters of multiple wells into sub-tasks, simultaneously running thermal-fluid coupling simulation and PSO search; the dynamic adjustment module processes the drilling data streams from multiple wells in parallel. A state switching module coordinates resource allocation, invoking a high-precision geological model during the initial planning phase and focusing on real-time data response during the dynamic optimization phase. Furthermore, a dual-threshold criterion for normalized results (total value exceeding the benchmark or insufficient number of qualified items) ensures a flexible and reliable optimization triggering mechanism. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system framework principle of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 and Figure 2 The main objective of this application is to provide a collaborative optimization method for wellbore trajectories in hydrothermal geothermal clusters, comprising the following steps: inputting a set of well trajectory parameters and geological information including at least seismic data, logging data, fracture network model, and drilling engineering constraints; constructing a three-dimensional geothermal reservoir geological model containing a fracture network and extracting reservoir permeability and temperature field parameters; constructing a machine learning model and calling a thermal-fluid coupling numerical simulation model set in the machine learning model to calculate the thermal recovery efficiency, drilling cost, and inter-well interference index for different wellbore trajectory combinations; and using a multi-objective function set in the machine learning model. Quantitative analysis and evaluation results To collect heat, For drilling costs, Inter-well interference index, α ,β, γThe weighting coefficients are used to perform dimensionless processing on the analysis and evaluation results and multiply them by the weights to generate normalized results. Based on the judgment benchmark value or the threshold value of qualified items, it is determined whether trajectory adjustment is needed. If adjustment is needed, the evaluation item with the largest normalized value is selected first. An improved multi-objective particle swarm optimization algorithm is used to adjust the trajectory parameters. Global / local search is dynamically balanced through adaptive inertial weights, and drilling engineering constraints are satisfied through the penalty function method. The adjusted trajectory parameters are fed back to the thermal-fluid coupling simulation step for iterative optimization until the termination condition is met. During drilling, real-time logging data is input, and the trajectory parameters are dynamically corrected through the random forest algorithm.

[0022] In the above embodiments, the input parameters mainly include a set of well trajectory parameters and geological information, wherein, The set of well trajectory parameters includes initial design parameters such as wellhead coordinates, target coordinates, well depth, inclination angle, and azimuth angle for each well. Geological information includes at least four types of data: seismic data: providing stratigraphic structure and fault distribution; well logging data: obtaining in-situ permeability, porosity, and temperature profiles; fracture network model: fracture geometry parameters (strike, dip, density) generated based on the Discrete Fracture Network (DFN) method; and drilling engineering constraints: build-up rate limits (typically ≤3° / 30m), anti-collision distance (≥5m), and fracture penetration conditions (requiring the trajectory to pass through fracture cells with permeability >100mD).

[0023] In the above embodiments, three-dimensional geothermal reservoir geological modeling involves constructing a structured mesh model by integrating input data, embedding a fracture network into the matrix rock blocks. The permeability field is generated by interpolating well logging data and coupling it with the DFN model, while the temperature field is constrained by the geothermal gradient equation. The key outputs are the permeability tensor K and temperature value T for each mesh cell.

[0024] In the above embodiments, the machine learning model, acting as a scheduling framework, calls the built-in thermal-fluid coupling numerical simulator to solve the following equations: Continuity equation: ; Darcy's Law: ; Energy equation: ; in, For fluid density, Viscosity, For the combined heat capacity of rock / fluid, The overall thermal conductivity is used to simulate and output the heat extraction efficiency. Heat harvesting (heat taken per unit time, unit: MW). Drilling cost (calculated based on trajectory length and construction difficulty). The inter-well interference index (defined as the inter-well pressure drop correlation coefficient).

[0025] In some embodiments, the normalization and adjustment criteria are as follows: for Heat collection, Drilling costs, Interference indices are dimensionless (e.g., divided by the historical maximum value) and then multiplied by weights to generate a comprehensive score. Two methods are used to determine adjustments: (a) Total value exceeds the benchmark: if the weighted total score is less than the preset threshold (e.g., 0.7), optimization is triggered; (b) Insufficient number of qualified items: if more than two scores are less than the sub-threshold (e.g., 0.6), optimization is triggered.

[0026] The improved PSO algorithm is calculated using the following formula:

[0027] in For constraint functions (such as collision avoidance distance requirements) ), As a penalty factor.

[0028] The adjusted parameters are fed back into the thermal-fluid coupling simulation for re-evaluation, and this process is repeated until the rate of change of F is less than 1% or the maximum number of iterations is reached. During the drilling phase, the logging-while-drilling data is analyzed using a random forest model, and the trajectory adjustment is output in real time.

[0029] In some embodiments, the machine learning model is further configured as follows: a switching module that switches between the initial planning state and the dynamic optimization state of the well trajectory parameter set; configuring the multiple well trajectories to be modeled in the machine learning model in the initial planning state; using the machine learning model to perform collaborative constraint optimization of the trajectory parameter set of multiple wells by switching the machine learning model from the initial planning state to the dynamic optimization state; dynamically correcting the well trajectory parameters based on the results of thermal-fluid coupling numerical simulation to maximize heat recovery, minimize inter-well interference, and minimize drilling costs; and dynamically adjusting the trajectory parameter set using a random forest algorithm based on real-time logging data during drilling.

[0030] In the above embodiments, the switching module is configured as follows: Initial planning state: Load the static geological model and configure the initial trajectory parameters of the well group. Dynamic optimization state: Initiate collaborative optimization based on real-time simulation results, focusing on handling inter-well interference (such as adjusting the well spacing to >300m) and fracture penetration (directional drilling into high-permeability zones).

[0031] In the above embodiments, within the multi-objective PSO framework, the overall F-value of the well group is used as the fitness function to simultaneously optimize the trajectory parameters of all wells. For example, to reduce interference, the distance between production wells and injection wells is automatically increased; to reduce costs, the length of the horizontal section is shortened while ensuring passage through densely fractured areas.

[0032] In the above embodiments, during dynamic adjustments while drilling, the training data for the random forest model includes historical logging curves and corresponding optimal trajectory corrections. During real-time drilling, data such as gamma and neutron porosity of the current well section are input, and the model outputs suggestions for adjusting the inclination / azimuth angle of the next well section (such as increasing the dip angle by 2° to track fracture zones).

[0033] In some embodiments, the drilling engineering constraints include build-up rate limits, anti-collision distances, and fracture penetration conditions; the optimized trajectory needs to be verified for spatial safety through three-dimensional anti-collision scanning, and to ensure maximum contact with high-permeability fracture zones.

[0034] In the above embodiments, drilling engineering constraints include: build-up rate limit: maximum allowable build-up rate k max≤3 。 / 30m, to prevent drill string jamming. Anti-collision distance: the minimum three-dimensional Euclidean distance between wellbores. d min≥5m. Fracture penetration condition: The trajectory must pass through at least 3 fracture elements with permeability >100mD. 3D anti-collision scanning: The wellbore trajectory is fitted using a B-spline curve, and the minimum spatial distance is calculated using the following formula: d min=min‖r1( s )-r2( t )‖( s , t If ∈[0,1]), d If the depth is less than 5m, the system will automatically mark high-risk well sections and adjust the target point.

[0035] In some embodiments, the normalization result is determined in one of the following ways: (a) calculating the total normalized value; if it exceeds the benchmark value, adjustment is triggered; (b) counting the number of qualified items that meet the benchmark; if it is below the threshold, adjustment is triggered. The total normalized value can be determined by applying a multi-objective function. The weighted calculation is performed. The baseline value for the weighted calculation is set to 0.7 times F.

[0036] In some embodiments, the improved PSO algorithm employs an adaptive inertia weighting strategy that increases global search capability in the early stages of iteration and enhances local search accuracy in the later stages; and a constraint handling strategy that forces compliance with slope limitation and collision avoidance distance through a penalty function method.

[0037] In some embodiments, the state switching of the switching module is configured as follows: initial planning state: configure the well trajectory parameters to the three-dimensional geological model; dynamic optimization state: collaboratively optimize the trajectory based on the thermal-fluid coupling simulation results and perform fracture network penetration optimization.

[0038] In some embodiments, the random forest algorithm outputs a set of trajectory adjustment suggestions for the next drilling segment and updates them synchronously to all associated wellbores. The random forest dynamic correction mechanism includes the following: Input: Real-time logging-while-drilling data (10Hz sampling rate), including resistivity, natural gamma, well diameter, etc. Output: A set of trajectory adjustment suggestions for the next drilling segment (typically 30m), such as: Δ well inclination angle ∈ [-2°, +2°]; Δ azimuth angle ∈ [-5∘, +5∘]; Δ target depth ∈ [-10m, +10m].

[0039] Synchronous Update: If adjustments affect related wells (such as changes in the location of injection wells), the system will automatically update their trajectory parameters and re-simulate.

[0040] This invention also provides a wellbore trajectory collaborative optimization system for hydrothermal geothermal well groups, used to implement the aforementioned method for collaborative optimization of wellbore trajectories for hydrothermal geothermal well groups. The system includes: an input module for inputting a set of well trajectory parameters and geological information including at least seismic data, logging data, a fracture network model, and drilling engineering constraints; a geological modeling module connected to the input module, generating a three-dimensional geothermal reservoir model containing a fracture network based on the geological information; and a machine learning model including a numerical simulation module, an analysis and evaluation module, a collaborative optimization module, an engineering constraint module, a dynamic adjustment module, and a switching module. The numerical simulation module performs heat-fluid coupling simulation, outputting thermal recovery efficiency and interference intensity. The analysis and evaluation module quantifies the evaluation results using a multi-objective function and performs weighted normalization. The collaborative optimization module embeds an improved multi-objective PSO algorithm to batch optimize the trajectory parameter set. The engineering constraint module verifies the build-up rate, anti-collision distance, and fracture penetration conditions, generating a three-dimensional anti-collision report. The dynamic adjustment module inputs real-time logging data and dynamically corrects the trajectory parameters using a random forest algorithm. The switching module seamlessly transitions between the initial planning state and the dynamic optimization state.

[0041] Furthermore, the collaborative optimization module and the dynamic adjustment module adopt a parallel computing architecture to simultaneously process the simulation, optimization and real-time correction of multi-wellbore trajectory combinations.

[0042] Furthermore, the improved PSO algorithm is integrated into the collaborative optimization module, which adjusts the search capability through adaptive inertia weights and uses the penalty function method to handle drilling engineering constraints.

[0043] In the foregoing description, examples have been described with reference to specific exemplary embodiments. However, it will be apparent that various modifications and changes can be made to the specific examples without departing from the scope set forth in the appended claims, and the claims are not limited to the specific examples described above.

Claims

1. A method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters, characterized in that, Includes the following steps: Input a set of well trajectory parameters and geological information including at least seismic data, well logging data, fracture network model, and drilling engineering constraints; A three-dimensional geological model of a thermal reservoir containing a fracture network was constructed, and reservoir permeability and temperature field parameters were extracted. A machine learning model is constructed, and a thermal-fluid coupled numerical simulation model set in the machine learning model is called to calculate the heat recovery efficiency, drilling cost and inter-well interference index of different well trajectory combinations. By setting a multi-objective function in the machine learning model Quantitative analysis and evaluation results To collect heat, For drilling costs, Inter-well interference index. α ,β, γ These are the weighting coefficients; The analysis and evaluation results are dimensionless and multiplied by weights to generate normalized results. Based on the judgment benchmark value or the threshold of qualified items, determine whether trajectory adjustment is needed. If adjustment is needed, select the evaluation item with the largest normalized value first. An improved multi-objective particle swarm optimization algorithm is used to adjust trajectory parameters, and global / local searches are dynamically balanced through adaptive inertial weights. Drilling engineering constraints are satisfied through the penalty function method. The adjusted trajectory parameters are fed back into the thermal-fluid coupling simulation step for iterative optimization until the termination condition is met. During the drilling process, real-time logging data is input, and trajectory parameters are dynamically corrected using a random forest algorithm.

2. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 1, characterized in that, The machine learning model is also configured to: A switching module that switches between the initial planning state and the dynamic optimization state of the well trajectory parameter set; The multiple well trajectories to be modeled are configured in the machine learning model in the initial planning state; By switching the machine learning model from the initial planning state to the dynamic optimization state, the machine learning model is used to perform collaborative constraint optimization on the trajectory parameter set of multiple wells; Based on the results of thermal-fluid coupling numerical simulation, the trajectory parameters of the group wells are dynamically adjusted to maximize heat recovery, minimize inter-well interference, and minimize drilling costs. During the drilling process, the set of trajectory parameters is dynamically adjusted based on real-time logging data using a random forest algorithm.

3. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 1, characterized in that, The drilling engineering constraints include build-up rate limits, collision avoidance distances, and fracture penetration conditions. The optimized trajectory needs to be verified for spatial safety through three-dimensional anti-collision scanning to ensure maximum contact with high-permeability fracture zones.

4. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 1, characterized in that, The normalization result is determined using any of the following methods: (a) Calculate the total normalized value; if it exceeds the baseline value, an adjustment is triggered. (b) Count the number of qualified items that meet the benchmark, and trigger adjustments if the number is lower than the threshold.

5. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 1, characterized in that, The adaptive inertia weight strategy of the improved multi-objective particle swarm optimization algorithm: In the early stages of iteration, the global search capability is increased, and in the later stages, the local search accuracy is enhanced; constraint handling strategy: the penalty function method is used to force the slope rate limit and collision avoidance distance to be met.

6. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 2, characterized in that, The state switching of the switching module is configured as follows: Initial planning status: Configure well trajectory parameters into the 3D geological model; Dynamic optimization state: The trajectory is collaboratively optimized based on the results of thermal-fluid coupling simulation, and crack network penetration optimization is performed.

7. The method for collaborative optimization of wellbore trajectories in hydrothermal geothermal clusters according to claim 1, characterized in that, The random forest algorithm outputs a set of trajectory adjustment suggestions for the next drilling segment and updates them synchronously to all associated wells.

8. A hydrothermal geothermal wellbore trajectory collaborative optimization system, used to implement the hydrothermal geothermal wellbore trajectory collaborative optimization method according to any one of claims 1-7, characterized in that, include: The input module is used to input a set of well trajectory parameters and geological information including at least seismic data, well logging data, fracture network model, and drilling engineering constraints. The geological modeling module is connected to the input module and generates a three-dimensional thermal reservoir model containing a fracture network based on the geological information. The machine learning model includes modules for numerical simulation, analysis and evaluation, collaborative optimization, engineering constraints, dynamic adjustment, and switching. The numerical simulation module is used to perform thermal-fluid coupling simulation and output the heat collection efficiency and interference intensity. The analysis and evaluation module quantifies the evaluation results through a multi-objective function and performs weighted normalization processing. The collaborative optimization module embeds an improved multi-objective particle swarm optimization algorithm to optimize the set of trajectory parameters in batches. The engineering constraint module is used to verify the slope, anti-collision distance and crack penetration conditions, and generate a three-dimensional anti-collision report. The dynamic adjustment module is used to input real-time logging-while-drilling data and dynamically correct trajectory parameters using a random forest algorithm. The switching module is used to seamlessly transition between the initial planning state and the dynamic optimization state.

9. The wellbore trajectory collaborative optimization system for hydrothermal geothermal cluster wells according to claim 8, characterized in that, The collaborative optimization module and the dynamic adjustment module adopt a parallel computing architecture to simultaneously process the simulation, optimization and real-time correction of multiple wellbore trajectory combinations.

10. The wellbore trajectory collaborative optimization system for hydrothermal geothermal cluster wells according to claim 8, characterized in that, The improved multi-objective particle swarm optimization algorithm is integrated into the collaborative optimization module. It adjusts the search capability through adaptive inertia weights and uses the penalty function method to handle drilling engineering constraints.

Images