Moderate and low temperature geothermal drilling and completion methods and drilling optimization system

Abstract

The application discloses a kind of low-temperature geothermal drilling and well completion optimization method and system.Method includes: constructing drilling equipment efficiency model (input energy, output footage) and thermal reservoir response model (input parameter, output capacity index), based on both to establish whole-process optimization model, with drilling cost minimization and thermal reservoir capacity maximization function as target, decision variable contains energy input, well structure and other parameters, iteration is solved by Benders decomposition algorithm and dynamically controlled operation.System contains equipment efficiency modeling, thermal reservoir response modeling, optimization solving engine and dynamic control terminal and other modules, can realize well structure design, drilling process parameter generation, completion sealing scheme decision and other functions, form closed-loop optimization.The method improves drilling efficiency and capacity, reduces cost, and is suitable for low-temperature geothermal resource development.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geothermal drilling and completion, in particular to a method for drilling and completing a medium-low temperature geothermal well and a drilling optimization system. BACKGROUND

[0002] The conventional two-opening wellbore structure (surface casing + technical casing) is prone to cause reservoir pollution in the fractured bedrock type geothermal reservoir due to the long well section. In addition, the cement sealing is insufficient, which causes the water layers to communicate with each other, affecting the normal operation of the geothermal well and the effective development of geothermal resources. The commonly used expansion rubber water stopper is prone to failure due to aging in the actual use process. Once it fails, low-temperature water will seep into the wellbore, not only reducing the water temperature, but also causing the single-well production capacity to decrease, affecting the utilization efficiency of geothermal resources.

[0003] The conventional drilling fluid positive circulation mode is prone to block the geothermal reservoir pores during drilling, hindering the flow of geothermal fluid. Although the clean water circulation has the advantage of environmental protection, its rock carrying capacity is weak, which will cause the cuttings to accumulate at the bottom of the well, thereby affecting the water yield of the geothermal well and reducing the development benefit of the geothermal well. In the recharging process, the mudstone section is prone to collapse, blocking the recharging channel. At the same time, fine silt is also prone to invade the formation, reducing the recharging efficiency and affecting the sustainable development and utilization of geothermal resources. SUMMARY

[0004] The present application aims to provide a method for optimizing drilling and completion of a medium-low temperature geothermal well, comprising the following steps:

[0005] Constructing a drilling equipment efficiency model: establishing an efficiency model for the drilling equipment subsystem, the input of which includes drilling energy input, and the output of which is drilling footage;

[0006] Constructing a geothermal reservoir response model: establishing a response model for the geothermal reservoir, the input of which includes drilling parameter combinations, and the output of which includes geothermal production capacity indicators;

[0007] Establishing a drilling optimization model: based on the efficiency model and the response model, establishing a drilling whole-process optimization model, the decision variables of which include drilling equipment energy input, wellbore structure parameters, and completion control parameters; the optimization target is a function of minimizing drilling cost and maximizing geothermal reservoir production capacity;

[0008] Iterative optimization solution: multiple iterations of the optimization model are solved by the Benders decomposition algorithm, and the drilling operation is dynamically controlled according to the solution result.

[0009] Further, the drilling equipment efficiency model specifically includes:

[0010] Obtaining historical drilling data, including: rig power, mud pump flow, air injection amount, formation leakage rate, and rate of penetration;

[0011] A first neural network model is established to obtain a drilling equipment efficiency model by training historical data, which is used to predict drilling efficiency under different input parameters.

[0012] Further, the building of the thermal reservoir response model specifically includes:

[0013] The historical thermal reservoir response data is obtained, including drilling fluid density, pH value, screen pipe type, gravel packing density, and well flushing parameters.

[0014] A second neural network model is established to obtain a thermal reservoir response model by training historical data, which is used to predict the mapping relationship between completion parameters and geothermal well water production.

[0015] Further, the iterative optimization solution adopts a Benders decomposition algorithm, including:

[0016] Temporary decision variables are generated: wellbore structure parameters and drilling process parameters are generated based on the current model.

[0017] Simulated drilling is performed: simulated instructions are sent to the thermal reservoir response model according to the temporary parameters.

[0018] Feedback data is received: including predicted water production, temperature decay rate and gradient information.

[0019] When the relative difference between the upper and lower bounds of the objective function is less than a threshold value, the final drilling scheme is output.

[0020] Further, the optimization objective function is: min(∑ɸ+ɵ-α×ʀ);

[0021] Wherein, ɸ is the drilling energy cost, ɵ is the completion material cost, ʀ is the predicted water production, and α is the production capacity weight coefficient.

[0022] The constraint conditions include: wellbore structure compatibility, sand control screen pipe strength limit, and underbalanced drilling pressure threshold.

[0023] Further, the wellbore structure design is integrated into the optimization model as a decision variable:

[0024] For sandstone pore type thermal reservoir, the technical casing diameter and screen pipe type of the second opening wellbore structure are determined.

[0025] For bedrock fissure type thermal reservoir, the open hole section length and wire wrapping and netting process parameters of the third opening wellbore structure are determined.

[0026] Further, the composite drilling process parameters are dynamically generated by the optimization model:

[0027] In the lost formation, the air-lift reverse circulation drilling parameters are generated, including the inner diameter of the double-wall drilling tool and the cuttings upward velocity.

[0028] The water-filling air-lifting underbalanced drilling parameters of the easy-leaking layer are generated, including the air injection ratio and the underpressure value.

[0029] Further, the well completion sealing scheme is decided by the optimization model:

[0030] The lithology combination mode of the casing and the sand control screen pipe is selected;

[0031] The thickness of the lead sealing layer and the taper angle of the special alloy hanger are optimized;

[0032] The gravel packing density and the screen pipe wire wrapping layer number are dynamically matched.

[0033] Further, a dynamic logging parameter library is established:

[0034] According to the logging scheme output by the optimization model, the reservoir permeability distribution is fed back in real time;

[0035] Based on the feedback data, the thermal reservoir response model is updated to form a closed-loop optimization.

[0036] A drilling optimization system for implementing the low-temperature geothermal drilling and completion optimization method, comprising:

[0037] An equipment efficiency modeling module for storing a drilling equipment efficiency model and receiving drilling rig and mud pump operation data in real time;

[0038] A thermal reservoir response modeling module for storing a thermal reservoir response model and integrating geological data and historical production capacity data;

[0039] An optimization solving engine for executing a Benders decomposition algorithm and outputting the optimal parameter combination of the wellbore structure, drilling process, and well completion sealing;

[0040] A dynamic control terminal for issuing the optimization parameters to a drilling equipment execution unit and adjusting the air-lift reverse circulation device and the adaptive drilling fluid injection system in real time;

[0041] The optimization solving engine further comprises:

[0042] A sand control process decision sub-module for dynamically selecting a sand control scheme among the wire-wrapped screen pipe, slotted liner, and wire-wrapped net based on the thermal reservoir fracture characteristics;

[0043] A sealing life prediction sub-module for optimizing the rubber-free sealing structure parameters through stress distribution simulation of the alloy hanger sealing layer;

[0044] The dynamic control terminal executes in real time:

[0045] A stepwise well flushing instruction for starting the air-lift reverse circulation well flushing, high-pressure jet flushing, and chemical plug removal agent injection in the optimized sequence;

[0046] Drilling fluid dynamic regulation is used to automatically adjust the drilling fluid performance according to the pH value and the density output by the model.

[0047] The application has the following beneficial effects:

[0048] The application has a double-model construction technology based on a neural network, which improves the prediction accuracy of drilling and thermal reservoir response. The technology trains a first neural network model by obtaining historical drilling data (drilling rig power, mud pump flow, etc.), constructs a drilling equipment efficiency model, and can accurately predict the drilling footage under different energy inputs. Compared with traditional empirical formulas, the neural network can capture nonlinear relationships, such as the coupling effect of formation leakage rate and mechanical drilling speed, avoiding efficiency misjudgment caused by parameter fluctuations. For example, in a sandstone formation, when the mud pump flow increases from 30 L / s to 40 L / s, the traditional model may linearly predict that the drilling speed will increase by 20%, while the neural network model learns the inhibitory effect of leakage rate increase on drilling speed from historical data.

[0049] At the same time, the second neural network model is trained using historical thermal reservoir response data (drilling fluid density, screen pipe type, etc.), and a thermal reservoir response model is established to accurately map the relationship between completion parameters and water production. Taking the gravel packing density as an example, the traditional method is empirically set to 1.8 g / cm 3 , while the model analyzes the permeability change under different densities and finds that when the screen pipe type is twisted wire wrapped screen, a packing density of 2.2 g / cm 3 can increase the water production of the sandstone thermal reservoir by 30%. Field verification shows that the water production of the well section under this parameter is 80 m 3 / h, which is 25% higher than that of the historical well of the same type. This double-model construction technology provides a reliable data basis for subsequent optimization and solves the problems of poor model generalization and large prediction deviation in traditional methods.

[0050] The application has an iterative optimization mechanism driven by Benders decomposition algorithm, which realizes the collaborative optimization of drilling whole-process cost and productivity. The mechanism generates temporary decision variables (wellbore structure, drilling process parameters), simulates the drilling process and receives thermal reservoir feedback data (water production, temperature decay rate, etc.), and outputs the optimal scheme when the relative difference between the upper and lower bounds of the objective function is less than a threshold. Its core advantage is to decompose the complex mixed integer programming problem into equipment efficiency sub-problems and thermal reservoir response sub-problems, reducing the solution dimension. The optimization objective function min(∑ɸ+ɵ-α×ʀ) considers the drilling energy cost, completion material cost, and predicted water production, and the introduction of constraint conditions (wellbore structure compatibility, sand control screen pipe strength, etc.) ensures that the optimization scheme is feasible in engineering practice, such as when underbalanced drilling, the model automatically controls the pressure within a threshold value to avoid the risk of blowout.

[0051] This application possesses a differentiated wellbore structure design technology based on reservoir type, addressing contamination and cross-contamination issues in different reservoirs. For sandstone porous reservoirs, this technology determines the casing diameter and screen type of the two-section wellbore structure. Traditional two-section structures in sandstone often suffer from mismatch between casing diameter and porosity, leading to cement penetration into the pores and reservoir contamination. For bedrock fractured reservoirs, this technology determines the open-hole length and wire mesh wrapping process parameters of the three-section wellbore structure. Excessively long traditional three-section sections easily lead to fracture leakage and water cross-contamination. A certain bedrock well adopted an optimized open-hole section of 200m, coupled with 3 layers of wire mesh wrapping and a gravel packing density of 2.0g / cm³. 3 Compared to the traditional scheme (350m open hole section, 2 layers), the temperature decay rate decreased from 15% / 100m to 8% / 100m, and no cross-contamination occurred, resulting in a 25% increase in single-well productivity. This differentiated design, by incorporating wellbore structural parameters as decision variables into the optimization model, achieves precise matching of "reservoir characteristics-wellbore structure-productivity," fundamentally solving the pollution and cross-contamination problems caused by the traditional uniform structure.

[0052] This application possesses a dynamic generation technology for composite drilling process parameters, adapting to safe and efficient drilling in both lost-circuit and easily lost-circuit formations. In lost-circuit formations, this technology generates air-lift reverse circulation drilling parameters (double-wall drill string inner diameter, cuttings return velocity). The optimization model dynamically generates a parameter combination of a 127mm double-wall drill string inner diameter and a return velocity of 2.5m / s based on the loss rate. Utilizing the negative pressure suction effect of air-lift reverse circulation, cuttings are promptly carried out, with the accumulation thickness controlled within 1cm / h. In easily lost-circuit formations, water-air-filled underbalanced drilling parameters (air injection ratio, underpressure value) are generated. Traditional water circulation has weak cuttings-carrying capacity, while improper underbalanced parameters can easily lead to wellbore collapse. In a certain easily lost-circuit mudstone section, the optimization model recommends an air injection ratio of 30% and an underpressure value of 0.5MPa. Compared to traditional water drilling (without air injection), and with the underpressure value controlled within the mudstone stability limit, this technology solves the efficiency and safety bottlenecks of traditional processes in complex formations by dynamically generating drilling parameters adapted to different formations.

[0053] This application employs a multi-parameter collaborative optimization technology for well completion sealing and sand control schemes, improving seal life and reinjection efficiency. The well completion sealing scheme optimizes the model by determining the lithological combination of casing and screen, the lead sealing layer parameters of the alloy hanger, and the matching of gravel packing density with the number of screen layers. Traditional expanding rubber stoppers are prone to aging and failure. The sand control process decision submodule dynamically selects sand control schemes based on the characteristics of thermal reservoir fractures, employing tiered well washing commands (airlift reverse circulation well washing → high-pressure jet flushing → chemical unblocking) and dynamic drilling fluid control (based on model output pH value 8.5, density 1.1 g / cm³). 3The synergistic effect of regulation significantly improves the reservoir permeability recovery rate after well completion. This technology solves the problems of sealing failure, sand intrusion, and reinjection blockage through multi-parameter synergistic optimization, thus extending the well completion seal life. Attached Figure Description

[0054] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0055] 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.

[0056] Reference Figure 1 This invention aims to provide a method for optimizing medium- and low-temperature geothermal drilling and completion, comprising the following steps: Constructing a drilling equipment efficiency model: establishing an efficiency model for the drilling equipment subsystem, with inputs including drilling energy input and output including drilling footage; Constructing a geothermal reservoir response model: establishing a response model for the geothermal reservoir, with inputs including a combination of drilling parameters and output including geothermal production capacity indicators; Establishing a drilling optimization model: based on the efficiency model and response model, establishing a full-process drilling optimization model, with decision variables including drilling equipment energy input, wellbore structural parameters, and completion control parameters; the optimization objective is a function of minimizing drilling cost and maximizing geothermal reservoir production capacity; Iterative optimization solution: using the Benders decomposition algorithm to iteratively solve the optimization model multiple times, and dynamically controlling drilling operations based on the solution results.

[0057] The theoretical basis and implementation mechanism of the dual-model construction are as follows:

[0058] Drilling Equipment Efficiency Model: Based on the nonlinear mapping capability of neural networks, this model establishes a correlation between energy input parameters such as drilling rig power and mud pump flow rate and drilling footage. The principle is to train the neural network using historical data, enabling it to learn the complex mapping relationship between "energy input - formation response - drilling efficiency." For example, when the formation loss rate increases, the effective utilization rate of the mud pump flow rate decreases. The neural network captures this nonlinear attenuation relationship by adjusting the weights of hidden layer nodes, thus avoiding the prediction bias of traditional linear models. The mathematical essence of this model is multivariate nonlinear regression. It optimizes the weight matrix through an error backpropagation algorithm to minimize the mean square error between the predicted and actual values, making it suitable for handling efficiency prediction problems involving multiple coupled parameters and strong nonlinearity during drilling.

[0059] The geothermal reservoir response model leverages the advantages of LSTM neural networks in processing time-series data to establish a dynamic mapping between parameters such as drilling fluid density and screen type and geothermal productivity indicators (water production). The gating mechanism of LSTM (forget gate, input gate, output gate) effectively memorizes the time-series characteristics of different parameter effects, such as the long-term impact of gravel packing density on permeability. During model training, the model learns the time-series patterns of "parameter application - reservoir response - productivity change" from historical data to predict future productivity. Its theoretical basis is the combination of reservoir flow mechanics and machine learning, fitting parameters that are difficult to quantify using theoretical models such as Darcy's law (e.g., the impact of screen type on fracture flowability) through a data-driven approach, thus solving the prediction error problem caused by excessive assumptions in traditional analytical models.

[0060] The optimization principle and iterative logic of the Benders decomposition algorithm are as follows:

[0061] Basic Algorithm Framework: Benders decomposition is a multi-level optimization algorithm that breaks down complex mixed-integer programming problems into upper-level master problems and lower-level subproblems. In this invention, the master problem is responsible for deciding integer parameters such as wellbore structure and drilling process (e.g., screen type, open-hole length), while the subproblems, given the parameters of the master problem, optimize continuous parameters (e.g., drilling fluid density, sealing layer thickness) and provide feedback on reservoir response data (water production rate, temperature decay rate). This decomposition reduces the dimensionality of high-dimensional optimization problems, lowers computational complexity, and is suitable for optimization scenarios where discrete and continuous parameters coexist throughout the drilling process.

[0062] Iterative convergence mechanism: In each iteration, the main problem generates temporary decision variables, and the subproblems calculate the objective function value and constraints using the thermal storage response model. If the convergence condition is not met (the relative difference between the upper and lower bounds of the objective function ≥ the threshold), a Benders cut (i.e., constraint condition) is generated and returned to the main problem to limit the subsequent search space. For example, when a subproblem finds that the length of a certain unsealed section causes the stress of the sealing layer to exceed the limit, a Benders cut of "unsealed section length ≤ 200m" will be generated, and the main problem will exclude the overly long unsealed section scheme in the next iteration. Through this iterative logic of "generation-verification-feedback-correction", the solution space is gradually narrowed, and finally converges to the global optimum. Its mathematical essence is to guide the search of the main problem through the dual information of the subproblems, ensuring the feasibility and optimality of the optimization result.

[0063] The mathematical modeling principle for the objective function and constraints is as follows:

[0064] The objective function is constructed as min(∑ɸ+ɵ-α×ʀ), which integrates multi-objective optimization ideas of cost and capacity. Drilling energy cost ɸ and completion material cost ɵ are linearly summed to reflect direct economic input; the predicted water output ʀ is converted into a cost deduction item through a capacity weighting coefficient α (0<α<1), reflecting the contribution of capacity to economic benefits. The mathematical meaning of this function is to establish a quantitative balance between cost and capacity. The value of α determines the optimization focus (e.g., when α=0.7, every 10 m³ / h increase in capacity is equivalent to a cost deduction of 70,000 yuan). This linear weighting method transforms multi-objective optimization into single-objective optimization, simplifying the solution process while retaining adaptability to different engineering needs (by adjusting α to prioritize cost control or increase capacity).

[0065] Constraints are established as follows: Wellbore structure compatibility constraints are based on the principle of geometric matching to ensure the mechanical stability of the casing and formation; sand control screen strength limit constraints are based on the mechanics of materials to avoid screen deformation and blockage; underbalanced drilling pressure threshold constraints are based on wellbore stability theory. These constraints transform safety specifications in engineering practice into mathematical expressions, ensuring the engineering feasibility of the optimization scheme and achieving the unity of "theoretical optimization - engineering feasibility".

[0066] The technical logic behind the differentiated design of thermal reservoir types and the dynamic generation of drilling processes is as follows:

[0067] The principle of wellbore structure differentiation: For sandstone porous geothermal reservoirs, "pollution prevention" is the core principle. Based on the porous media seepage theory, reservoir damage is reduced by increasing the diameter of the technical casing (reducing the cement intrusion radius) and optimizing the screen type (such as matching the gap size of the wire-wound screen to the sand particle size). The theoretical formula is: Cement intrusion depth L = K × (P / μ) × t (K is the permeability coefficient, P is the grouting pressure, μ is the cement viscosity, and t is the time). Optimizing the casing diameter can reduce the product of P and t, keeping L within a safe range. For bedrock fractured geothermal reservoirs, "production preservation" is the goal, based on the cubic law of fracture seepage Q = (gb 3 / 12μ)×ΔP (g is the acceleration due to gravity, b is the fracture width, and ΔP is the pressure difference), by optimizing the open hole length (increasing the exposed fracture area) and the wire wrapping parameters (reducing fracture blockage), the flow rate Q is increased. The essence of this differentiated design is to match the corresponding wellbore structure parameters to the seepage mechanism of different reservoirs, thereby maximizing energy transfer efficiency.

[0068] The dynamic generation logic of the composite drilling process is as follows: In lost circulation formations, the negative pressure pumping principle of air-lift reverse circulation is as follows: By injecting high-pressure gas (density ρg) and drilling fluid (density ρm) to form a mixed fluid, the annular fluid column pressure Pc = ρg × h (where h is the well depth) is reduced. When Pc < formation pressure Pf, a negative pressure difference ΔP = Pf - Pc is generated, promoting the return of cuttings. The model dynamically calculates ΔP based on the loss rate, matching the inner diameter d of the double-wall drill string with the return velocity v, so that the cuttings carrying rate η = 0.85 × (v / vt) (where vt is the critical velocity) ≥ 80%. In easily lost circulation formations, water-air-filled underbalanced drilling utilizes the compressibility of air. By controlling the air injection ratio β, the annular fluid density ρa = ρm × (1 - β) + ρg × β is adjusted, so that the underpressure value ΔP = Pf - ρa × h is controlled within the safe range of mudstone collapse pressure. At the same time, the increase of β increases the fluid kinetic energy and enhances the cuttings carrying capacity. The technical principle of this process is to establish "pressure-safe and rock-carrying-efficient" drilling conditions in complex formations by controlling fluid dynamic parameters, thus solving the problem of insufficient adaptability of traditional processes.

[0069] The closed-loop optimization mechanism for well completion sealing and dynamic logging is as follows:

[0070] Multi-parameter synergistic optimization of well completion sealing: Based on elastoplastic mechanics, the stress distribution σ(r,θ) of the lead sealing layer is related to the thickness h and the cone angle θ. A constraint condition of σ(r,θ) ≤ σs (yield strength) is established through finite element simulation to optimize the combination of h and θ. Simultaneously, the lithological combination of casing and screen follows the principle of "elastic modulus matching," such as a modulus difference of ≤10% between steel casing (E=210GPa) and wire-wound screen (E=190GPa) to reduce thermal stress caused by temperature changes. The matching of gravel packing density ρ and the number of screen layers n is based on particle gradation theory, ρ=ρ max ×(1-ε)(ρ max (where ε is the maximum dry density and ε is the porosity), and when n≥3 layers, it can effectively block sand particles with a diameter >0.074mm. The essence of this multi-parameter collaborative optimization is to establish a reliability model of the sealing system through multi-physics field coupling analysis (mechanics, thermodynamics, fluid mechanics), forming a complete technology chain from material selection and structural design to parameter matching, thereby improving the sealing life and sand prevention effect.

[0071] Closed-loop optimization of dynamic logging parameter library: Based on geostatistics, the reservoir permeability distribution k(x,y,z) from logging feedback is used to update the input parameters of the thermal reservoir response model through Kriging interpolation, forming a closed loop of "optimized model output logging scheme → logging data feedback → model parameter correction → new round of optimization". Its mathematical principle is Bayesian updating, which uses new logging data to correct the model's prior probability distribution to obtain the posterior distribution, allowing the model to continuously approximate the actual reservoir characteristics. For example, when logging reveals that the permeability of a certain area is lower than the model's predicted value, the permeability coefficient k of that area is updated, and the model automatically adjusts the well cleaning parameters for that section in the next round of optimization (such as increasing jet pressure) to compensate for the impact of insufficient permeability on production capacity. This closed-loop optimization mechanism breaks the traditional "open-loop design-passive execution" model, achieving adaptive adjustment during the drilling and completion process, and significantly enhancing the system's robustness to geological uncertainties.

[0072] The following are specific examples.

[0073] A medium-low temperature geothermal field is located in the Guanzhong Plain. The target reservoir is a fractured limestone reservoir of the Lower Paleozoic Cambrian period, buried at a depth of 1800-2200m, with a formation temperature of 60-80℃. The expected design production capacity of a single well is 50m³. 3 / h. Preliminary exploration revealed complex geological conditions in the area, including uneven development of bedrock fissures, localized lost circulation zones (loss rate 15%-30%), and mudstone interlayers. Traditional drilling methods encountered the following problems during adjacent well operations: second-stage wellbore structure led to reservoir contamination, resulting in a 25% reduction in water production compared to predictions; rubber seal failure caused a 10-15℃ drop in water temperature; and positive circulation drilling achieved a mechanical drilling rate of only 1.5 m / h in lost circulation zones, with a construction period as long as 45 days.

[0074] Construction of drilling equipment efficiency model:

[0075] Data Acquisition: Drilling data from 10 historical wells in the area were collected, including drilling rig power (400-500kW), mud pump flow rate (25-40L / s), and air injection rate (0-50m³ / s). 3 The drilling speed is calculated as follows: drilling rig power ( / min), formation loss rate (5%-40%), and mechanical drilling rate (0.8-3.0 m / h). For example, in a certain well, with a drilling rig power of 450 kW, mud pump flow rate of 35 L / s, and loss rate of 20%, the mechanical drilling rate is 2.1 m / h.

[0076] Model training: A three-layer backpropagation (BP) neural network (5 nodes in the input layer, 10 nodes in the hidden layer, and 1 node in the output layer) was used, with 70% of the data as the training set and 30% as the test set. During training, the learning rate was set to 0.01, the number of iterations was 1000, and the loss function was mean squared error. After training, the model's average error in predicting the mechanical drilling rate on the test set was 4.8%. For example, if the actual drilling rate of a test sample was 2.5 m / h, the model predicted 2.62 m / h, with an error of 5.2%, which meets the engineering accuracy requirements.

[0077] Construction of thermal reservoir response model:

[0078] Data integration: Compile historical well completion parameters and productivity data, including drilling fluid density (1.05-1.2 g / cm³). 3 pH value (7.0-9.0), sieve tube type (wire-wound sieve tube, wire-wrapped mesh), gravel packing density (1.8-2.4 g / cm³). 3 Well washing parameters (gas lift time, jet pressure) and water output (30-70m³) 3 / h). For example, a well uses wire-wound screen pipe with a packing density of 2.2 g / cm³. 3 When the well-washing jet pressure is 20MPa, the water output reaches 65m³. 3 / h.

[0079] Model Establishment: An LSTM neural network (5 nodes in the input layer, 50 nodes in the hidden layer, and 1 node in the output layer) was used to train the model using historical data to capture the temporal relationship between parameters and water output. Testing showed that the model's average prediction error for water output was 6.3%, and the actual water output of a certain validation sample was 55 m³ / s. 3 / h, predicted value 58.2m 3 / h, with an error of 5.8%, accurately reflects the impact of well completion parameters on production capacity.

[0080] Drilling optimization model establishment and solution:

[0081] Decision variables are set as follows: Wellbore structural parameters are the length of the three-section open hole (180-250m) and the number of wire mesh layers (2-4 layers); Drilling process parameters are the inner diameter of the gas lift reverse circulation double-wall drill string (108-139mm) and the cuttings return velocity (2.0-3.0m / s); Completion control parameters are the lead sealing layer thickness (10-20mm) and the cone angle (25°-35°).

[0082] Objective function and constraints: The objective function is min(∑ɸ+ɵ-0.7×ʀ), where ɸ is the drilling energy cost (unit price 0.8 yuan / kW·h), ɵ is the completion material cost (screen pipe, hanger, etc.), and ʀ is the predicted water production (m³ / h). 3 / h). Constraints include: length of the naked eye section ≤ 250m, stress of the sealing layer ≤ yield strength of lead (18MPa), and underbalanced pressure ≥ 0.3MPa and ≤ 0.8MPa.

[0083] Benders decomposition iteration: The first iteration generates temporary parameters: 220m open hole section, 3 mesh layers, drill string inner diameter 127mm, up-and-down velocity 2.5m / s, sealing thickness 15mm, and cone angle 30°. After inputting into the reservoir model, the predicted water yield is 68m³. 3 The energy cost is 280,000 yuan, the material cost is 150,000 yuan, and the objective function value is 280,000 + 150,000 - 0.7 × 68 = 15.4. After 5 iterations, the relative difference between the upper and lower bounds decreased from 12% to 2.3% (< the threshold of 5%), converging to the optimal solution: 200m open hole section, 4 mesh layers, 127mm drill string inner diameter, 2.8m / s return velocity, 16mm sealing thickness, and 32° cone angle. At this point, the water output is 72m³ / h. 3 / h, energy cost 250,000 yuan, material cost 180,000 yuan, objective function value = 250,000 + 180,000 - 0.7 × 72 = 12.6, which is 20.8% better than the initial plan.

[0084] Dynamic control and on-site implementation:

[0085] Drilling process execution: In the lost circulation formation (loss rate 25%), air lift reverse circulation was initiated. The inner diameter of the double-wall drill string was adjusted to 127mm according to the model parameters, and the cuttings return velocity was controlled at 2.8m / s. The actual mechanical drilling rate reached 3.2m / h, which is 113% higher than the traditional forward circulation, and no cuttings accumulation occurred. In the mudstone interlayer section, water-air-filled underbalanced drilling was adopted, with an air injection ratio of 28% and an underpressure value of 0.6MPa. The wellbore was stable, and the cuttings carrying efficiency was improved by 55%. The construction cycle for this section was shortened from the planned 10 days to 6 days.

[0086] Well completion sealing and sand control: A lithological combination of steel casing and 4 layers of wire-wound screen pipe was selected. The lead sealing layer thickness was optimized to 16mm and the cone angle to 32°. Stress simulation showed that the maximum stress of the sealing layer was 16.5MPa < 18MPa, meeting the strength requirements. The gravel packing density was 2.3g / cm³ as output by the model. 3 The implementation demonstrates significant sand control effectiveness, with sand intrusion rate <0.8 kg / m³. 3 The cascade well washing was performed in the following sequence: "gas lift reverse circulation (4 hours) → high pressure jet (pressure 25MPa, 2 hours) → chemical unblocking (pH 8.8 unblocking agent, 1 hour)". After well washing, the reservoir permeability was restored to 96%.

[0087] Dynamic logging and model updates: Real-time logging feedback on reservoir permeability distribution revealed that the permeability in the 2050-2100m section was 15% lower than the model prediction. The thermal reservoir response model parameters were immediately updated, and the well-washing jet pressure in this section was adjusted to 28MPa, ultimately resulting in a water production of 70m³. 3 / h, exceeding the designed capacity by 40%, with an outlet water temperature of 78℃, 13℃ higher than the adjacent well, and no cross-contamination.

[0088] Implementation Results and Comparison: The total construction period for this well was 32 days, 13 days shorter than the traditional method. Drilling costs were reduced by 18% (energy savings of 50,000 yuan and material optimization savings of 30,000 yuan), single-well productivity increased by 40%, and the predicted sealing life exceeded 20 years. Compared with adjacent wells, the traditional well's water production was 50m³. 3 The well operated at 65°C per hour, but after one year, the temperature dropped to 55°C due to rubber seal failure. In this example, after three years of operation, the well's water output stabilized at 68 m³ / h. 3 The method was validated by achieving a recharge rate of 82% at a temperature of 76℃ and a recharge rate of 1000 m / h.

[0089] 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 optimizing medium- and low-temperature geothermal drilling and completion, characterized in that, Includes the following steps: Constructing a drilling equipment efficiency model: Establishing an efficiency model for the drilling equipment subsystem, with inputs including drilling energy input and outputs including drilling footage; Constructing a geothermal reservoir response model: Establishing a response model for the geothermal reservoir, with inputs including a combination of drilling parameters and outputs including geothermal production capacity indicators; Establish a drilling optimization model: Based on the efficiency model and response model, establish a full-process drilling optimization model. The decision variables include drilling equipment energy input, wellbore structural parameters, and completion control parameters. The optimization objective is a function of minimizing drilling cost and maximizing thermal reservoir production capacity. Iterative optimization solution: The optimization model is solved through multiple iterations using the Benders decomposition algorithm, and the drilling operation is dynamically controlled based on the solution results; Constructing a drilling equipment efficiency model specifically includes: Obtain historical drilling data, including: drilling rig power, mud pump flow rate, air injection volume, formation loss rate, and mechanical drilling speed; A first neural network model is established, and a drilling equipment efficiency model is obtained by training it with historical data, which is used to predict drilling efficiency under different input parameters. The specific steps involved in constructing a thermal reservoir response model include: Acquire historical thermal reservoir response data, including: drilling fluid density, pH value, screen type, gravel packing density, and well washing parameters; A second neural network model was established and trained using historical data to obtain a geothermal reservoir response model, which was used to predict the mapping relationship between well completion parameters and geothermal well water production. The iterative optimization solution uses the Benders decomposition algorithm, which includes: Generate temporary decision variables: Generate wellbore structure parameters and drilling process parameters based on the current model; Perform simulated drilling: Send simulation commands to the thermal reservoir response model based on temporary parameters; Receive feedback data, including predicted water output, temperature decay rate, and gradient information; When the relative difference between the upper and lower bounds of the objective function is less than the threshold, the final drilling plan is output. This also includes establishing a dynamic logging parameter database: Based on the logging scheme output by the optimized model, the reservoir permeability distribution is fed back in real time. The thermal reservoir response model is updated based on feedback data to form a closed-loop optimization.

2. The method for optimizing medium- and low-temperature geothermal drilling and completion according to claim 1, characterized in that, The optimized objective function is: ; in, For drilling energy costs, For well completion material costs, To predict water output This is the production capacity weighting coefficient; The constraints include: wellbore structural compatibility, sand control screen strength limit, and underbalanced drilling pressure threshold.

3. The method for optimizing medium- and low-temperature geothermal drilling and completion according to claim 1, characterized in that, Wellbore structural design is incorporated as a decision variable into the optimization model: For sandstone porous geothermal reservoirs, determine the casing diameter and screen type for the second wellbore structure; For bedrock fractured geothermal reservoirs, determine the length of the open hole section and the wire wrapping process parameters for the three-section wellbore structure.

4. The method for optimizing medium- and low-temperature geothermal drilling and completion according to claim 1, characterized in that, Composite drilling process parameters are dynamically generated through an optimization model: Generate air-lift reverse circulation drilling parameters in the lost formation, including the inner diameter of the double-wall drill string and the cuttings return velocity; In the easily lost circulation zone, generate water-air-filled underbalanced drilling parameters, including the air injection ratio and underpressure value.

5. The method for optimizing medium- and low-temperature geothermal drilling and completion according to claim 4, characterized in that, Well completion sealing schemes are determined through optimization model decisions: Select the lithological combination method of casing and sand control screen; Optimize the lead sealing layer thickness and cone angle of the specially designed alloy suspension; Dynamically match gravel filling density with the number of wire wrapping layers on the screen tube.

6. A drilling optimization system for implementing the medium-low temperature geothermal drilling and completion optimization method according to any one of claims 1-5, characterized in that, include: The equipment efficiency modeling module is used to store drilling equipment efficiency models and receive drilling rig and mud pump operation data in real time. The thermal reservoir response modeling module is used to store thermal reservoir response models and integrate geological data with historical production data. The optimization solution engine is used to execute the Benders decomposition algorithm and output the optimal combination of parameters for well structure, drilling process, and completion sealing. The dynamic control terminal is used to send optimized parameters to the drilling equipment execution unit and adjust the gas lift reverse circulation device and the adaptive drilling fluid injection system in real time. The optimization solution engine also includes: The sand control process decision submodule is used to dynamically select sand control schemes from wire-wound screens, slotted liners, and wire-wound meshes based on the characteristics of thermal reservoir fractures. The seal life prediction submodule is used to optimize the parameters of rubber-free seal structures by simulating the stress distribution of the alloy hanger seal layer. The dynamic control terminal executes in real time: The cascade well-washing command is used to initiate gas lift reverse circulation well-washing, high-pressure jet flushing, and chemical unblocking agent injection in an optimized sequence; Dynamic control of drilling fluid is used to automatically adjust the performance of drilling fluid according to the pH value and density output by the model.

Images