Fracturing effect dynamic evaluation and regulation and control system based on artificial intelligence

Abstract

The invention discloses a fracturing effect dynamic evaluation and regulation system based on artificial intelligence, and relates to the technical field of tight reservoir fracturing reformation, the system comprises a basic module for collecting construction related data and constructing a prediction model, and the system is based on a ternary coupling mechanism. Comprising an error and physical field coupling acquisition unit, a cross-scale physical field analysis unit, an error and process and topology mapping unit, a chaos adaptive regulation and control unit and a reservoir repair feedback unit. According to the method, a five-unit closed-loop cooperative technology of error and physical field coupling acquisition, cross-scale physical field analysis, error-process-topology three-layer mapping, chaos adaptive regulation and control and reservoir repair feedback under a ternary coupling mechanism is utilized; the problems that an existing AI fracturing evaluation regulation and control system does not excavate reservoir dynamic information in errors, lacks a multi-scale error decoupling mechanism and depends on empirical parameters, so that evaluation deviation is large, error causes are difficult to position, and the fracturing effect is not matched with reservoir requirements are solved.

Description

Technical Field

[0001] This invention relates to the field of tight reservoir fracturing technology, specifically to an artificial intelligence-based dynamic evaluation and control system for fracturing effects. Background Technology

[0002] As oil and gas resource development extends to tight and low-permeability reservoirs, fracturing has become a core technology for enhancing oil recovery. Artificial intelligence-based dynamic evaluation and control systems for fracturing effects are gradually replacing traditional manual control methods due to their real-time response advantages. Existing AI systems primarily collect real-time data such as fracturing pressure, displacement, and sand ratio, combine this data with static reservoir parameters to construct predictive models, achieve dynamic evaluation of fracturing effects, and adjust fracturing parameters based on the evaluation results.

[0003] However, existing technologies have several problems: First, the prediction error of AI assessment models is always considered a confounding factor. Technicians attempt to reduce the error by optimizing algorithms and expanding samples, but fail to uncover the dynamic feedback information of the reservoir contained in the error, leading to a deviation between the assessment results and the actual state of the reservoir. Second, the system lacks a multi-scale error decoupling mechanism, making it impossible to distinguish the different causes of well-section-level macroscopic errors and pore-level microscopic errors, and making it difficult to specifically correlate core parameters of fracturing operations. Third, the existing control logic only revolves around matching the prediction results, without establishing a correlation between error characteristics and proppant placement and fracturing fluid flow patterns. The uniformity of proppant placement, as a key indicator determining fracture conductivity, still relies on empirical parameters for control, often resulting in wasted dosage or uneven placement. At the same time, the complex characteristics of the error, such as the fractal dimension, are not utilized, failing to provide a basis for the dynamic switching of proppant cluster-dispersion states, making it difficult to accurately match the fracturing effect with reservoir requirements, thus restricting the efficiency and economy of tight reservoir fracturing stimulation.

[0004] Therefore, an artificial intelligence-based dynamic evaluation and control system for fracturing effects is provided to overcome the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based dynamic evaluation and control system for fracturing effects, so as to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, this invention provides an artificial intelligence-based dynamic evaluation and control system for fracturing effects, comprising a basic module for collecting construction-related data and constructing a predictive model.

[0007] The system is based on a three-element coupling mechanism, including an error and physical field coupled acquisition unit, a cross-scale physical field analysis unit, an error, process and topology mapping unit, a chaotic adaptive control unit, and a reservoir repair feedback unit. The five units work together to form a closed loop of sensing, analysis, mapping, control and repair.

[0008] The error and physical field coupling acquisition unit synchronously acquires error sequences and reservoir temperature field, stress field, and seepage field data, and extracts coupling features; the cross-scale physical field analysis unit analyzes the bidirectional transmission effect between the well section-level macroscopic physical field and the pore-level microscopic physical field; the error, process, and topology mapping unit establishes a three-layer mapping of error coupling features, construction process status, and fracture network topology; the chaotic adaptive control unit achieves adaptive matching control based on error chaotic features and cross-scale transmission coefficients; and the reservoir repair feedback unit simultaneously completes reservoir microfracture plugging repair and clay expansion inhibition, and provides feedback optimization.

[0009] Furthermore, the error and physical field coupled acquisition unit includes a multi-parameter synchronous acquisition module and a coupled feature extraction module. The multi-parameter synchronous acquisition module acquires the error sequence of AI model prediction of pressure, displacement, sand ratio and measured values ​​through a fiber optic sensor with a sampling frequency of 1kHz, and acquires reservoir temperature field, stress field and seepage field data through distributed fiber optic temperature measurement, microseismic monitoring and downhole flow sensor respectively. The coupled feature extraction module calculates the correlation between the error sequence and each physical field parameter through the mutual information entropy algorithm, extracts 15 coupled features including 6 correlation parameters, the maximum Lyapunov exponent of the error sequence, and the physical field gradient value, and forms a [0,1] standardized dataset.

[0010] Furthermore, the cross-scale physical field analytical unit defines the macroscopic physical field with a unit of 50 meters and the microscopic physical field based on the dynamic digital twin model of reservoir core CT scan; the macroscopic-to-microscopic transmission coefficient K1 and the microscopic-to-macroscopic transmission coefficient K2 are calculated, where K1 = Δr mi / ΔG σ , Δr mi ΔG represents the change in the radius of the micropore throat. σ This represents the change in macroscopic stress gradient; K2 = Δv m / Δα mi Δv m Let Δα be the change in flow rate in the macroscopic seepage field. mi The change in the microscopic clay swelling rate is represented by K1 and K2. The root cause of the positioning error is analyzed by correlation analysis between K1, K2 and the coupled feature set.

[0011] Furthermore, in the three-layer mapping model of the error, process, and topology mapping unit, the input layer is a set of 15 error and physical field coupled features, the middle layer outputs the proppant placement uniformity, fracturing fluid turbulence intensity, and proppant clustering coefficient, and the output layer outputs the fracture density, fracture connectivity, and main fracture length. A dual discriminator collaborative training is adopted, with the topology parameters measured by the drilling imaging tool and the process status measured by the fiber optic sensor as labels, respectively. The topology parameter thresholds are dynamically corrected according to the cross-scale transmission coefficients K1 and K2.

[0012] Furthermore, the macroscopic parameter control of the chaotic adaptive control unit is calculated using the displacement adjustment ΔQ = 0.5 × λ × (75% - η), where λ is the maximum Lyapunov exponent of the error sequence and η is the fracture connectivity; the microscopic parameter control is achieved using the clay inhibitor concentration adjustment ΔC = 0.3 × K² × (α) mi The calculation is based on -15%), where ΔC is the clay mineral expansion rate. The proppant particle size is adjusted according to the pore throat radius. Topology optimization control is calculated using the pulse frequency f = λ × 1.5 Hz, and the sand addition direction is adjusted simultaneously through microseismic monitoring to maintain the crack density at 2-3 cracks per meter.

[0013] Furthermore, the reservoir repair feedback unit uses ±0.5 MPa fluid pressure pulses to impact the proppant powder blocking the microfractures. Microfracture repair is considered complete when the mutual information entropy I(ΔQ,v) between the discharge error and the seepage field increases from <0.3 to 0.3-0.6. Clay swelling is suppressed by adjusting the fracturing fluid ratio; when the clay swelling rate α... mi When the concentration of the low-concentration inhibitor decreases from >15% to ≤12%, the low-concentration inhibitor maintenance mode is activated, and the repaired physical field parameters are fed back to the error and physical field coupling acquisition unit to update the coupling feature set.

[0014] Furthermore, the system's workflow includes a coupled acquisition stage, a cross-scale analysis stage, a three-layer mapping stage, a chaos control stage, and a repair feedback stage. The coupled acquisition stage extracts 15 coupled features. The cross-scale analysis stage calculates K1 and K2 and locates the root cause of the error. The three-layer mapping stage generates the process state and topology parameters of the adversarial network output through dual discriminators. The chaos control stage initiates adaptive control. The repair feedback stage monitors the repair effect and iteratively optimizes.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] Improve assessment accuracy: By coupling error with physical field acquisition and extracting 15 coupled features, the error is transformed into a dynamic sensing signal of the reservoir, which significantly reduces the deviation between the assessment results and the actual state of the reservoir.

[0017] Precisely pinpoint the root cause of errors: By using the bidirectional transmission coefficient (macro to micro, micro to macro) of the cross-scale physical field analytical unit, precise correlation analysis between macro and micro errors can be achieved, greatly improving the efficiency of error cause location.

[0018] Optimize construction parameter control: By using a three-layer mapping model (error-process-topology) and dual discriminators for collaborative training, combined with chaotic adaptive control logic, the uniformity of proppant placement and the stability of fracturing fluid flow can be improved, completely eliminating the dependence on empirical parameters.

[0019] Improving fracture network quality: By topology optimization and control (pulse propulsion + micro-seismic guidance), the fracture density is precisely maintained at 2-3 fractures / meter, improving fracture connectivity, expanding the effective drainage area of ​​tight reservoirs, and thus improving oil recovery.

[0020] Protecting reservoir performance: The reservoir repair feedback unit simultaneously completes microfracture plugging repair (fluid pressure pulse impact) and clay expansion inhibition (fracturing fluid ratio adjustment), effectively improving the reservoir permeability recovery rate.

[0021] Reduce construction costs: Based on the adaptive control of error chaos characteristics, the amount of fracturing fluid and inhibitor used is reduced, construction energy consumption and tool wear are reduced, and tool service life is extended.

[0022] Provides optimization basis: The macroscopic stress and microscopic porosity transmission model established through cross-scale physical field analysis provides new geological parameter support for subsequent fracturing scheme optimization, shortening the scheme optimization cycle. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the artificial intelligence-based dynamic evaluation and control system for fracturing effect of the present invention. Detailed Implementation

[0024] 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.

[0025] Please see Figure 1 The present invention provides a technical solution:

[0026] See Figure 1 As shown, an example of an AI-based dynamic evaluation and control system for fracturing performance is presented:

[0027] I. System:

[0028] This system is based on a three-element coupling mechanism, including an error and physical field coupled acquisition unit, a cross-scale physical field analysis unit, an error, process, and topology mapping unit, a chaotic adaptive control unit, and a reservoir repair feedback unit. These five units work together to form a closed loop of sensing, analysis, mapping, control, and repair, as detailed below:

[0029] (I) Error and Physical Field Coupled Acquisition Unit:

[0030] Breaking away from the traditional model of only collecting error data, this method simultaneously collects error sequences along with reservoir temperature field, stress field, and seepage field data, constructing a coupled dataset of error and physical field.

[0031] Multi-parameter synchronous acquisition module:

[0032] Error sequence acquisition: The errors ΔP(t), ΔQ(t), and ΔS(t) between the AI ​​model's predicted pressure, displacement, and sand ratio and the measured values ​​are acquired through fiber optic sensors (sampling frequency 1kHz); where ΔP(t) is the pressure error at time t; ΔQ(t) is the displacement error at time t; and ΔS(t) is the sand ratio error at time t, expressed as a percentage.

[0033] Physical field acquisition: Distributed fiber optic thermometry is used to acquire the reservoir temperature field T(x,y,z,t), microseismic monitoring is used to inversely calculate the stress field σ(x,y,z,t), and downhole flow sensors are used to acquire the seepage field v(x,y,z,t); where T(x,y,z,t) is the temperature at coordinate (x,y,z,t) at time t; σ(x,y,z,t) is the stress at coordinate (x,y,z,t) at time t; and v(x,y,z,t) is the seepage velocity at coordinate (x,y,z,t) at time t.

[0034] Coupled feature extraction module:

[0035] Algorithm for the correlation between design error and physical field: Calculate the mutual information entropy (reflecting the correlation strength) between the error sequence and each physical field parameter. For example, I(ΔQ,σ) represents the correlation between pressure error and temperature field, and I(ΔQ,σ) represents the correlation between displacement error and stress field. The formula for calculating mutual information entropy is:

[0036]

[0037] in:

[0038] X is the error sequence, Y is the physical field parameter sequence, p(x,y) is the joint probability distribution of X and Y, p(x) is the marginal probability distribution of X, and p(y) is the marginal probability distribution of Y.

[0039] Extracting the coupling feature set: including mutual information entropy I1 to I6 (6 physical field and error correlation parameters), chaotic eigenvalues ​​λ (maximum Lyapunov exponent) of the error sequence, and physical field gradient values ​​G (such as temperature gradient G). T Stress gradient G σ A total of 15 coupled features are used to form a [0,1] standardized dataset; where λ is the maximum Lyapunov exponent of the error sequence, dimensionless, reflecting the chaotic characteristics of the error sequence; G T For temperature gradient; G σ The stress gradient is given.

[0040] (II) Analytical Unit of Cross-Scale Physics Fields:

[0041] Breaking through the separation of macroscopic and microscopic modes, this study analyzes the bidirectional transmission effect of macroscopic physical fields at the well section level and microscopic physical fields at the pore level, revealing the fundamental cause of positioning errors.

[0042] Definition of cross-scale physical fields:

[0043] Macroscopic physical field (well section level): Calculate the mean stress field σ within a 50-meter unit. m Temperature field fluctuation amplitude ΔT m Flow rate v in seepage field m The associated macroscopic parameters are (rock layer thickness h, geostress difference Δσ).

[0044] Microscopic physical field (pore level): Calculate the pore throat radius r based on a dynamic digital twin model (real-time updated pore structure) constructed from reservoir core CT scans. mi Clay mineral swelling rate α mi Microcrack opening w mi Correlation of microscopic parameters (clay content C, porosity) ).

[0045] Calculation of bidirectional transmission coefficient:

[0046] Macroscopic to microscopic transmission coefficient K1: used to calculate the macroscopic stress gradient G σ The degree of influence on the radius of the micropore throat:

[0047] K1=Δr mi / ΔG σ ;

[0048] Where K1 is in micrometers per unit, Δr mi ΔG represents the change in the radius of the micropore throat. σ K1 represents the change in macroscopic stress gradient; K1 reflects the evolution of microscopic porosity induced by macroscopic stress changes.

[0049] Microscopic to macroscopic transmission coefficient K2: Calculation of microscopic clay swelling rate α mi For macroscopic seepage field flow rate v m Degree of impact:

[0050] K2=Δv m / Δα mi ;

[0051] Where K2 is in cubic meters per cubic meter, Δv m Δα represents the change in flow rate in the macroscopic seepage field. mi K2 represents the change in the microscopic clay swelling rate; K2 reflects the change in macroscopic seepage capacity caused by microscopic clay swelling.

[0052] Error cause location: Through correlation analysis of K1, K2 and the coupling feature set, the root cause of the error was located (e.g., K1>0.8 and I(ΔP,σ)>0.7, indicating that the change in macroscopic stress field is the main cause of pressure error).

[0053] K2>0.6 and I(ΔQ,v)>0.6, indicating that microscopic clay swelling is the main cause of displacement error.

[0054] (III) Error and Process Mapping Unit:

[0055] Breaking away from the traditional model of only mapping construction process parameters, a three-layer mapping is established, which includes error coupling characteristics, construction process status, and crack network topology, to achieve multi-dimensional status perception.

[0056] Three-layer mapping model:

[0057] Input layer: Error and physical field coupling feature set (15 parameters);

[0058] Intermediate layer (process status output): Generator G1 outputs proppant placement uniformity R, fracturing fluid turbulence intensity vt, and proppant clustering coefficient K.

[0059] Output layer (topology output): Generator G2 outputs the fracture network topology parameters (fracture density ρ, fracture connectivity η, main fracture length L) based on the intermediate layer results.

[0060] Dual discriminators: D1 uses the measured topological parameters of the drilling imaging tool (acquiring one set of crack images every 3 minutes) as labels, and D2 uses the measured process status of the fiber optic sensor as labels. The dual discriminators are trained together to reduce mapping errors.

[0061] Dynamic threshold correction: The topology parameter threshold is corrected according to the cross-scale transmission coefficients K1 and K2. For example, when K1>0.8 (macro-stress dominant), the qualified threshold of crack connectivity η is reduced from 80% to 75% to adapt to the stress-induced crack propagation characteristics.

[0062] (iv) Chaotic Adaptive Control Unit:

[0063] Breaking away from the traditional model of fixed formula control, this method achieves adaptive matching of construction parameters, physical fields, and topological structures based on the chaotic characteristics of errors and cross-scale transmission coefficients, thus simultaneously solving all root cause problems of errors.

[0064] Chaotic regulation decision-making model:

[0065] Macroscopic parameter control (for K1-dominant error): When K1 > 0.8 and η < 75% (insufficient crack connectivity), initiate stress-adaptive displacement adjustment. Displacement adjustment amount:

[0066] ΔQ = 0.5 × λ × (75% - η);

[0067] Where ΔQ is the displacement adjustment amount; λ is the chaotic characteristic value of the error sequence (maximum Lyapunov exponent); η is the crack connectivity; this formula enables the displacement change to resonate with the stress field fluctuation, promoting crack propagation.

[0068] Micro-parameter tuning (for K2-dominant error): When K2 > 0.6 and α mi >15% (severe clay swelling), initiate dynamic mixing of clay-inhibiting fracturing fluid, adjust the concentration of clay inhibitor in the fracturing fluid as follows:

[0069] ΔC=0.3×K2×(α mi -15%);

[0070] Where ΔC is the adjustment amount of clay inhibitor concentration; K2 is the microscopic to macroscopic transmission coefficient; α mi The swelling rate of clay minerals is adjusted; simultaneously, the proppant particle size (r) is adjusted. mi Use 100 / 140 mesh for sizes <5 micrometers; 5 micrometers

[0071] ≤r mi Use 70 / 140 mesh for particles smaller than 10 micrometers to solve micropore blockage.

[0072] Topology optimization control (new feature): When ρ < 2 cracks per meter (low crack density), activate the pulsed sand injection and microseismic guidance coordinated mode, pulse frequency:

[0073] f = λ × 1.5 Hz;

[0074] Where f is the pulse frequency; λ is the chaotic characteristic value of the error sequence (maximum Lyapunov exponent); at the same time, the direction of sand addition is adjusted in real time through microseismic monitoring, so that the crack density is increased to 2-3 cracks per meter.

[0075] The control objective is to maintain the coupling characteristics within the optimal range (I1 to I6 ∈ [0.3, 0.6], λ ∈ [0.2, 0.4], K1 ∈ [0.4, 0.7], K2 ∈ [0.3, 0.5]), rather than pursuing the minimization of error values.

[0076] (V) Reservoir Repair Feedback Unit:

[0077] By utilizing changes in the physical field during the regulation process, reservoir micro-fracture plugging and repair and clay expansion inhibition can be achieved, forming a synergistic effect of regulation and repair.

[0078] Microcrack repair: When the discharge rate fluctuates, a fluid pressure pulse (ΔP = ±0.5 MPa) is used to impact the proppant powder blocking the microcracks. The change of I(ΔQ,v) is monitored by the seepage field. When I(ΔQ,v) increases from <0.3 to 0.3-0.6, the microcrack repair is considered complete. Here, ΔP is the amplitude of the fluid pressure pulse in MPa; I(ΔQ,v) is the mutual information entropy between the discharge rate error and the seepage field.

[0079] Clay swelling inhibition: The clay swelling rate α is monitored in real time by adjusting the fracturing fluid ratio. mi When α mi When the concentration of inhibitor decreases from >15% to ≤12%, the low-concentration inhibitor maintenance mode is activated to avoid reservoir damage caused by excessive use of inhibitors.

[0080] Feedback on repair results: The repaired physical field parameters (such as v) will be provided. m Increase, α mi The reduced amount is fed back to the error and physical field coupling acquisition unit to update the coupling feature set and optimize subsequent control.

[0081] II. Workflow:

[0082] Coupled acquisition phase: After the system starts up, it synchronously acquires error sequences and reservoir temperature, stress, and seepage field data, and extracts 15 coupled features;

[0083] Cross-scale analysis stage: Calculate the bidirectional transmission coefficients K1 and K2 of macroscopic and microscopic physical fields, and locate the root cause of the error (macroscopic stress or microscopic clay expansion);

[0084] Three-layer mapping stage: The dual discriminator generative adversarial network model outputs the construction process status (proppant, fracturing fluid) and fracture network topology parameters;

[0085] Chaotic control phase: Based on K1, K2 and topological parameters, adaptive control (displacement rate, fracturing fluid ratio, pulse sand addition) is initiated;

[0086] Repair feedback phase: Monitor reservoir repair effects (micro-fracture unblocking, clay suppression), update coupling feature set, and iteratively optimize.

[0087] Summarize:

[0088] By coupling the error with the physical field, the error is transformed into a dynamic sensing signal of the reservoir, which significantly reduces the deviation between the assessment results and the actual state of the reservoir.

[0089] By using bidirectional transmission coefficients K1 and K2, the root cause of error can be accurately located, improving the efficiency of macroscopic and microscopic error correlation analysis.

[0090] By employing three-layer mapping and chaotic regulation, the uniformity of proppant placement and the stability of fracturing fluid flow are improved, completely resolving the problem of dependence on empirical parameters.

[0091] and:

[0092] Fracture network topology optimization: Enables precise control of fracture density and connectivity during fracturing, expands the effective drainage area of ​​tight reservoirs, and improves oil recovery;

[0093] Dynamic reservoir repair: Simultaneously complete micro-fracture plugging repair and clay expansion inhibition during fracturing operations to improve reservoir permeability recovery rate;

[0094] Reduced construction energy consumption: Based on the adaptive control of error chaos characteristics, the amount of fracturing fluid used is reduced, thus reducing construction energy consumption and tool wear, and extending tool life;

[0095] By analyzing physical fields across scales, a model for the transmission of macroscopic stress and microscopic pores is established, providing new geological parameters for the optimization of subsequent fracturing schemes and shortening the optimization cycle.

Claims

1. A dynamic evaluation and control system for fracturing effects based on artificial intelligence, comprising a basic module for collecting construction-related data and constructing a predictive model, characterized in that: The system is based on a three-element coupling mechanism, including an error and physical field coupled acquisition unit, a cross-scale physical field analysis unit, an error, process and topology mapping unit, a chaotic adaptive control unit, and a reservoir repair feedback unit. The five units work together to form a closed loop of sensing, analysis, mapping, control and repair. The error and physical field coupling acquisition unit synchronously acquires error sequences and reservoir temperature field, stress field, and seepage field data, and extracts coupling features; the cross-scale physical field analysis unit analyzes the bidirectional transmission effect between the well section-level macroscopic physical field and the pore-level microscopic physical field; the error, process, and topology mapping unit establishes a three-layer mapping of error coupling features, construction process status, and fracture network topology; the chaotic adaptive control unit achieves adaptive matching control based on error chaotic features and cross-scale transmission coefficients; and the reservoir repair feedback unit simultaneously completes reservoir microfracture plugging repair and clay expansion inhibition, and provides feedback optimization.

2. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: The error and physical field coupled acquisition unit includes a multi-parameter synchronous acquisition module and a coupled feature extraction module. The multi-parameter synchronous acquisition module acquires the error sequence of AI model prediction of pressure, displacement, sand ratio and measured values ​​through a fiber optic sensor with a sampling frequency of 1kHz. It also acquires reservoir temperature field, stress field and seepage field data through distributed fiber optic temperature measurement, microseismic monitoring and downhole flow sensor respectively. The coupled feature extraction module calculates the correlation between the error sequence and each physical field parameter through the mutual information entropy algorithm, extracts 15 coupled features including 6 correlation parameters, the maximum Lyapunov exponent of the error sequence and the physical field gradient value, and forms a [0,1] standardized dataset.

3. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: The cross-scale physical field analytical unit defines the macroscopic physical field with a unit of 50 meters and the microscopic physical field based on the dynamic digital twin model of reservoir core CT scan; the macroscopic-to-microscopic transmission coefficient K1 and the microscopic-to-macroscopic transmission coefficient K2 are calculated, where K1 = Δr mi / ΔG σ , Δr mi ΔG represents the change in the radius of the micropore throat. σ This represents the change in macroscopic stress gradient; K2 = Δv m / Δα mi Δv m Let Δα be the change in flow rate in the macroscopic seepage field. mi The change in the microscopic clay swelling rate is represented by K1 and K2. The root cause of the positioning error is analyzed by correlation analysis between K1, K2 and the coupled feature set.

4. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: In the three-layer mapping model of the error, process and topology mapping unit, the input layer is a set of 15 error and physical field coupled features, the middle layer outputs the proppant placement uniformity, fracturing fluid turbulence intensity and proppant clustering coefficient, and the output layer outputs fracture density, fracture connectivity and main fracture length. The dual discriminator is trained in a collaborative manner, with the topology parameters measured by the drilling imaging tool and the process status measured by the fiber optic sensor as labels, respectively. The topology parameter thresholds are dynamically corrected according to the cross-scale transmission coefficients K1 and K2.

5. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: The macroscopic parameter control of the chaotic adaptive control unit is calculated using the displacement adjustment ΔQ = 0.5 × λ × (75% - η), where λ is the maximum Lyapunov exponent of the error sequence and η is the fracture connectivity. The microscopic parameter control is calculated using the clay inhibitor concentration adjustment ΔC = 0.3 × K² × (α) mi The calculation is based on -15%), where ΔC is the clay mineral swelling rate, and the proppant particle size is adjusted according to the pore throat radius. Topology optimization control is calculated using pulse frequency f = λ × 1.5 Hz, and the direction of sand addition is adjusted simultaneously through microseismic monitoring to maintain the crack density at 2-3 cracks per meter.

6. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: The reservoir repair feedback unit uses ±0.5 MPa fluid pressure pulses to impact the proppant powder blocking microfractures. Microfracture repair is considered complete when the mutual information entropy I(ΔQ,v) between the discharge error and the seepage field increases from <0.3 to 0.3-0.

6. Clay swelling is suppressed by adjusting the fracturing fluid ratio; when the clay swelling rate α... mi When the concentration of the low-concentration inhibitor decreases from >15% to ≤12%, the low-concentration inhibitor maintenance mode is activated, and the repaired physical field parameters are fed back to the error and physical field coupling acquisition unit to update the coupling feature set.

7. The artificial intelligence-based dynamic evaluation and control system for fracturing effect as described in claim 1, characterized in that: The system's workflow includes a coupled acquisition phase, a cross-scale analysis phase, a three-layer mapping phase, a chaos regulation phase, and a repair feedback phase; the coupled acquisition phase extracts 15 coupled features. The cross-scale analysis stage calculates K1 and K2 and locates the root cause of the error; the three-layer mapping stage generates the process state and topology parameters of the adversarial network output through dual discriminators; the chaotic regulation stage initiates adaptive regulation. The repair feedback phase monitors the repair effectiveness and iteratively optimizes the process.

Images