Deep geothermal fluid cause and migration path tracing method

Abstract

The invention discloses a deep geothermal fluid cause and migration path tracing method, and relates to the technical field of earth scientific exploration, and the method specifically comprises the following steps: A1, collecting a geothermal fluid sample, A2, fusing high-precision three-dimensional seismic data, logging data and drilling data, constructing a multi-scale fracture network model, quantifying fracture connectivity and permeability anisotropy, and calculating the fracture connectivity and permeability anisotropy according to the fracture connectivity and permeability anisotropy. A3, developing a long short-term memory (LSTM) network model, inputting temperature, pressure and chemical field data monitored in real time, dynamically predicting a fluid migration path, and introducing a generative adversarial network (GAN) optimization model, and A4, deploying a distributed optical fiber sensor network. According to the sensing method for the aviation high-speed moving target, through the core-shell structure tracer agent, high fluorescence efficiency and geological compatibility are considered, a complete innovation chain is formed in the aspects of material design, detection method and environmental adaptability of the nano tracer agent, and the precision and sustainability of tracing of a migration path of deep geothermal fluid are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of earth science exploration technology, and in particular to a method for tracing the genesis and migration path of deep geothermal fluids. Background Art

[0002] As a clean, renewable energy source, the efficient exploration and development of deep geothermal resources is of strategic importance for promoting green and sustainable development amidst the global energy transition. Accurately uncovering the genesis and migration pathways of deep geothermal fluids is a key prerequisite for the scientific development and rational utilization of geothermal resources.

[0003] The development of deep geothermal resources relies on the accurate identification of the origin and migration paths of geothermal fluids. Existing technologies have the following drawbacks: Traditional methods rely on single isotopes or geochemical parameters and are easily disturbed by the mixing of multi-source fluids, leading to errors in the identification of genesis: Numerical simulations based on two-dimensional geological profiles cannot reflect the fluid migration patterns in complex three-dimensional fracture networks, and lack the ability to track the real-time changes in the coupled temperature-pressure-chemical fields during geothermal fluid migration. To this end, the present invention provides a method for tracing the genesis and migration paths of deep geothermal fluids. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method for tracing the genesis and migration path of deep geothermal fluids, which solves the problem that traditional methods rely on single isotopes or geochemical parameters and are easily disturbed by the mixing of multi-source fluids, resulting in errors in the cause identification.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for tracing the genesis and migration path of deep geothermal fluids, specifically comprising the following steps:

[0006] A1. Collect geothermal fluid samples, determine isotope composition, noble gas composition, trace element fingerprints and water-rock reaction indicators, and establish a multidimensional feature database;

[0007] A2. Integrate high-precision 3D seismic data, well logging data, and borehole data to construct a multi-scale fracture network model, quantify fracture connectivity and permeability anisotropy, couple geochemical fingerprint data with fracture network parameters, and generate a fluid migration probability field through Monte Carlo simulation.

[0008] A3. Develop a long short-term memory (LSTM) model, input real-time monitored temperature, pressure, and chemical field data, dynamically predict fluid migration paths, introduce a generative adversarial network (GAN) optimization model, generate virtual path scenarios through historical data training, and improve prediction robustness.

[0009] A4. Deploy a distributed fiber optic sensor network to achieve multi-parameter synchronous monitoring of temperature, strain and chemical fields.

[0010] Preferably, the determination of isotopic composition comprises: 、 、 、 , rare gas components include: He, Ne, Ar, trace element fingerprints include: Li, B, As, Sb and water-rock reaction indicators include , pH, Eh, the A1 introduces environmentally friendly nanotracers to achieve high-sensitivity in situ detection through laser induced breakdown spectroscopy (LIBS).

[0011] Preferably, the specific method of introducing environmentally friendly nano-tracers to achieve high-sensitivity in-situ detection by laser-induced breakdown spectroscopy (LIBS) is as follows:

[0012] B1, select CdSe / ZnS core-shell structure quantum dots, the emission wavelength covers the visible light to near infrared region (500-800nm), and achieve multi-color coding by adjusting the particle size (2-10nm), using improved Sol-gel method, dense deposition on the surface of quantum dots Shell (thickness 10-20nm), forming The core-shell structure blocks direct contact between quantum dots and the external environment;

[0013] B2, through silane coupling agent The surface is modified to introduce active functional groups to enhance the dispersibility of the nanoparticles in aqueous solution and their compatibility with geothermal fluid components. The outer layer is coated with polyethylene glycol or zwitterionic polymer to reduce the nonspecific adsorption of nanoparticles in the geothermal system and prolong the circulation time. Then, nanotracers are injected and their migration is monitored. Depending on the volume V of the geothermal system and the monitoring requirements, nanotracers are injected at a concentration of 0.1-1ppm, reducing the total dosage by more than 90% compared to traditional dyes. The tracer solution is slowly injected into the target layer along the borehole wall through a dedicated injection pump to avoid disturbing the fracture network. Optical fibers are arranged along the geothermal wellbore and potential migration paths, with monitoring points set up at intervals of 10-50m to achieve simultaneous acquisition of temperature, strain, and chemical fields. A portable LIBS system is installed at the key monitoring wellhead. The laser wavelength is 1064nm, the pulse energy is 10mJ, and the focused spot diameter is <100μm, which can penetrate the wellbore fluid to directly excite the nanoparticles.

[0014] B3. Laser excitation: High-energy pulsed laser is focused on the fluid sample, instantly vaporizing and ionizing the nanoparticles to generate plasma. During the cooling process of the plasma, characteristic spectra are emitted. The spectra are separated by a spectrometer (resolution 0.1nm) and recorded by a CCD detector. The emission peaks of quantum dots include Cd (226.5nm), Se (196.0nm), and Zn (213.8nm) element lines, which are used to quantify the tracer concentration. The matrix peaks include Si (288.2nm) and O (777.4nm) lines to verify the integrity of the nanoparticles. Finally, a calibration curve between the spectral intensity I and the tracer concentration C was established. , n=0.5-1.2, eliminate the matrix effect interference by multivariate linear regression, combined with the fiber optic sensor temperature and strain The particle filter algorithm is used to reconstruct the migration trajectory of nanoparticles based on the data, with a spatial resolution of 1m.

[0015] Preferably, the specific method of constructing a multi-scale fracture network model and quantifying fracture connectivity and permeability anisotropy is:

[0016] C1. Constrain the macroscopic fracture distribution using 3D seismic data. Extract the spatial location of the fault zone through seismic attribute analysis. Combined with sequence stratigraphy, delineate the top and bottom interfaces of the thermal reservoir, and construct a macroscopic fracture network skeleton. Then, use well logging data to correct the mesoscopic fracture density: Use imaging logging (FMI, UBI) to identify the fracture direction and aperture (0.1-10mm) around the wellbore. Combined with full-wavelength acoustic logging, invert the fracture filling type and establish a fracture density attenuation model with depth (exponential function: , is the attenuation coefficient), and finally the micro-fracture connectivity is calibrated by drilling data. The micro-fracture porosity (0.1%-5%) is statistically analyzed by core slice observation. The pore throat radius distribution (0.01-1μm) is obtained by combining mercury injection experiments. The micro-fracture connectivity probability function ( ), k is the connectivity coefficient);

[0017] C2. Discrete Fracture Network (DFN) modeling and anisotropy quantification, multi-scale fracture generation, based on fractal theory, generating fracture networks at macro (length > 100m), meso (1-100m), and micro (<1m) scales, using a Poisson process to control the distribution of fracture spacing, and Monte Carlo sampling to determine fracture azimuths (obeying Fisher distribution, concentration parameter =5-20), then calculate the permeability anisotropy and assign a permeability tensor to each fracture ( , t is the opening), the regional equivalent permeability tensor is calculated by volume averaging method ( ), quantify the main permeability direction (K_max / K_min ratio reaches - );

[0018] C3, Monte Carlo simulation and migration probability field generation, parameter random sampling: fracture opening (t~N( , ))、Permeability(K~LN( , )), geochemical indicators (such as (-10‰,-5‰)) performs Latin hypercube sampling (LHS) to generate - The improved random walk particle tracking (RWPT) algorithm is used to combine the parameters of the group. Each particle carries the geochemical fingerprint characteristics. When moving along the fracture network, the fingerprint value is updated according to the fracture properties (openness, permeability) and fluid mixing rules (complete mixing / laminar flow). The path frequency of all particles reaching the observation point is counted, and the posterior probability (P( )=P( ) , is the fracture parameter, D is the geochemical observation data), and a migration probability density cloud map (resolution 10m×10m×1m) is generated.

[0019] Preferably, the calculation formula for developing the long short-term memory network (LSTM) model in A3 is:

[0020]

[0021] express The predicted geothermal fluid migration path vector, which contains key information such as migration direction and speed, is used to describe the migration state of geothermal fluid at that moment. is a weight coefficient used to balance the contribution of the LSTM model prediction results and the GAN optimization results. It can be adjusted according to actual data and prediction requirements. The Long Short-Term Memory Network (LSTM) is based on the time The prediction function of the real-time monitoring data outputs the predicted value of the geothermal fluid migration path. Represents pressure, temperature, chemical field data, Generative Adversarial Network (GAN) is based on historical data The generated virtual path scenario is used to optimize the prediction results of the LSTM model and enhance the robustness of the prediction.

[0022] Preferably, the calculation formula of the adversarial generative network (GAN) optimization model in A2 is:

[0023]

[0024] represents the input vector (temperature, pressure, chemical field data) at time t, Indicates the hidden state at the last moment. Indicates the cell state at the previous moment, represents element-wise multiplication, represents the hyperbolic tangent function, Indicates how much historical information to retain. Indicates how much new information to add. Indicates which information is output.

[0025] Preferably, the algorithm formula of the Generative Adversarial Network (GAN) optimization model in A2 is:

[0026]

[0027] represents the objective function of GAN, Represents the discriminator, judging whether the input data is real data, Dummy data generated by the generator, Represents the discriminator's response to real data The discriminant output of represents a generator, Indicates receiving random noise, Indicates the generation of a virtual geothermal fluid migration path scenario.

[0028] Beneficial effects

[0029] The present invention provides a method for tracing the genesis and migration path of deep geothermal fluids. Compared with existing technologies, it has the following advantages:

[0030] By using core-shell tracers, LIBS technology achieves in-situ, non-destructive, and real-time detection downhole, balancing high fluorescence efficiency and geological compatibility. This replaces the traditional sampling-laboratory analysis process, reduces usage by over 90%, and reduces ecological risks to 1 / 50 of traditional methods. This is in line with the trend of green geothermal development. This technology forms a complete innovation chain in nanotracer material design, detection methods, and environmental adaptability, significantly improving the accuracy and sustainability of deep geothermal fluid migration path tracing. Furthermore, the fracture parameters corresponding to low-probability paths in Monte Carlo simulations are fed back to the DFN model, and fracture connectivity is optimized using a simulated annealing algorithm (objective function: ), realize model self-calibration, compare the thermal breakthrough time predicted by the probability field with the measured data of the optical fiber sensor (error <5%), verify the reliability of the model, realize the deep integration of geochemical information and three-dimensional fracture physical field, transform the traditional static geological model into a dynamic probability prediction tool, and significantly improve the analytical accuracy of deep geothermal fluid migration paths. By combining the three-dimensional fracture network model with the LSTM-GAN algorithm, the real-time prediction and error self-correction of the migration path are realized, and the isotope-rare gas-trace element joint tracing is used to improve the accuracy of cause discrimination. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Flowchart of the perception method of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0033] See also Figure 1 , the present invention provides a technical solution:

[0034] The method for tracing the genesis and migration path of deep geothermal fluids specifically includes the following steps:

[0035] A1. Collect geothermal fluid samples, determine isotope composition, noble gas composition, trace element fingerprints and water-rock reaction indicators, and establish a multidimensional feature database;

[0036] A2. Integrate high-precision 3D seismic data, well logging data, and borehole data to construct a multi-scale fracture network model, quantify fracture connectivity and permeability anisotropy, couple geochemical fingerprint data with fracture network parameters, and generate a fluid migration probability field through Monte Carlo simulation.

[0037] A3. Develop a long short-term memory (LSTM) model, input real-time monitored temperature, pressure, and chemical field data, dynamically predict fluid migration paths, introduce a generative adversarial network (GAN) optimization model, generate virtual path scenarios through historical data training, and improve prediction robustness.

[0038] A4. Deploy a distributed fiber optic sensor network to achieve multi-parameter synchronous monitoring of temperature, strain and chemical fields.

[0039] In an embodiment of the present invention, the isotopic composition determination includes: 、 、 、 , rare gas components include: He, Ne, Ar, trace element fingerprints include: Li, B, As, Sb and water-rock reaction indicators include , pH, Eh, the A1 introduces environmentally friendly nano-tracers, and achieves high-sensitivity in-situ detection through laser-induced breakdown spectroscopy (LIBS). Geothermal fluid samples at different depths (500m, 1000m, 1500m) are drilled and measured. (-8.5‰ to -6.2‰), (1.2×10⁻ 6 to 2.8×10⁻ 6 ) and Li / B ratio (0.5 to 1.2), combined with the regional geological background, it was determined that the fluid was a mixture of deep circulating metamorphic water and magmatic hydrothermal fluid.

[0040] In the embodiment of the present invention, the specific method of introducing environmentally friendly nano-tracers to achieve high-sensitivity in-situ detection by laser-induced breakdown spectroscopy (LIBS) is as follows:

[0041] B1, select CdSe / ZnS core-shell structure quantum dots, the emission wavelength covers the visible light to near infrared region (500-800nm), and achieve multi-color coding by adjusting the particle size (2-10nm), using improved Sol-gel method, dense deposition on the surface of quantum dots Shell (thickness 10-20nm), forming The core-shell structure blocks direct contact between quantum dots and the external environment;

[0042] B2, through silane coupling agent The surface is modified to introduce active functional groups to enhance the dispersibility of the nanoparticles in aqueous solution and their compatibility with geothermal fluid components. The outer layer is coated with polyethylene glycol or zwitterionic polymer to reduce the nonspecific adsorption of nanoparticles in the geothermal system and prolong the circulation time. Then, nanotracers are injected and their migration is monitored. Depending on the volume V of the geothermal system and the monitoring requirements, nanotracers are injected at a concentration of 0.1-1ppm, reducing the total dosage by more than 90% compared to traditional dyes. The tracer solution is slowly injected into the target layer along the borehole wall through a dedicated injection pump to avoid disturbing the fracture network. Optical fibers are arranged along the geothermal wellbore and potential migration paths, with monitoring points set up at intervals of 10-50m to achieve simultaneous acquisition of temperature, strain, and chemical fields. A portable LIBS system is installed at the key monitoring wellhead. The laser wavelength is 1064nm, the pulse energy is 10mJ, and the focused spot diameter is <100μm, which can penetrate the wellbore fluid to directly excite the nanoparticles.

[0043] B3. Laser excitation: High-energy pulsed laser is focused on the fluid sample, instantly vaporizing and ionizing the nanoparticles to generate plasma. During the cooling process of the plasma, characteristic spectra are emitted. The spectra are separated by a spectrometer (resolution 0.1nm) and recorded by a CCD detector. The emission peaks of quantum dots include Cd (226.5nm), Se (196.0nm), and Zn (213.8nm) element lines, which are used to quantify the tracer concentration. The matrix peaks include Si (288.2nm) and O (777.4nm) lines to verify the integrity of the nanoparticles. Finally, a calibration curve between the spectral intensity I and the tracer concentration C was established. , n=0.5-1.2, eliminate the matrix effect interference by multivariate linear regression, combined with the fiber optic sensor temperature and strain The particle filter algorithm is used to reconstruct the migration trajectory of nanoparticles based on the data, with a spatial resolution of 1m.

[0044] The above-mentioned QDs@SiO2 core-shell structure tracer takes into account high fluorescence efficiency and geological compatibility. LIBS technology realizes in-situ, non-destructive and real-time detection underground, replacing the traditional sampling-laboratory analysis process, reducing the usage by more than 90%, and reducing the ecological risk to 1 / 50 of the traditional method. It is in line with the trend of green geothermal development, forming a complete innovation chain in material design, detection methods and environmental adaptability of nanotracers, and significantly improving the accuracy and sustainability of tracing the migration path of deep geothermal fluids.

[0045] In the embodiment of the present invention, the specific method of constructing a multi-scale fracture network model and quantifying fracture connectivity and permeability anisotropy is:

[0046] C1. Constrain the macroscopic fracture distribution using 3D seismic data. Extract the spatial location of the fault zone through seismic attribute analysis. Combined with sequence stratigraphy, delineate the top and bottom interfaces of the thermal reservoir, and construct a macroscopic fracture network skeleton. Then, use well logging data to correct the mesoscopic fracture density: Use imaging logging (FMI, UBI) to identify the fracture direction and aperture (0.1-10mm) around the wellbore. Combined with full-wavelength acoustic logging, invert the fracture filling type and establish a fracture density attenuation model with depth (exponential function: , is the attenuation coefficient), and finally the micro-fracture connectivity is calibrated by drilling data. The micro-fracture porosity (0.1%-5%) is statistically analyzed by core slice observation. The pore throat radius distribution (0.01-1μm) is obtained by combining mercury injection experiments. The micro-fracture connectivity probability function ( , k is the connectivity coefficient);

[0047] C2. Discrete Fracture Network (DFN) modeling and anisotropy quantification, multi-scale fracture generation, based on fractal theory, generating fracture networks at macro (length > 100m), meso (1-100m), and micro (<1m) scales, using a Poisson process to control the distribution of fracture spacing, and Monte Carlo sampling to determine fracture azimuths (obeying Fisher distribution, concentration parameter =5-20), then calculate the permeability anisotropy and assign a permeability tensor to each fracture ( , t is the opening), the regional equivalent permeability tensor is calculated by volume averaging method ( ), quantify the main permeability direction (K_max / K_min ratio reaches - );

[0048] C3, Monte Carlo simulation and migration probability field generation, parameter random sampling: fracture opening ( )、Permeability(K~LN( )), geochemical indicators (such as (-10‰,-5‰)) performs Latin hypercube sampling (LHS) to generate - The improved random walk particle tracking (RWPT) algorithm is used to combine the parameters of the group. Each particle carries the geochemical fingerprint characteristics. When moving along the fracture network, the fingerprint value is updated according to the fracture properties (openness, permeability) and fluid mixing rules (complete mixing / laminar flow). The path frequency of all particles reaching the observation point is counted, and the posterior probability (P( )=P( ) , is the fracture parameter, D is the geochemical observation data), and a migration probability density cloud map (resolution 10m×10m×1m) is generated.

[0049] The fracture parameters corresponding to the low-probability paths in the Monte Carlo simulation are fed back to the DFN model, and the fracture connectivity is optimized using the simulated annealing algorithm (objective function: , ), realize model self-calibration, compare the thermal breakthrough time predicted by the probability field with the actual data measured by the optical fiber sensor (error <5%), verify the reliability of the model, realize the deep integration of geochemical information and three-dimensional fracture physical field, transform the traditional static geological model into a dynamic probability prediction tool, and significantly improve the analysis accuracy of deep geothermal fluid migration path.

[0050] In the embodiment of the present invention, the calculation formula for developing the long short-term memory network (LSTM) model in A3 is:

[0051]

[0052] express The predicted geothermal fluid migration path vector, which contains key information such as migration direction and speed, is used to describe the migration state of geothermal fluid at that moment. is a weight coefficient used to balance the contribution of the LSTM model prediction results and the GAN optimization results. It can be adjusted according to actual data and prediction requirements. The Long Short-Term Memory Network (LSTM) is based on the time The prediction function of the real-time monitoring data outputs the predicted value of the geothermal fluid migration path. Represents pressure, temperature, chemical field data, Generative Adversarial Network (GAN) is based on historical data The generated virtual path scenario is used to optimize the prediction results of the LSTM model and enhance the robustness of the prediction.

[0053] In this embodiment of the present invention, the calculation formula of the Generative Adversarial Network (GAN) optimization model in A2 is:

[0054]

[0055] Representative Moment Input vectors (temperature, pressure, chemical field data), Indicates the hidden state at the last moment. Indicates the cell state at the previous moment, represents element-wise multiplication, represents the hyperbolic tangent function, Indicates how much historical information to retain. Indicates how much new information to add. Indicates which information is output.

[0056] In the embodiment of the present invention, the algorithm formula of the generative adversarial network (GAN) optimization model in A2 is:

[0057]

[0058] represents the objective function of GAN, Represents the discriminator, judging whether the input data is real data, Dummy data generated by the generator, Represents the discriminator's response to real data The discriminant output of represents a generator, Indicates receiving random noise, Indicates the generation of a virtual geothermal fluid migration path scenario.

[0059] In the embodiment of the present invention, a distributed optical fiber sensor network is deployed in A4 to realize the simultaneous monitoring of multiple parameters of temperature, strain and chemical field. The edge computing module is used to realize real-time data processing and online model updating. The algorithm formula of the edge computing module is: , is the model parameter at the current moment, Denoted as the learning rate, Represented as the gradient of the loss function calculated at the edge, represents the preprocessed sensor data, It represents the real-time path prediction target. Through the edge-cloud collaborative computing architecture, the present invention realizes the "perception-decision-control" closed loop of geothermal fluid migration monitoring, significantly improving the safety and economy of deep geothermal resource development.

[0060] In summary, the data of high-speed moving targets is collected through acquisition equipment, and the data is integrated after preprocessing to establish a spatial model to identify the position and angle changes of high-speed moving targets. After learning and optimization, it is applied to the identification and acquisition equipment, which can identify and collect high-speed moving targets, and calculate the position of high-speed moving targets based on data conversion, and realize real-time update of data according to the movement of the moving target, and predict the future trajectory of the target. At the same time, it is convenient for the equipment to perceive high-speed moving targets and make real-time adjustments, thereby improving the accuracy of the data.

[0061] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0062] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0063] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for tracing the genesis and migration path of deep geothermal fluids, characterized by: The specific steps include: A1. Collect geothermal fluid samples, determine isotope composition, noble gas composition, trace element fingerprints and water-rock reaction indicators, and establish a multidimensional feature database; A2. Integrate high-precision 3D seismic data, well logging data, and borehole data to construct a multi-scale fracture network model, quantify fracture connectivity and permeability anisotropy, couple geochemical fingerprint data with fracture network parameters, and generate a fluid migration probability field through Monte Carlo simulation. A3. Develop a long-short-term memory network model, input real-time monitored temperature, pressure, and chemical field data, dynamically predict fluid migration paths, introduce a generative adversarial network optimization model, and generate virtual path scenarios through historical data training to improve prediction robustness. A4. Deploy a distributed fiber optic sensor network to achieve multi-parameter synchronous monitoring of temperature, strain and chemical fields.

2. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: Determining the isotopic composition includes: 、 、 、 , rare gas components include: He, Ne, Ar, trace element fingerprints include: Li, B, As, Sb and water-rock reaction indicators include , pH, Eh, the A1 introduces an environmentally friendly nanotracer and realizes high-sensitivity in situ detection through laser-induced breakdown spectroscopy.

3. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 2, characterized in that: The specific method of introducing environmentally friendly nano-tracers and achieving high-sensitivity in-situ detection through laser-induced breakdown spectroscopy is as follows: B1, select CdSe / ZnS core-shell structure quantum dots, the emission wavelength covers the visible light to the near infrared region, and multi-color coding is achieved by adjusting the particle size. Sol-gel method, dense deposition on the surface of quantum dots Shell, forming The core-shell structure blocks direct contact between quantum dots and the external environment; B2, through silane coupling agent The surface is modified to introduce active functional groups to enhance the dispersibility of the nanoparticles in aqueous solution and their compatibility with geothermal fluid components. The outer layer is coated with polyethylene glycol or zwitterionic polymer to reduce the nonspecific adsorption of nanoparticles in the geothermal system and prolong the circulation time. Nanotracers are then injected and their migration is monitored. Depending on the volume of the geothermal system and monitoring requirements, nanotracers are injected at a concentration of 0.1-1ppm, reducing the total amount by more than 90% compared to traditional dyes. A dedicated injection pump is used to slowly inject the tracer solution into the target layer along the borehole wall to avoid disturbing the fracture network. Optical fibers are arranged along the geothermal wellbore and potential migration paths, with monitoring points set up at intervals of 10-50m to achieve simultaneous acquisition of temperature, strain, and chemical fields. A portable LIBS system is installed at the key monitoring wellhead. The laser wavelength is 1064nm, the pulse energy is 10mJ, and the focused spot diameter is <100μm, which can penetrate the wellbore fluid to directly excite the nanoparticles. B3. Laser excitation: High-energy pulsed laser is focused on the fluid sample, instantly vaporizing and ionizing the nanoparticles to generate plasma. The plasma emits characteristic spectra during cooling. The spectrometer separates the spectra and the CCD detector records the spectral signals. The quantum dot emission peaks include Cd, Se, and Zn element lines, which are used to quantify the tracer concentration. The matrix peaks include Si and O lines, which verify the integrity of the nanoparticles. Finally, a calibration curve between spectral intensity and tracer concentration is established. The interference of matrix effects is eliminated through multivariate linear regression. Combined with the temperature and strain data of the optical fiber sensor, the particle filter algorithm is used to reconstruct the migration trajectory of the nanoparticles with a spatial resolution of 1m.

4. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: The specific method for constructing a multi-scale fracture network model and quantifying fracture connectivity and permeability anisotropy is: C1. 3D seismic data constrains the distribution of macroscopic fractures. Seismic attribute analysis is used to extract the spatial location of fault zones. Sequence stratigraphy is used to delineate the top and bottom interfaces of thermal reservoirs, and a macroscopic fracture network framework is constructed. Well logging data is then used to calibrate the mesoscopic fracture density. Imaging logging is used to identify the direction and aperture of fractures around the wellbore. Full-wavelength acoustic logging is used to invert the type of fracture filling material, and a model for fracture density attenuation with depth is established. Finally, borehole data is used to calibrate microscopic fracture connectivity. Core section observations are used to calculate the microfracture porosity. Mercury injection experiments are used to obtain the pore throat radius distribution, and a microfracture connectivity probability function is constructed. C2. Discrete fracture network modeling and anisotropy quantification. Multi-scale fracture generation. Based on fractal theory, fracture networks are generated at a macroscale. A Poisson process is used to control the distribution of fracture spacing. Monte Carlo sampling is used to determine the fracture azimuth. Permeability anisotropy is then calculated. A permeability tensor is assigned to each fracture. The regional equivalent permeability tensor is calculated using the volume averaging method to quantify the main permeability direction. C3, Monte Carlo simulation and migration probability field generation, parameter random sampling: Latin hypercube sampling is performed on fracture aperture, permeability, and geochemical indicators to generate - A parameter combination and a path tracing algorithm are then used. An improved random walk particle tracing method is used. Each particle carries a geochemical fingerprint feature. When moving along the fracture network, the fingerprint value is updated according to the fracture properties and fluid mixing rules. The path frequency of all particles reaching the observation point is counted, and the posterior probability is updated in combination with the Bayesian theorem to generate a migration probability density cloud map.

5. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: The calculation formula of the long short-term memory network (LSTM) model developed in A3 is: ； in express Predicted geothermal fluid migration path vector, is the weight coefficient, which is used to balance the contribution of the LSTM model prediction results and the GAN optimization results. The Long Short-Term Memory Network (LSTM) is based on the time The prediction function of the real-time monitoring data outputs the predicted value of the geothermal fluid migration path. Represents pressure, temperature, chemical field data, Generative Adversarial Network (GAN) is based on historical data Generated virtual path scene.

6. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: The calculation formula of the adversarial generative network optimization model in A3 is: ； in represents the input vector at time t, including temperature, pressure, and chemical field data, Indicates the hidden state at the last moment. Indicates the cell state at the previous moment, represents element-wise multiplication, represents the hyperbolic tangent function, Indicates how much historical information to retain. Indicates how much new information to add. Indicates which information is output.

7. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: The algorithm formula for dynamically predicting fluid migration path in A3 is: ； in represents the objective function of GAN, Represents the discriminator, judging whether the input data is real data, Dummy data generated by the generator, Represents the discriminator's response to real data The discriminant output of represents a generator, Indicates receiving random noise, Indicates the generation of a virtual geothermal fluid migration path scenario.

8. The method for tracing the genesis and migration path of deep geothermal fluids according to claim 1, characterized in that: A4 deploys a distributed fiber optic sensor network to achieve simultaneous monitoring of multiple parameters of temperature, strain, and chemical fields. The edge computing module enables real-time data processing and online model updates. The algorithm formula of the edge computing module is: , is the model parameter at the current moment, Denoted as the learning rate, Represented as the gradient of the loss function calculated at the edge, represents the preprocessed sensor data, Represents the real-time path prediction target.

Images