Multi-source information fusion and intelligent interpretation technology in hot dry rock crack identification

Abstract

The invention discloses a multi-source information fusion and intelligent interpretation technology in hot dry rock fracture identification, and the technology specifically comprises the following steps: 1, data collection: collecting multi-source data including seismic reflected waves, logging lithologic parameters, electromagnetic abnormal data and thermal infrared remote sensing images, and building a hot dry rock reservoir physical property parameter database, and 2, data preprocessing: carrying out denoising and offset correction on seismic data, normalizing logging data, inverting resistivity distribution of electromagnetic data, and extracting a temperature abnormal region of a thermal infrared image. The invention relates to the technical field of geothermal development. According to the multi-source information fusion and intelligent interpretation technology in hot dry rock fracture identification, by integrating multi-modal information such as geology, geophysics, remote sensing and drilling data and combining a feature level and decision level fusion strategy, a sensing blind area of a single sensor is effectively made up, the data are preprocessed, and the recognition accuracy is improved. And the capability of effectively inhibiting complex environmental factors such as seismic wave noise and electromagnetic interference is improved.

Description

Technical Field

[0001] This invention relates to the field of geothermal development technology, specifically to multi-source information fusion and intelligent interpretation technology in the identification of fissures in dry hot rocks. Background Technology

[0002] Hot dry rock is a high-temperature rock mass with a temperature generally above 200°C, buried at a depth of several thousand meters, and containing no fluid or only a small amount of underground fluid. The composition of such rock masses can vary greatly. Most of them are intermediate to acidic intrusive rocks from the Mesozoic era, but they can also be metamorphic rocks from the Mesozoic and Cenozoic eras, or even massive sedimentary rocks with huge thicknesses. Hot dry rock is mainly used to extract heat from its interior, so its main industrial indicator is the temperature inside the rock mass.

[0003] Referring to patent publication number "CN116360008A", a method for characterizing natural fractures in hot dry rock reservoirs is disclosed, including the following steps: S1: Conducting a natural fracture survey in the hot dry rock mining area and its surroundings, performing geological statistical analysis, and qualitatively determining the dominant orientation and occurrence information of natural fractures; S2: Conducting natural fracture analysis of well cores, and determining the dip and dip angle attributes of natural fractures in the reservoir by combining the analysis of natural fractures in the reservoir cores with paleomagnetic analysis of the cores; S3: Conducting statistical analysis of alteration minerals in rock cuttings, and quantitatively determining the density of natural fracture spatial development by continuously statistically analyzing the content of alteration minerals in the rock cuttings.

[0004] As shown in the above technology, existing detection methods are not robust to random noise in geophysical data, such as electromagnetic interference and seismic multiples, and are prone to false fracture misjudgments. They have not established a probability model for the existence of fractures, and cannot distinguish between high-confidence targets and ambiguous areas, which leads to increased drilling planning risks. Moreover, a single data source cannot fully reflect the three-dimensional distribution and connectivity of fractures. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-source information fusion and intelligent interpretation technology for identifying fissures in hot dry rocks, which solves the problems of misjudgment and the inability of single data to reflect the spatial distribution characteristics of deep underground layers in existing hot dry rock fissure identification.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks, specifically including the following steps:

[0007] Step 1: Data Acquisition: Collect multi-source data including seismic reflection waves, well logging lithology parameters, electromagnetic anomaly data, and thermal infrared remote sensing images, and establish a database of dry hot rock reservoir physical property parameters.

[0008] Step 2: Data preprocessing: Denoising and migration correction are performed on seismic data, well logging data is normalized, resistivity distribution is inverted from electromagnetic data, and temperature anomaly areas are extracted from thermal infrared images.

[0009] Step 3: Multi-source feature extraction: Seismic data: extract fracture reflection features using convolutional neural networks; Well logging data: extract lithological sequence correlation features based on recurrent neural networks; Electromagnetic data: use spectral clustering algorithm to divide resistivity anomaly areas; Thermal infrared data: use threshold segmentation to extract high temperature anomaly areas.

[0010] Step 4: Information Fusion and Fracture Modeling: Primary Fusion: The spatial features of seismic and electromagnetic data are superimposed to generate a fracture probability distribution map. Advanced Fusion: The DS evidence theory is introduced, and well logging lithology parameters and thermal infrared anomaly data are combined to correct the fracture probability model, construct a three-dimensional fracture network model, and mark high-confidence fracture areas.

[0011] Step 5: Optimization and Validation: Optimize model parameters through transfer learning to reduce overfitting. Use cross-validation to compare the recognition accuracy before and after fusion. Output the final fracture distribution map and mining suggestions. The intelligent identification method for rock mass structure surfaces based on multi-source data processes the data in a unified manner. After constructing a scientific model and optimizing the algorithm, it automatically interprets the structural surface information. By establishing a data sample library, the algorithm can use these sample libraries for learning and training.

[0012] Preferably, in step one, when acquiring seismic wave data, a broadband seismograph is used to acquire three-dimensional seismic reflection waveforms, and the propagation characteristics of P-waves and S-waves are recorded simultaneously for inverting the spatial distribution of fractures. In step one, when acquiring well logging data, lithological parameters are obtained through logging while drilling, including density, porosity, sonic transit time, and well temperature gradient. Gamma ray spectroscopy logging is used to identify radioactive anomaly zones. In step one, when acquiring electromagnetic data, magnetotellurics is used to measure resistivity distribution, and controlled-source audio-frequency magnetotellurics is used to enhance shallow resolution. In step one, when acquiring thermal infrared remote sensing data, Landsat-8 thermal infrared band images are acquired, and nighttime AST_08 data is used to eliminate solar radiation interference and extract surface temperature anomaly zones.

[0013] Preferably, in step one, when acquiring preprocessed seismic data, wavelet transform is used to remove random noise, and Kirchhoff migration correction is used to eliminate the influence of formation dip angle. In step one, when acquiring well logging data, Z-score standardization is used to eliminate dimensional differences, and lithological sequences are smoothed using filters. In step one, when acquiring electromagnetic data, the nonlinear conjugate gradient method is used to invert the three-dimensional resistivity distribution model, and static effects caused by topography are eliminated. In step one, when acquiring thermal infrared data, the radiative transfer equation method is applied to invert surface temperature, and high-temperature anomaly targets are extracted by combining elevation zoning threshold segmentation.

[0014] Preferably, during the seismic feature extraction, an improved ResNet-34 network is constructed, and seismic profile slices are input. The slices have a pixel size of 256×256. The fracture reflection features are extracted through multi-scale convolutional layers, and the dominant frequency attenuation characteristics are extracted by combining spectral analysis to quantify the wave velocity anomalies caused by the fracture zone.

[0015] Preferably, the well logging sequence analysis uses a bidirectional LSTM network to process the well logging parameter sequence, captures the characteristic correlation of lithological abrupt change points, and generates lithology-fracture probability curves by weighting key layers through an attention mechanism.

[0016] Preferably, the electromagnetic and thermal anomaly fusion applies a spectral clustering algorithm to the electromagnetic inversion results, divides resistivity anomaly blocks, marks potential fracture development areas, fuses thermal infrared temperature anomaly maps and electromagnetic anomaly blocks, calculates joint confidence using DS evidence theory, and eliminates isolated noise points.

[0017] Preferably, in step four, the primary fusion spatially superimposes seismic features and electromagnetic anomaly confidence levels to generate a fracture probability density distribution map, introduces well logging lithology constraints, performs Bayesian correction on the probability map, and suppresses false positive signals caused by non-fractures.

[0018] Preferably, in step four, the advanced fusion is based on the results of thermal infrared anomaly zone and seismic-electromagnetic fusion. The Delaunay triangulation algorithm is used to construct a three-dimensional topological network of fractures. Through fracture connectivity analysis, the distribution of main fracture channels and branch fractures is marked, and a permeability classification map is output.

[0019] Beneficial effects

[0020] This invention provides a multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks. Compared with existing technologies, it has the following advantages:

[0021] 1. The multi-source information fusion and intelligent interpretation technology for identifying dry hot rock fractures integrates multi-modal information such as geological, geophysical, remote sensing and borehole data, and combines feature-level and decision-level fusion strategies to effectively compensate for the perception blind spots of single sensors. Furthermore, the data preprocessing effectively enhances the ability to suppress complex environmental factors such as seismic wave noise and electromagnetic interference.

[0022] 2. The multi-source information fusion and intelligent interpretation technology in the identification of hot dry rock fractures uses electromagnetic data to locate macroscopic anomalies, seismic data to depict detailed morphology, thermal infrared data to correlate with surface thermal effects, and well logging data to provide vertical constraints; dynamic modeling: it integrates DS evidence theory and deep learning to upgrade from static probability distribution to dynamic fracture network, solving the problem of repeated modeling caused by differences in geological conditions in traditional methods, and effectively improving modeling efficiency.

[0023] 3. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks automates the entire process from data preprocessing to 3D visualization, supports parallel interpretation of multiple borehole nodes, and achieves significant time savings compared to traditional interpretation methods. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the steps involved in identifying fissures in dry, hot rocks according to the present invention. Detailed Implementation

[0025] 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.

[0026] Please see Figure 1 This invention provides a technical solution: a multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks, specifically including the following steps:

[0027] Step 1: Data Acquisition: Acquire multi-source data including seismic reflection waves, well logging lithological parameters, electromagnetic anomaly data, and thermal infrared remote sensing images, and establish a database of dry hot rock reservoir physical parameters. In Step 1, when acquiring seismic wave data, a broadband seismograph is used to acquire three-dimensional seismic reflection waveforms, simultaneously recording the propagation characteristics of P-waves and S-waves for inverting fracture spatial distribution. In Step 1, when acquiring well logging data, lithological parameters are obtained through logging while drilling, including density, porosity, sonic transit time, and well temperature gradient. Gamma ray spectroscopy logging is used to identify radioactive anomaly zones. In Step 1, when acquiring electromagnetic data, magnetotellurics is used to measure resistivity distribution, and controlled-source audio-frequency magnetotellurics is used to enhance shallow layer resolution. In Step 1, when acquiring thermal infrared remote sensing data, Landsat-8 thermal infrared band images are acquired, and nighttime AST_08 data is used to eliminate solar radiation interference and extract surface temperature anomaly zones.

[0028] In Step 1, when acquiring and preprocessing seismic data, wavelet transform is used to remove random noise, and Kirchhoff migration correction is used to eliminate the influence of stratigraphic dip angle. In Step 1, when acquiring well logging data, Z-score standardization is used to eliminate dimensional differences, and lithological sequences are smoothed using filters. In Step 1, when acquiring electromagnetic data, the nonlinear conjugate gradient method is used to invert the three-dimensional resistivity distribution model, and static effects caused by topography are eliminated. In Step 1, when acquiring thermal infrared data, the radiative transfer equation method is applied to invert surface temperature, and high-temperature anomaly targets are extracted by combining elevation zoning threshold segmentation.

[0029] By integrating multimodal information such as geological, geophysical, remote sensing, and borehole data, and combining feature-level and decision-level fusion strategies, the system effectively compensates for the blind spots of single sensors. Furthermore, by preprocessing the data, it effectively enhances the ability to suppress complex environmental factors such as seismic wave noise and electromagnetic interference.

[0030] Step 2: Data preprocessing: Denoising and migration correction are performed on seismic data, well logging data is normalized, resistivity distribution is inverted from electromagnetic data, and temperature anomaly areas are extracted from thermal infrared images.

[0031] Step 3: Multi-source feature extraction: Seismic data: fracture reflection features are extracted using a convolutional neural network; Well logging data: lithological sequence correlation features are extracted based on a recurrent neural network; Electromagnetic data: resistivity anomaly zones are divided using a spectral clustering algorithm; Thermal infrared data: high-temperature anomaly zones are extracted using threshold segmentation. An improved ResNet-34 network is constructed for seismic feature extraction. Seismic profile slices are input, with a pixel size of 256×256. Fracture reflection features are extracted through multi-scale convolutional layers, and the dominant frequency attenuation characteristics are extracted by combining spectral analysis to quantify wave velocity anomalies caused by fracture zones.

[0032] When analyzing the logging sequence, a bidirectional LSTM network is used to process the logging parameter sequence, capture the characteristic correlation of lithological abrupt change points, and generate lithology-fracture probability curves by weighting key intervals through an attention mechanism.

[0033] The electromagnetic and thermal anomalies were fused. A spectral clustering algorithm was applied to the electromagnetic inversion results to divide resistivity anomaly blocks, mark potential fracture development areas, fuse thermal infrared temperature anomaly maps and electromagnetic anomaly blocks, calculate joint confidence using DS evidence theory, and remove isolated noise points.

[0034] Step 4: Information Fusion and Fracture Modeling: Primary Fusion: Spatial features of seismic and electromagnetic data are superimposed to generate a fracture probability distribution map. Advanced Fusion: DS evidence theory is introduced, and well logging lithology parameters and thermal infrared anomaly data are combined to correct the fracture probability model, construct a three-dimensional fracture network model, and mark high-confidence fracture areas. In Step 4, primary fusion spatially superimposes seismic features and electromagnetic anomaly confidence to generate a fracture probability density distribution map. Well logging lithology constraints are introduced, and Bayesian correction is applied to the probability map to suppress false positive signals caused by non-fractures.

[0035] In step four, the advanced fusion is based on the results of thermal infrared anomaly zones and seismic-electromagnetic fusion. The Delaunay triangulation algorithm is used to construct a three-dimensional fracture topology network. Through fracture connectivity analysis, the distribution of main fracture channels and branch fractures is marked, and a permeability classification map is output.

[0036] Step 5: Optimization and Validation: Optimize model parameters through transfer learning to reduce overfitting. Use cross-validation to compare the recognition accuracy before and after fusion. Output the final fracture distribution map and mining suggestions. The intelligent identification method for rock mass structure surfaces based on multi-source data processes the data in a unified manner. After constructing a scientific model and optimizing the algorithm, it automatically interprets the structural surface information. By establishing a data sample library, the algorithm can use these sample libraries for learning and training.

[0037] Electromagnetic data is used to locate macroscopic anomalies, seismic data to depict detailed morphology, thermal infrared data to correlate surface thermal effects, and well logging data to provide vertical constraints. Dynamic modeling: By integrating DS evidence theory and deep learning, the model upgrades from static probability distribution to dynamic fracture network, solving the problem of repetitive modeling caused by differences in geological conditions in traditional methods. The modeling efficiency is effectively improved. The entire process from data preprocessing to 3D visualization is automated, supporting parallel interpretation of multiple borehole nodes, resulting in significant time savings compared to traditional interpretation methods.

[0038] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0039] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks, characterized by: Specifically, the following steps are included: Step 1: Data Acquisition: Acquire multi-source data including seismic reflection waves, well logging lithology parameters, electromagnetic anomaly data, and thermal infrared remote sensing images, and establish a database of dry hot rock reservoir physical property parameters; Step 2: Data preprocessing: Denoising and migration correction are performed on seismic data, well logging data is normalized, resistivity distribution is inverted from electromagnetic data, and temperature anomaly areas are extracted from thermal infrared images. Step 3: Multi-source feature extraction: Seismic data: extract fracture reflection features using convolutional neural networks; Well logging data: extract lithological sequence correlation features based on recurrent neural networks; Electromagnetic data: use spectral clustering algorithm to divide resistivity anomaly areas; Thermal infrared data: use threshold segmentation to extract high temperature anomaly areas. Step 4: Information Fusion and Fracture Modeling: Primary Fusion: The spatial features of seismic and electromagnetic data are superimposed to generate a fracture probability distribution map. Advanced Fusion: The DS evidence theory is introduced, and well logging lithology parameters and thermal infrared anomaly data are combined to correct the fracture probability model, construct a three-dimensional fracture network model, and mark high-confidence fracture areas. Step 5: Optimization and Validation: Optimize model parameters through transfer learning to reduce overfitting. Use cross-validation to compare the recognition accuracy before and after fusion. Output the final fracture distribution map and mining suggestions. The intelligent identification method for rock mass structure surfaces based on multi-source data processes the data in a unified manner. After constructing a scientific model and optimizing the algorithm, it automatically interprets the structural surface information. By establishing a data sample library, the algorithm can use these sample libraries for learning and training.

2. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: In step one, when acquiring seismic wave data, a broadband seismograph is used to acquire three-dimensional seismic reflection waveforms, and the propagation characteristics of P-waves and S-waves are recorded simultaneously for inverting the spatial distribution of fractures. In step one, when acquiring well logging data, lithological parameters are obtained through logging while drilling, including density, porosity, sonic transit time, and well temperature gradient. Gamma ray spectroscopy logging is used to identify radioactive anomaly zones. In step one, when acquiring electromagnetic data, magnetotellurics is used to measure resistivity distribution, and controlled-source audio-frequency magnetotellurics is used to enhance shallow resolution. In step one, when acquiring thermal infrared remote sensing data, Landsat-8 thermal infrared band images are acquired, and nighttime AST_08 data is used to eliminate solar radiation interference and extract surface temperature anomaly zones.

3. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: In step one, when acquiring preprocessed seismic data, wavelet transform is used to remove random noise, and Kirchhoff migration correction is used to eliminate the influence of formation dip angle. In step one, when acquiring well logging data, Z-score standardization is used to eliminate dimensional differences, and lithological sequences are smoothed using filters. In step one, when acquiring electromagnetic data, the nonlinear conjugate gradient method is used to invert the three-dimensional resistivity distribution model, and static effects caused by topography are eliminated. In step one, when acquiring thermal infrared data, the radiative transfer equation method is applied to invert surface temperature, and high-temperature anomaly targets are extracted by combining elevation zoning threshold segmentation.

4. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: The earthquake feature extraction process involves constructing an improved ResNet-34 network, inputting seismic profile slices with a pixel size of 256×256, extracting fracture reflection features through multi-scale convolutional layers, and combining spectral analysis to extract the dominant frequency attenuation characteristics, thereby quantifying the wave velocity anomalies caused by fracture zones.

5. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: The well logging sequence analysis employs a bidirectional LSTM network to process the well logging parameter sequence, captures the characteristic correlation of lithological abrupt change points, and generates lithology-fracture probability curves by weighting key layers through an attention mechanism.

6. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: The electromagnetic and thermal anomaly fusion method applies a spectral clustering algorithm to the electromagnetic inversion results, divides resistivity anomaly blocks, marks potential fracture development areas, merges thermal infrared temperature anomaly maps and electromagnetic anomaly blocks, calculates joint confidence using DS evidence theory, and eliminates isolated noise points.

7. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: In step four, the primary fusion spatially superimposes seismic features and electromagnetic anomaly confidence levels to generate a fracture probability density distribution map. Well logging lithology constraints are introduced, and Bayesian correction is applied to the probability map to suppress false positive signals caused by non-fractures.

8. The multi-source information fusion and intelligent interpretation technology for identifying fractures in hot dry rocks according to claim 1, characterized in that: In step four, the advanced fusion is based on the results of thermal infrared anomaly zone and seismic-electromagnetic fusion. The Delaunay triangulation algorithm is used to construct a three-dimensional fracture topology network. Through fracture connectivity analysis, the distribution of main fracture channels and branch fractures is marked, and a permeability classification map is output.

Images