Geothermal resource prediction method and system based on multi-source fusion data

Abstract

The invention discloses a geothermal resource prediction method and system based on multi-source fusion data, and belongs to the technical field of geothermal resource exploration. Unified space reference registration and feature extraction are carried out on multi-source heterogeneous data such as geology, geophysics, geochemistry, remote sensing and wellholes, a geothermal resource prediction model capable of adaptively selecting an algorithm model is constructed, and finally a quantitative prediction result including a thermal storage probability, three-dimensional distribution of a geothermal field and a thermal reserve level is output. And generating an exploration target suggestion. According to the method, deep collaborative utilization and dynamic closed-loop optimization of multi-source data are realized, and the precision and reliability of geothermal resource prediction and the exploration decision efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of geothermal resource exploration technology, and more specifically to a geothermal resource prediction method and system based on multi-source fusion data. Background Technology

[0002] Currently, geothermal resources, as a clean and sustainable new energy source, hold significant strategic value in promoting the green transformation of the energy structure and alleviating pressure on traditional energy sources. The core prerequisite for their efficient development is the accurate prediction of resource distribution, temperature fields, and reserves. Current geothermal resource prediction work primarily relies on multi-dimensional geoscientific data, including geological data recording lithology and stratigraphic structure information, geophysical data encompassing gravity, magnetics, and magnetotelluric information, geochemical data including water sample ions and gas sample composition, remote sensing data reflecting surface temperature anomalies, and existing well data recording temperature measurement and production capacity information. These data present the geological background and resource characteristics of geothermal reservoirs from different perspectives, providing fundamental support for prediction work.

[0003] However, existing geothermal resource prediction methods still have room for further optimization in practical applications. Some methods do not adequately unify the spatial benchmark processing of multi-source data, and the accuracy of data correlation analysis is easily affected by differences in coordinate systems or mismatches in grid scales. In the feature extraction stage, the depth of mining for key parameters directly related to geothermal reservoir formation, such as reservoir thickness, low resistivity zone distribution, and geothermal gradient, is insufficient, which may limit the effectiveness of model input information. At the same time, the adaptability of model algorithm selection to data types needs to be improved, making it difficult to fully adapt to the spatial neighborhood characteristics, topological relationships, or nonlinear features of different data, and lacking a closed-loop optimization mechanism that continuously incorporates new exploration data, which to some extent affects the dynamic updating capability and reliability of prediction results.

[0004] Therefore, how to propose a geothermal resource prediction method and system based on multi-source fusion data, realize the deep fusion and synergistic utilization of multi-source heterogeneous data, and improve the accuracy and reliability of geothermal resource prediction is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a geothermal resource prediction method and system based on multi-source fusion data. By integrating multi-source heterogeneous data, it realizes in-depth mining and collaborative utilization of data value, effectively solves the problem of insufficient data fusion in traditional methods, improves the adaptability of prediction models to data types, and achieves dynamic updates and performance improvements.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: On the one hand, this invention proposes a geothermal resource prediction method based on multi-source fusion data, comprising the following steps: Acquire multi-source geoscience data; The multi-source geoscience data are subjected to unified spatial benchmark processing to obtain preprocessed data; Feature extraction is performed on the preprocessed data to obtain a set of geothermal reservoir characteristic parameters; Construct a geothermal resource prediction model; the geothermal resource prediction model selects the corresponding machine learning or deep learning algorithm based on the data structure of the geothermal reservoir characteristic parameter set; The geothermal resource prediction model is trained based on the set of geothermal reservoir characteristic parameters and training samples labeled with known well data. The set of geothermal reservoir characteristic parameters obtained after processing the region to be predicted is input into the trained geothermal resource prediction model, and the quantitative prediction results of geothermal resources are output.

[0007] Preferably, the multi-source geoscience data includes geological data, geophysical data, geochemical data, remote sensing data, and existing well data.

[0008] Preferably, the multi-source geoscientific data undergoes spatial benchmark unification to obtain preprocessed data, including the following data registration steps: Establish a unified geographic coordinate system and grid scale; The raster data is resampled to the uniform grid scale using a resampling algorithm. Vector data is transformed to the unified geographic coordinate system, and topology checks and corrections are performed. Calculate the root mean square error after data registration to verify accuracy. If the error exceeds the allowable accuracy range, repeat the data registration steps until the accuracy meets the requirements.

[0009] Preferably, feature extraction is performed on the preprocessed data to obtain a set of geothermal reservoir feature parameters, including: Extract the lithological area ratio, fault density, and stratum burial depth from the geological data; Gravity gradient, magnetic field distribution, resistivity distribution, and seismic velocity parameters are extracted from the geophysical data. Extract from the geochemical data the reservoir temperature based on water sample ion concentration, the groundwater recharge source based on isotopic composition, and the heat source type based on gas sample composition; The location and area of ​​surface temperature anomalies, linear structure density, and number of intersections are extracted from the remote sensing data. The geothermal gradient and productivity per unit depth are extracted from the existing well data.

[0010] Preferably, the geothermal resource prediction model selects a corresponding machine learning or deep learning algorithm based on the data structure of the geothermal reservoir characteristic parameter set, including: If the data structure is a grid containing spatial neighborhood information, then a convolutional neural network is selected; If the data structure contains complex topological relationships, then a graph neural network is selected; If the data structure is in the form of feature vectors and needs to handle nonlinear relationships, then random forest should be selected.

[0011] Preferably, training the geothermal resource prediction model includes: The set of geothermal reservoir characteristic parameters is associated with the corresponding tag data to form a sample set, which is then divided into a training set, a validation set, and a test set according to a preset ratio. The hyperparameters of the geothermal resource prediction model are tuned using either a grid search method or a Bayesian optimization method. The geothermal resource prediction model is trained using the training set, and the weight parameters are updated using the gradient descent method. The optimization is iteratively performed until the termination condition is met. The performance of the geothermal resource prediction model was evaluated using the test set, with evaluation metrics including mean absolute error and coefficient of determination.

[0012] Preferably, the quantitative prediction results of the geothermal resources include a geothermal reservoir probability map, a three-dimensional geothermal field distribution map, and estimated geothermal reservoir temperature and geothermal reservoir level. Based on the thermal reservoir probability map, the three-dimensional distribution map of the geothermal field, and the thermal reservoir level, spatial range information for indicating the exploration target area is generated and output.

[0013] Preferably, a geothermal resource prediction method based on multi-source fusion data further includes closed-loop optimization of the geothermal resource prediction model, specifically including: Collect new exploration data; The new exploration data is processed and features are extracted using a unified spatial benchmark, and then integrated into the original set of geothermal reservoir feature parameters and training label data to update the training dataset. The geothermal resource prediction model was retrained using the updated training dataset, and the model performance was evaluated. When the model performance improves, the improved model will be applied to subsequent geothermal resource predictions.

[0014] On the other hand, the present invention also proposes a geothermal resource prediction system based on multi-source fusion data, used to implement the above-mentioned geothermal resource prediction method based on multi-source fusion data, comprising: The data acquisition module is used to acquire multi-source geoscience data; The data processing module is used to perform unified spatial benchmark processing on the multi-source geoscience data to obtain preprocessed data; The feature extraction module is used to extract features from the preprocessed data to obtain a set of geothermal reservoir feature parameters; The model building module is used to build a geothermal resource prediction model; the geothermal resource prediction model selects the corresponding machine learning or deep learning algorithm according to the data structure of the geothermal reservoir characteristic parameter set; The model training module trains the geothermal resource prediction model based on the set of geothermal reservoir characteristic parameters and training samples labeled with known well data. The result output module inputs the set of geothermal reservoir characteristic parameters obtained after processing the area to be predicted into the trained geothermal resource prediction model and outputs the quantitative prediction results of geothermal resources. The three-dimensional display module is used to visualize the quantitative prediction results in three dimensions. The three-dimensional visualization includes overlaying and rendering the thermal reservoir probability map, the three-dimensional distribution map of the geothermal field, and the geographic geological environment data. It also supports users to perform rotation, scaling, and sectioning operations through the human-computer interaction interface, and outputs an exploration target area prediction report containing the spatial range information.

[0015] As described above, this invention discloses a geothermal resource prediction method and system based on multi-source fusion data. By performing unified spatial benchmark registration and feature extraction on heterogeneous multi-source data such as geological, geophysical, geochemical, remote sensing, and wellbore data, a geothermal resource prediction model capable of adaptively selecting algorithm models is constructed. The final output includes quantitative prediction results encompassing reservoir probability, three-dimensional geothermal field distribution, and reservoir level, and generates exploration target area suggestions. This invention achieves deep collaborative utilization and dynamic closed-loop optimization of multi-source data, significantly improving the accuracy, reliability, and exploration decision-making efficiency of geothermal resource prediction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a flowchart of a geothermal resource prediction method based on multi-source fusion data. Figure 2 This is a framework diagram of a geothermal resource prediction system based on multi-source fusion data. Detailed Implementation

[0018] 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.

[0019] On the one hand, embodiments of the present invention propose a geothermal resource prediction method based on multi-source fusion data, such as... Figure 1 As shown, it includes the following steps: S1. Acquire multi-source geoscientific data, including geological data, geophysical data, geochemical data, remote sensing data, and existing well data.

[0020] The acquisition of geological data depends on existing geological maps of the study area and core data obtained from drilling projects. Geological maps should include information on lithological distribution, fault structure orientation, and stratigraphic age. Core data should record key parameters such as lithology, porosity, and permeability, which directly reflect the storage capacity and fluid migration characteristics of underground rock strata.

[0021] Geophysical data are collected using gravity measurement equipment, magnetic measurement equipment, magnetotelluric measurement equipment, and seismic exploration equipment. Gravity data can reveal underground density anomalies, magnetic data is used to identify the distribution of magnetic bodies, magnetotelluric data reflects the underground electrical structure, and seismic exploration data provides stratigraphic interfaces and velocity information. These physical parameters provide indirect but important evidence for inferring the distribution of geothermal resources.

[0022] Geochemical data are obtained by collecting water and gas samples from hot springs or geothermal springs within the study area. Water samples need to be tested for the types, concentrations, and isotopic composition of water chemical ions, while gas samples need to be analyzed for the types and contents of gas components. These chemical indicators can be used to estimate the temperature of geothermal reservoirs, determine the type of heat source, and identify groundwater circulation characteristics.

[0023] The remote sensing data was acquired using thermal infrared remote sensing satellites. The surface temperature anomaly information was retrieved and the linear structure distribution was interpreted through remote sensing data processing technology. The surface temperature anomaly may indicate underground thermal activity, while the linear structure may be related to deep hydrothermal upwelling channels.

[0024] The existing well data collection includes drilling logs of drilled wells in the research area. Temperature measurement data at different depths of the wells are extracted, and production capacity data, including water output and water temperature per unit time, are statistically analyzed. This data is used as training labels for direct model learning.

[0025] S2. Perform unified spatial benchmark processing on multi-source geoscientific data, register all data to the same geographic coordinate system and grid scale, and obtain preprocessed data.

[0026] The geographic coordinate system adopts the national geodetic coordinate system to ensure the consistency of spatial reference. The grid scale is determined according to the area size of the study area and the resolution of multi-source geoscientific data. Generally, a moderate scale is selected to balance computational efficiency and detail preservation.

[0027] For raster data, resampling algorithms are used to adjust it to a uniform grid scale. Commonly used algorithms include nearest neighbor method, bilinear interpolation or cubic convolution method. During the resampling process, the spatial distribution characteristics of the data should be kept unchanged to avoid introducing distortion.

[0028] For vector data such as fault lines and stratigraphic boundaries in geological maps, coordinate transformation is performed. Coordinate transformation algorithms are used to convert their coordinates to a unified geographic coordinate system. At the same time, topological checks are performed on the vector data to correct topological errors such as overlaps and gaps, so as to ensure the accuracy of spatial relationships.

[0029] After data registration, the root mean square error needs to be calculated to verify the accuracy. If the error exceeds the allowable range, the registration parameters are readjusted until the accuracy meets the requirements. This step is the key to ensuring the spatial consistency of multi-source data.

[0030] S3. Perform feature extraction on the preprocessed data to obtain a set of geothermal reservoir characteristic parameters.

[0031] The characteristic parameters extracted from geological data include the division of the range based on lithological distribution and the statistical analysis of the area proportion. Lithological distribution affects the lithological combination and reservoir performance of the geothermal reservoir. The length, density and extension depth are calculated based on the strike of the fault structure. The fault structure can serve as a channel for hydrothermal migration. The burial depth and thickness are determined in combination with the stratigraphic age. These parameters together control the spatial distribution and thermal occurrence conditions of the geothermal reservoir.

[0032] Feature parameters extracted from geophysical data include calculating gravity gradient values ​​and determining the location and range of gradient zones from gravity data, which are often related to tectonic boundaries; inverting the distribution of magnetic bodies from magnetic data, which may reflect igneous rocks or alteration zones; inverting the resistivity distribution of subsurface media from magnetotelluric data and extracting parameters for low-resistivity zones, which may correspond to hydrothermal alteration or fluid enrichment zones; and interpreting stratigraphic interfaces and extracting velocity parameters from seismic exploration data, where velocity anomalies can indicate lithological changes or fracture development zones.

[0033] Characteristic parameters extracted from geochemical data include calculating silicon and sodium-potassium thermometers based on water sample ion concentrations to estimate reservoir temperatures. Geothermal thermometers are a reliable method based on the water-rock balance theory. Isotopic composition can be used to determine the source of groundwater recharge and circulation depth. Isotopes such as oxo-18 and deuterium can reveal the source and evolution history of water bodies. Gas sample composition can be used to determine the type of heat source. For example, carbon dioxide and helium isotopes can distinguish between magmatic and metamorphic heat sources.

[0034] The feature parameters extracted from remote sensing data include determining the location, temperature value, and area of ​​the anomaly zone based on surface temperature anomaly information. Surface thermal anomalies are a direct manifestation of underground thermal activity. The density and number of intersection points are calculated based on the distribution of linear structures. Areas with dense linear structures may be favorable locations for the upwelling of deep heat flow.

[0035] The characteristic parameters extracted from existing well data include geothermal gradient calculated based on temperature measurement data at different depths. Geothermal gradient is an important indicator for measuring geothermal background. Production capacity per unit depth is calculated based on production capacity data. This indicator reflects the production capacity of the thermal reservoir.

[0036] S4. Construct a geothermal resource prediction model; the geothermal resource prediction model selects the corresponding machine learning or deep learning algorithm based on the data structure of the geothermal reservoir characteristic parameter set.

[0037] Specifically, if the data is in the form of a grid containing spatial neighborhood information, then a convolutional neural network is chosen, as it can effectively capture spatially related features. If the data contains complex topological relationships, then a graph neural network is chosen, as it is suitable for processing graph-structured data such as fractured networks. If the data is in the form of feature vectors and nonlinear relationships need to be processed, then a random forest is chosen, as it has good interpretability and resistance to overfitting.

[0038] The specific criteria and procedures for selecting the appropriate algorithm based on the type of multi-source geoscientific data are as follows: Data format determination: If the data is in grid form (such as remote sensing imagery, geothermal field raster data) and contains spatial neighborhood information (such as temperature correlation between adjacent grids), it is determined to "contain spatial neighborhood information" and a convolutional neural network (CNN) is selected. Topological relationship determination: If the data is not in grid form, but has a node-edge relationship structure (such as a fault structure network or a stratigraphic contact relationship map), it is determined to "contain complex topological relationships" and a graph neural network (GNN) is selected. Nonlinear relationship determination: If the data is in the form of feature vectors (such as geochemical ion concentration, geothermal gradient values), determine the relationship by plotting a "feature-label" scatter plot: If the scatter points are distributed as a curve (non-linear fitting, R² < 0.6), it is determined that "nonlinear relationship needs to be processed", and random forest is selected.

[0039] S5. Based on the set of geothermal reservoir characteristic parameters and training samples labeled with known well data, train a geothermal resource prediction model.

[0040] The extracted feature parameters are correlated with the temperature measurement data and production capacity data in the existing wellbore data to form a sample set. The sample set is divided into a training set, a validation set and a test set according to a preset ratio. When dividing, it is ensured that the training set can provide sufficient learning information, usually using a ratio of 7:2:1.

[0041] The hyperparameters of the model are tuned using grid search or Bayesian optimization. The tuning aims to minimize the prediction error on the validation set. The hyperparameters that need to be adjusted include the size of the convolution kernel and the number of decision trees, which directly affect the model's complexity and generalization ability.

[0042] The optimized model is trained using the training set. During training, the model's weight parameters are updated using gradient descent. The loss value between the predicted value and the training label is calculated through backpropagation. The optimization is iteratively performed until the loss value converges or the preset number of iterations is reached.

[0043] The model performance is evaluated using a test set. Evaluation metrics include mean absolute error and coefficient of determination. Mean absolute error measures the average deviation of the predicted values, while the coefficient of determination reflects the proportion of variance explained by the model. If the evaluation metrics do not meet the requirements, the algorithm or hyperparameters are readjusted, and the training and evaluation are repeated until the model performance meets the requirements.

[0044] S6. Input the set of geothermal reservoir characteristic parameters obtained after processing the area to be predicted into the trained geothermal resource prediction model, and output the quantitative prediction results of geothermal resources.

[0045] The quantitative prediction results of geothermal resources include a geothermal reservoir probability map, a three-dimensional geothermal field distribution map, and estimated geothermal reservoir temperature and reservoir capacity level. Based on the thermal reservoir probability map, the three-dimensional distribution map of the geothermal field, and the thermal reservoir level, spatial range information for indicating the exploration target area is generated and output.

[0046] When generating the thermal reservoir probability map, the probability values ​​of thermal reservoirs in the grid cells output by the model are divided into intervals, corresponding to the thermal reservoir potential level. The distribution is displayed by color rendering and the area is counted. The higher the probability value, the greater the possibility of thermal reservoirs.

[0047] When constructing a three-dimensional distribution map of the geothermal field, the predicted temperature values ​​at different depths are used as a basis, combined with the stratum depth data, and spatial interpolation algorithms such as Kriging or inverse distance weighting are used to expand it into a three-dimensional data volume, thereby comprehensively displaying the distribution of geothermal temperature in three-dimensional space.

[0048] When calculating the estimated reservoir temperature, the average temperature within the reservoir depth range is taken from the three-dimensional geothermal field distribution map. If a temperature gradient exists, the temperature is calculated segment by segment and then averaged to accurately represent the thermal state of the reservoir. When classifying reservoir levels, the estimated reservoir temperature, reservoir thickness, reservoir lithology thermal conductivity, and reservoir area are used to calculate the value using the reservoir calculation formula. The levels are then classified according to industry standards. The reservoir calculation formula generally uses the volumetric method to consider the heat capacity of the rock and fluids. When analyzing the prediction results, areas with high reservoir probability, high temperature, and high reserve level are identified, and their central location and boundary range are determined. This provides preliminary target area suggestions for exploration and improves the exploration success rate.

[0049] This embodiment also includes uncertainty analysis to assess the reliability of the prediction results. Sources of uncertainty include differences in accuracy of multi-source geoscientific data, feature extraction errors, inherent errors in the fusion model algorithm, and biases in the training label data. For each source of uncertainty, an assessment is conducted, data measurement error reports are collected, and the error range is statistically analyzed to determine its impact on feature extraction. For example, errors in geophysical data acquisition may lead to uncertainty in the inversion results. Feature parameters are repeatedly extracted, and the standard deviation is calculated. The stability of feature extraction is assessed using the standard deviation; a larger standard deviation indicates less stable feature extraction. Cross-validation is used to calculate the dispersion of multiple sets of prediction results. The uncertainty of the model algorithm is assessed using the dispersion; a large dispersion indicates large fluctuations in model predictions. The acquisition methods for wellbore temperature data are compared to examine the impact of biases in the training label data; for example, differences in the accuracy of downhole thermometers may lead to label errors. Monte Carlo simulation is used for uncertainty analysis. Based on the error distribution, multiple sets of random error parameters are generated. These random error parameters are substituted into the model for multiple predictions. The mean, standard deviation, and confidence intervals of the multiple prediction results are statistically analyzed to quantify the uncertainty of the prediction. Output uncertainty analysis results, display the mean, standard deviation and confidence interval of the prediction results in chart form, mark the high confidence area and low confidence area in the prediction results, generate an assessment report, analyze the contribution of each uncertainty source, provide risk reference for exploration decision-making, prioritize high confidence and high potential areas, and reduce exploration risk.

[0050] In the Monte Carlo simulation, the error distribution assumptions are as follows: data measurement errors and feature extraction errors adopt a normal distribution, specifically determined based on the mean and standard deviation of the data measurement reports. The bias of the training label data adopts a uniform distribution, specifically determined based on the accuracy range of the temperature measurement equipment.

[0051] Furthermore, this embodiment also includes closed-loop optimization of the geothermal resource prediction model, specifically including: Collect new exploration data, including temperature data, core data, and production data from new wells in high-potential areas, as well as supplementary geophysical data, geochemical data, and remote sensing updates. The new data provides additional information for model optimization.

[0052] The new data is organized according to the data acquisition requirements, ensuring that its type and format are consistent with the existing data to guarantee data consistency. Then, the new data is registered according to the unified spatial reference processing procedure, integrating the registered new data into the existing data system and updating the entire dataset.

[0053] Update the training dataset by adding the feature parameters extracted from the new data and their corresponding labels to the original sample set. Re-divide the training, validation, and test sets according to the original sample set's partitioning ratio to maintain the rationality of the data partitioning. Retrain the original model using the updated training set, adjusting the model's hyperparameters using the original parameter tuning methods to adapt the model to the new data.

[0054] By comparing the mean absolute error and coefficient of determination of the model on the test set before and after training, we can determine whether the model performance has improved. If the model performance has improved, the new model will be used for subsequent geothermal resource prediction. If the model performance has not improved, we will analyze the reasons for the lack of improvement, check the data quality or adjust the algorithm and train again until the model performance improves, thus achieving continuous optimization of the model.

[0055] On the other hand, reference Figure 2 The present invention also proposes a geothermal resource prediction system based on multi-source fusion data, used to implement the above-mentioned geothermal resource prediction method based on multi-source fusion data, comprising: The data acquisition module is used to acquire multi-source geoscientific data. This module receives data uploaded by users through a human-computer interaction interface, verifies the format, stores the data according to type, and creates an index. It also supports the uploading and updating of new data.

[0056] The data processing module performs unified spatial benchmarking on multi-source geoscientific data to obtain preprocessed data. This module calls the data stored in the data acquisition module, performs coordinate transformation, resampling, and topology checking operations, and then transmits the processed preprocessed data to the feature extraction module.

[0057] The feature extraction module is used to extract features from the preprocessed data to obtain a set of geothermal reservoir feature parameters. Based on the preprocessed data, this module extracts various features through numerical calculation and inversion algorithms, and then transmits the obtained geothermal reservoir feature parameter set to the model building module and the model training module.

[0058] The model building module is used to build a geothermal resource prediction model; the geothermal resource prediction model selects the corresponding machine learning or deep learning algorithm based on the data structure of the geothermal reservoir characteristic parameter set; The model training module trains a geothermal resource prediction model based on a set of geothermal reservoir characteristic parameters and training samples labeled with known well data. During the training process, the gradient descent method is used to update the weight parameters and iteratively optimize until the termination condition is met.

[0059] The results output module takes the set of geothermal reservoir characteristic parameters obtained after processing the area to be predicted and inputs it into the trained geothermal resource prediction model, outputting the quantitative prediction results of geothermal resources. The quantitative prediction results include a reservoir probability map, a three-dimensional distribution map of the geothermal field, and the estimated reservoir temperature and reservoir quantity level.

[0060] The 3D visualization module is used to visualize the quantitative prediction results in 3D. The 3D visualization includes overlaying and rendering the thermal reservoir probability map, the 3D distribution map of the geothermal field, and the geographic geological environment data. It also supports users to perform rotation, scaling, and sectioning operations through the human-computer interaction interface, and outputs exploration target area prediction reports containing spatial range information.

[0061] The interaction process of each module is as follows: The data acquisition module receives multi-source geoscientific data uploaded by users, classifies and stores it after verification; the data processing module calls the stored data, performs unified spatial benchmark processing, and then transmits it to the feature extraction module; the feature extraction module extracts a set of geothermal reservoir characteristic parameters based on the preprocessed data and transmits it to the model building module and the model training module; the model building module builds a geothermal resource prediction model according to the data structure of the feature parameter set; the model training module trains the model using the feature parameter set and training samples; the result output module outputs quantitative prediction results using the trained model; and the 3D display module provides 3D visualization and interactive operation for the prediction results. The entire system achieves full-process automation of geothermal resource prediction through modular design, improving prediction efficiency and accuracy.

[0062] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0063] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for geothermal resource prediction based on multi-source fusion data, characterized in that, The method comprises the following steps: obtaining multi-source geology data; performing unified spatial reference processing on the multi-source geology data to obtain preprocessed data; performing feature extraction on the preprocessed data to obtain a set of geothermal reservoir formation feature parameters; constructing a geothermal resource prediction model; the geothermal resource prediction model selects a corresponding machine learning or deep learning algorithm according to the data structure of the set of geothermal reservoir formation feature parameters; training the geothermal resource prediction model based on the set of geothermal reservoir formation feature parameters and training samples with known well data as labels; inputting the set of geothermal reservoir formation feature parameters obtained after processing the area to be predicted into the trained geothermal resource prediction model to output a quantitative prediction result of geothermal resources. 2.The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, The multi-source geology data includes geological data, geophysical data, geochemical data, remote sensing data, and existing well data. 3.The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, The spatial reference of the multi-source geology data is unified to obtain preprocessed data, including the following data registration steps: determining a unified geographic coordinate system and grid scale; adjusting grid type data to the unified grid scale using a resampling algorithm; performing coordinate conversion processing on vector type data to the unified geographic coordinate system, and performing topology checking and correction; calculating the root mean square error after data registration to verify the accuracy, and repeating the data registration steps until the accuracy meets the requirements if the error exceeds the allowed accuracy range.

4. The geothermal resource prediction method based on multi-source fusion data according to claim 2, characterized in that, Performing feature extraction on the preprocessed data to obtain a set of geothermal reservoir formation feature parameters, including: extracting lithology area proportion, fault structure density, and stratum burial depth from the geological data; extracting gravity gradient, magnetic body distribution, resistivity distribution, and seismic velocity parameters from the geophysical data; extracting geothermal reservoir temperature based on water sample ion concentration, groundwater recharge source based on isotope composition, and heat source type based on gas sample composition from the geochemical data; extracting surface temperature anomaly area position and area, linear structure density, and intersection point number from the remote sensing data; extracting geothermal gradient and unit depth productivity index from the existing well data.

5. The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, The geothermal resource prediction model selects a corresponding machine learning or deep learning algorithm according to the data structure of the set of geothermal reservoir formation feature parameters, including: if the data structure is a grid form containing spatial neighborhood information, a convolutional neural network is selected; if the data structure contains complex topological relationships, a graph neural network is selected; if the data structure is a feature vector form and needs to process nonlinear relationships, a random forest is selected.

6. The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, Training the geothermal resource prediction model includes: associating the set of geothermal reservoir formation feature parameters with corresponding label data to form a sample set, and dividing it into a training set, a validation set, and a test set according to a predetermined proportion; using a grid search method or a Bayesian optimization method to optimize the hyperparameters of the geothermal resource prediction model; using the training set to train the geothermal resource prediction model, updating the weight parameters through gradient descent method, and iteratively optimizing until the termination condition is reached; using the test set to evaluate the performance of the geothermal resource prediction model, and the evaluation indicators include mean absolute error and determination coefficient.

7. The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, The quantitative prediction result of the geothermal resource includes a geothermal reservoir probability map, a three-dimensional distribution map of the geothermal field, an estimated geothermal reservoir temperature, and a geothermal reservoir volume grade. Based on the geothermal reservoir probability map, the three-dimensional distribution map of the geothermal field, and the geothermal reservoir volume grade, spatial range information indicating an exploration target area is generated and output. 8.The geothermal resource prediction method based on multi-source fusion data according to claim 1, characterized in that, The method also includes closed-loop optimization of the geothermal resource prediction model, specifically including: Collecting new exploration data; After the new exploration data is processed by unified spatial reference and feature extraction, the original geothermal reservoir forming feature parameter set and training label data are fused to update the training data set; The geothermal resource prediction model is retrained using the updated training data set, and the model performance is evaluated; When the model performance improves, the model with improved performance is applied to subsequent geothermal resource prediction.

9. A geothermal resource prediction system based on multi-source fusion data, used to implement a geothermal resource prediction method based on multi-source fusion data according to any one of claims 1-8, characterized in that, The method includes: A data acquisition module for acquiring multi-source geoscience data; A data processing module for performing unified spatial reference processing on the multi-source geoscience data to obtain preprocessed data; A feature extraction module for extracting features from the preprocessed data to obtain a geothermal reservoir forming feature parameter set; A model construction module for constructing a geothermal resource prediction model; the geothermal resource prediction model selects a corresponding machine learning or deep learning algorithm according to the data structure of the geothermal reservoir forming feature parameter set; A model training module for training the geothermal resource prediction model based on the geothermal reservoir forming feature parameter set and training samples with known well data as labels; A result output module for inputting the geothermal reservoir forming feature parameter set obtained after processing the area to be predicted into the trained geothermal resource prediction model and outputting a quantitative prediction result of the geothermal resource.

10. The geothermal resource prediction system based on multi-source fusion data according to claim 9, characterized in that, A three-dimensional display module for three-dimensional visualization of the quantitative prediction result, including superimposing and fusing rendering of the geothermal reservoir probability map, the three-dimensional distribution map of the geothermal field, and the geographic and geological environment data, supporting user rotation, scaling, and sectioning operations through a human-computer interaction interface, and outputting an exploration target area prediction report containing the spatial range information.

Images