Prediction method for ground subsidence caused by crustal stress change in geothermal exploitation

Abstract

The invention provides a method for predicting ground subsidence caused by ground stress change in geothermal exploitation, which comprises the following steps: collecting geological survey data of a target geothermal exploitation area, and constructing a three-dimensional geomechanical model; based on the constructed three-dimensional geomechanical model, simulating crustal stress evolution processes in different mining stages by adopting a numerical simulation method to obtain crustal stress tensor time sequence change data; inputting the crustal stress tensor time sequence change data into a three-dimensional geomechanical model as a boundary condition, and calculating the deformation of each rock stratum in the vertical direction to obtain a ground surface settlement estimated value; constructing a crustal stress-settlement prediction model based on the nonlinear mapping relationship between the crustal stress tensor time sequence change data and the settlement estimated value, and training; and inputting subsequent terrestrial heat exploitation scheme parameters, and predicting a future ground surface settlement trend by using the trained ground stress-settlement prediction model. The method is wide in application range and can be suitable for ground subsidence prediction of the geothermal exploitation area under different geological conditions.

Description

Technical Field

[0001] The present invention belongs to the field of geological prediction, and in particular relates to a method for predicting ground subsidence caused by ground stress changes in geothermal mining. Background Art

[0002] With the continuous growth of global energy demand and increasing environmental awareness, the development and utilization of geothermal resources, as a clean, renewable energy source, has attracted widespread attention. However, during geothermal extraction, the extraction of hot water from underground causes changes in the pore pressure of the geothermal reservoir, which in turn leads to a redistribution of geostress. This change in geostress causes deformation in the surrounding rock formations. When this deformation is transmitted to the surface, it can trigger geological disasters such as ground subsidence.

[0003] Land subsidence not only damages surface infrastructure like buildings, roads, and bridges, but can also lead to a range of environmental problems, including falling groundwater levels and seawater intrusion, severely impacting people's lives and the ecological environment. Therefore, accurately predicting land subsidence caused by changes in geostress during geothermal extraction is crucial for developing rational extraction plans, implementing effective prevention and control measures, and ensuring the sustainable development of geothermal resources.

[0004] Currently, the main methods for predicting ground subsidence caused by geothermal extraction include empirical formulas and numerical simulation. The empirical formula method, based on previous engineering experience, uses regression analysis to determine the relationship between subsidence and parameters such as extraction volume. However, its accuracy is low and its applicability is limited. Numerical simulation can account for the complex structure and mechanical properties of geological bodies and simulate changes in ground stress and ground subsidence. However, traditional numerical simulation methods are computationally intensive and time-consuming, and they are unable to quickly respond to the impact of subsequent adjustments to extraction plans on ground subsidence.

[0005] Based on the above situation, there is an urgent need for a method that can accurately and efficiently predict ground subsidence caused by changes in ground stress during geothermal mining. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the existing geothermal mining ground subsidence prediction methods, such as low accuracy, limited scope of application, large amount of calculation and difficulty in quickly responding to mining plan adjustments, and to provide a prediction method for ground subsidence caused by ground stress changes in geothermal mining.

[0007] The technical solutions of the present invention are as follows:

[0008] A method for predicting ground subsidence caused by ground stress changes during geothermal mining, the method comprising:

[0009] Collect geological survey data of the target geothermal mining area and construct a three-dimensional geomechanical model;

[0010] Based on the constructed three-dimensional geomechanical model, numerical simulation methods were used to simulate the evolution of geostress under different mining stages and obtain the temporal variation data of geostress tensor.

[0011] The temporal variation data of the ground stress tensor are input into a three-dimensional geomechanical model as boundary conditions, and the deformation of each rock layer in the vertical direction is calculated to obtain an estimated value of surface settlement;

[0012] Based on the nonlinear mapping relationship between the time series change data of the ground stress tensor and the estimated settlement value, a ground stress-settlement prediction model is constructed and trained;

[0013] Input the parameters of the subsequent geothermal mining plan and use the trained ground stress-settlement prediction model to predict future surface settlement trends.

[0014] Furthermore, the geological survey data includes stratum lithology parameters, groundwater level distribution, initial pore pressure and geothermal reservoir temperature field data of the target area.

[0015] Furthermore, the three-dimensional geomechanical model includes a stratum structure module, a mechanical parameter module and a boundary condition module;

[0016] Among them, the stratum structure module constructs the spatial distribution morphology of different rock layers based on the stratum lithologic parameters, the mechanical parameter module integrates the physical and mechanical properties of each rock layer, and the boundary condition module is used to set the displacement constraints and stress constraints of the model.

[0017] Furthermore, the physical and mechanical properties include elastic modulus, Poisson's ratio, compressive strength, porosity and permeability.

[0018] Furthermore, the numerical simulation method is a finite difference method or a finite element method.

[0019] Furthermore, the temporal variation data of the ground stress tensor are input into the three-dimensional geomechanical model as boundary conditions, and the deformation of each rock layer in the vertical direction is calculated to obtain the estimated value of the surface settlement:

[0020] Based on the ground stress tensor data at different mining stages, the spatial coordinates are matched to the corresponding rock units in the three-dimensional geomechanical model, the stress boundary conditions of each rock unit are set, and the vertical displacement of each rock unit is calculated based on the stress-strain relationship;

[0021] The vertical deformation of each rock layer is accumulated to the surface using the settlement integration method to obtain the estimated value of surface settlement evolving with time.

[0022] Furthermore, the vertical displacement of each rock layer unit is calculated by the stress-strain relationship as follows:

[0023] The stress-strain relationship is established using the generalized Hooke's law to calculate the strain tensor of each rock unit;

[0024] The vertical strain component in the strain tensor is extracted and combined with the rock layer thickness to calculate the vertical displacement.

[0025] Furthermore, a geostress-settlement prediction model is constructed based on the nonlinear mapping relationship between the temporal variation data of the geostress tensor and the estimated settlement value, and the training is carried out as follows:

[0026] First, the acquired time series change data of the ground stress tensor are preprocessed to obtain the ground stress characteristic matrix; at the same time, the estimated surface settlement values ​​are sorted according to the corresponding time series to form a settlement label set;

[0027] Build a deep learning model based on long short-term memory network, including input layer, two layers of LSTM units, and fully connected output layer;

[0028] The ground stress feature matrix is ​​used as the input sequence of the model, the settlement label set is used as the output sequence, the mean square error is used as the loss function, and the Adam optimizer is used for training to obtain the ground stress-settlement prediction model.

[0029] Compared with the prior art, the present invention has the following advantages:

[0030] By constructing a three-dimensional geomechanical model, the present invention can truly reflect the geological structure and mechanical properties of the target geothermal mining area, providing a reliable basis for the simulation of the geostress evolution process and the calculation of surface subsidence.

[0031] The numerical simulation method is used to obtain the time series change data of the ground stress tensor, and the surface settlement estimate is calculated by combining the stress-strain relationship and the settlement integral method, which improves the accuracy of ground settlement prediction.

[0032] A geostress-subsidence prediction model is constructed based on the long short-term memory network, which can capture the nonlinear mapping relationship between the time-series change data of the geostress tensor and the settlement estimation value. The trained model can quickly respond to the input of subsequent mining plan parameters and efficiently predict future surface subsidence trends, providing a timely and accurate reference basis for the safe and efficient development of geothermal resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings illustrate various embodiments generally by way of example and not limitation, and together with the description and claims, serve to explain embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive of the embodiments of the present apparatus or method.

[0034] Figure 1A schematic flow chart of the method steps of the present invention is shown. DETAILED DESCRIPTION

[0035] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0036] like Figure 1 As shown, the present invention provides a method for predicting ground subsidence caused by ground stress changes in geothermal mining, comprising:

[0037] Collect geological survey data of the target geothermal mining area and construct a three-dimensional geomechanical model;

[0038] Geological survey data includes:

[0039] Stratigraphic lithologic parameters: including stratum distribution depth, lithologic type (sandstone, mudstone, tuff, etc.), and layer thickness;

[0040] Groundwater level distribution: Determine the water level changes at different locations through dynamic observation holes;

[0041] Initial pore pressure: converted according to formation depth and water head pressure;

[0042] Geothermal reservoir temperature field: 3D temperature distribution obtained through multi-depth thermocouples and thermal response tests;

[0043] Rock physical and mechanical parameters: elastic modulus, Poisson's ratio, compressive strength, porosity and permeability, etc. obtained through laboratory tests.

[0044] Construct a three-dimensional geomechanical model;

[0045] Use FLAC3D finite difference software or ABAQUS finite element analysis platform to build a three-dimensional geomechanical model. The model includes the following modules:

[0046] Stratigraphic structure module: Based on geological profile data, a multi-layer structural model is established to distinguish reservoirs, caprocks, bedrock, etc.

[0047] Mechanical parameter module: assign corresponding elastic modulus, Poisson's ratio, permeability, density and other properties to each rock layer;

[0048] Boundary condition module: Set horizontal constraints on the sides of the model, set fixed constraints at the bottom, set the top as a free surface, apply gravity loads and consider the influence of geothermal gradients.

[0049] The model area is divided into three-dimensional grids, and the reservoir and surrounding areas are divided using hexahedral or tetrahedral elements. Each unit body is bound to the corresponding lithology and mechanical properties to ensure simulation accuracy.

[0050] Simulation of geostress evolution

[0051] This example uses the finite element method (FEM) to simulate geostress. The simulation process consists of multiple stages, corresponding to the initial, mid-stage, stable, and final stages of geothermal fluid extraction. For each stage, the corresponding extraction volume, extraction rate, and reservoir pressure changes are input as loading conditions.

[0052] The output includes the stress tensor (principal stresses σ1, σ2, σ3, shear stress τ), strain tensor, and pore pressure change of each grid cell at each time step.

[0053] Calculation of estimated settlement values

[0054] At each simulation stage, the stress tensor data of each element of the model is extracted, and the stress-strain relationship is constructed in combination with the generalized Hooke's law (applicable to isotropic elastic bodies) to calculate the strain tensor:

[0055] Extract the vertical strain component and calculate the vertical displacement of each unit based on the rock layer thickness;

[0056] The estimated surface settlement is calculated by integrating the vertical deformation of each rock layer from the bottom to the surface:

[0057] Build and train a ground stress-settlement prediction model

[0058] (1) Data preprocessing

[0059] Construction of the Geostress Feature Matrix: The acquired geostress tensor time series data undergoes preprocessing, including data cleaning, normalization, and feature extraction. First, outliers and missing values ​​are removed, and missing values ​​are filled using linear interpolation. Then, the three principal stress values ​​of the geostress tensor are normalized to fall within the range [0, 1]. Finally, principal component analysis is used to extract features from the normalized geostress data, reducing the high-dimensional geostress tensor data to 10 key features to construct the geostress feature matrix.

[0060] Construction of the settlement label set: The surface settlement estimates were sorted into corresponding time series to form a settlement label set. To improve the generalization ability of the model, the settlement data was standardized to a mean of 0 and a standard deviation of 1.

[0061] (2) Model construction

[0062] Build a deep learning model based on the Long Short-Term Memory (LSTM) network and implement it using the TensorFlow framework. The model structure is as follows:

[0063] Input layer: Receives the ground stress feature matrix, with an input dimension of 10 (number of features) × 30 (time steps).

[0064] The first layer of LSTM units: contains 64 neurons, uses the ReLU activation function, and returns a sequence output.

[0065] The second layer of LSTM units: contains 32 neurons, uses the ReLU activation function, and does not return sequence output.

[0066] Fully connected output layer: contains 1 neuron, uses a linear activation function, and outputs the predicted settlement value.

[0067] (3) Model training

[0068] The geostress feature matrix and the settlement label set were divided into training and test sets in a ratio of 7:3. During training, the mean squared error (MSE) was used as the loss function, and the Adam optimizer was used for parameter optimization. The learning rate was initially set to 0.001 and decayed by 0.9 every 10 epochs. The number of training iterations was 100 epochs, and the batch size was set to 32. Early stopping was used to prevent model overfitting. Training was terminated when the validation set loss function stopped decreasing after 10 consecutive epochs.

[0069] After training, the performance of the model was evaluated on the test set, and the mean absolute error (MAE) was 2.5 mm and the root mean square error (RMSE) was 3.8 mm, indicating that the model has high prediction accuracy.

[0070] Input the parameters of the subsequent geothermal mining plan and use the trained ground stress-settlement prediction model to predict future surface settlement trends.

[0071] The above is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, can make equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for predicting ground subsidence caused by ground stress changes during geothermal mining, characterized in that: The method comprises: Collect geological survey data of the target geothermal mining area and construct a three-dimensional geomechanical model; Based on the constructed three-dimensional geomechanical model, numerical simulation methods were used to simulate the evolution of geostress under different mining stages and obtain the temporal variation data of geostress tensor. The temporal variation data of the ground stress tensor are input into a three-dimensional geomechanical model as boundary conditions, and the deformation of each rock layer in the vertical direction is calculated to obtain an estimated value of surface settlement; Based on the nonlinear mapping relationship between the time series change data of the ground stress tensor and the estimated settlement value, a ground stress-settlement prediction model is constructed and trained; Input the parameters of the subsequent geothermal mining plan and use the trained ground stress-settlement prediction model to predict future surface settlement trends.

2. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 1, characterized in that: The geological survey data include stratum lithology parameters, groundwater level distribution, initial pore pressure and geothermal reservoir temperature field data of the target area.

3. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 1, characterized in that: The three-dimensional geomechanical model includes a stratum structure module, a mechanical parameter module and a boundary condition module; Among them, the stratum structure module constructs the spatial distribution morphology of different rock layers based on the stratum lithologic parameters, the mechanical parameter module integrates the physical and mechanical properties of each rock layer, and the boundary condition module is used to set the displacement constraints and stress constraints of the model.

4. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 3, characterized in that: The physical and mechanical properties include elastic modulus, Poisson's ratio, compressive strength, porosity and permeability.

5. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 1, characterized in that: The numerical simulation method is a finite difference method or a finite element method.

6. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 1, characterized in that: The temporal variation data of the ground stress tensor are input into the three-dimensional geomechanical model as boundary conditions, and the deformation of each rock layer in the vertical direction is calculated to obtain the estimated value of the surface settlement: Based on the ground stress tensor data at different mining stages, the spatial coordinates are matched to the corresponding rock units in the three-dimensional geomechanical model, the stress boundary conditions of each rock unit are set, and the vertical displacement of each rock unit is calculated based on the stress-strain relationship; The vertical deformation of each rock layer is accumulated to the surface using the settlement integration method to obtain the estimated value of surface settlement evolving with time.

7. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 6, characterized in that: The vertical displacement of each rock layer unit is calculated by the stress-strain relationship as follows: The stress-strain relationship is established using the generalized Hooke's law to calculate the strain tensor of each rock unit; The vertical strain component in the strain tensor is extracted and combined with the rock layer thickness to calculate the vertical displacement.

8. The method for predicting ground subsidence caused by ground stress changes in geothermal mining according to claim 1, characterized in that: Based on the nonlinear mapping relationship between the time series change data of the ground stress tensor and the estimated settlement value, a ground stress-settlement prediction model is constructed and trained as follows: First, the acquired time series change data of the ground stress tensor are preprocessed to obtain the ground stress characteristic matrix; at the same time, the estimated surface settlement values ​​are sorted according to the corresponding time series to form a settlement label set; Build a deep learning model based on long short-term memory network, including input layer, two layers of LSTM units, and fully connected output layer; The ground stress feature matrix is ​​used as the input sequence of the model, the settlement label set is used as the output sequence, the mean square error is used as the loss function, and the Adam optimizer is used for training to obtain the ground stress-settlement prediction model.

Images