Prediction method for heat conductivity coefficient of marine natural gas hydrate reservoir

Abstract

The invention relates to a method for predicting the heat conductivity coefficient of a marine natural gas hydrate reservoir, and the method comprises the steps: obtaining the reservoir characteristics of a natural gas hydrate reservoir of which the heat conductivity coefficient is to be predicted, the reservoir characteristics comprising environmental characteristics, reservoir physical property characteristics, reservoir composition characteristics and reservoir structure characteristics; inputting the reservoir characteristics into a pre-trained prediction model to obtain a heat conductivity coefficient of the natural gas hydrate reservoir corresponding to the reservoir characteristics; the method specifically comprises the steps that a preset large number of natural gas hydrate reservoir heat conduction coefficient values under different reservoir characteristics are obtained; constructing a data set by using the reservoir heat conduction coefficient values and the corresponding reservoir features, preprocessing the data set, and dividing the data set into a training set and a test set; and extracting feature data of all reservoir features and heat conduction coefficient values in the training set.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration technology, specifically to a method for predicting the thermal conductivity of marine natural gas hydrate reservoirs. Background Technology

[0002] Natural gas hydrate resources are mainly found in deep-sea sedimentary layers and terrestrial permafrost layers, with approximately 27% of the land and 90% of the seabed considered as potential natural gas hydrate reservoirs.

[0003] The effective thermal conductivity of the porous media in marine natural gas hydrate reservoirs directly affects the hydrate extraction efficiency. Currently, there is limited research on the effective thermal conductivity of complex components in marine silty mud reservoirs, and the heat transfer mechanism during phase change processes remains unclear, hindering the optimization of extraction technologies and the selection of appropriate conditions for natural gas hydrate extraction from marine sediments. Summary of the Invention

[0004] This invention provides a method for predicting the thermal conductivity of marine natural gas hydrate reservoirs. Based on machine learning, it solves the problems of cumbersome experimental operations and high time costs in existing methods. It can quickly and efficiently obtain the predicted thermal conductivity of porous media systems in natural gas hydrate reservoirs in different sea areas, greatly improving the prediction accuracy and efficiency.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, this application provides a method for predicting the thermal conductivity of marine natural gas hydrate reservoirs, comprising: To obtain the characteristics of a natural gas hydrate reservoir with a predicted thermal conductivity, the characteristics include environmental characteristics, reservoir physical property characteristics, reservoir composition characteristics, and reservoir structural characteristics; The environmental characteristics include ambient temperature (T) and pressure (P); the reservoir physical characteristics include reservoir sediment material (quartz sand, montmorillonite, illite, etc.), reservoir sediment density (ρ), sediment porosity (φ), and reservoir sediment grain size (r); the reservoir composition characteristics include quartz content, montmorillonite content, and illite content; and the reservoir structural characteristics include component phase, water saturation (S_w), hydrate saturation (S_h), salt concentration (c), hydrate occurrence mode (pore-filling type / granular encapsulation type), and component contact mode. The thermal conductivity of the natural gas hydrate reservoir is input into a pre-trained prediction model to obtain the thermal conductivity (λ) of the natural gas hydrate reservoir under the specified reservoir characteristics.

[0006] The prediction model is trained on a pre-built machine learning model based on the thermal conductivity of natural gas hydrate reservoirs under different reservoir characteristics, specifically including: A large number of thermal conductivity values ​​of natural gas hydrate reservoirs under different reservoir characteristics were obtained, and a dataset was constructed with the corresponding reservoir characteristics. The dataset was preprocessed and divided into a training set and a test set. The thermal conductivity values ​​of natural gas hydrate reservoirs under different reservoir characteristics are calculated from the physical data obtained by thermal conductivity test experiments of natural gas hydrate reservoirs under different reservoir characteristics; The experimental physical data mentioned are the temperature (T), pressure (P), real-time voltage (U), and real-time current (I) of the natural gas hydrate reservoir thermal conductivity process under different reservoir characteristics, obtained by steady-state testing using the natural gas hydrate sediment thermal conductivity measurement experimental system. The aforementioned experimental system for measuring the thermal conductivity of sediments containing natural gas hydrates comprises a high-pressure reactor module, a constant temperature module, an electric heating module, a gas injection module, and a data acquisition module. The high-pressure reactor module is a stainless steel high-pressure reactor containing a long hollow cylinder, with heating tubes distributed along the axis of the cylinder for heating. Each heating tube consists of a 1 mm diameter outer protective shell and a heating wire. The protective shell is made of stainless steel with good thermal conductivity, and the heating wire is platinum wire with a thermal resistance length of (L). The ends of the heating tube are insulated and sealed with epoxy resin. Copper rods are welded to both ends of the heating tube and connected to a DC power supply to provide a stable voltage (U) to the heating wire. Three Pt1000 thermocouples (T1, T2, T3) are installed at the same horizontal plane in the middle of the stainless steel high-pressure reactor, distributed at different radial positions (r1, r2, r3). ​​In steady state, an isothermal zone exists along the axial direction in the central region of the cylinder, within which a stable one-dimensional heat flow (I) is generated radially from the cylinder axis to the outer wall of the cylinder. A data acquisition module is used to record physical data such as pressure (P), temperature (T), real-time voltage (U), and real-time current (I) during the experiment.

[0007] A calculation module was established, which uses a Python program to calculate the thermal conductivity (λ) value of natural gas hydrate reservoirs under different reservoir characteristics from the physical data.

[0008] The dataset preprocessing includes outlier handling, missing value handling, and normalization. The normalization process unifies the reservoir characteristics and thermal conductivity values ​​in numerical form. Feature data of all reservoir features and thermal conductivity values ​​in the training set are extracted. Based on the feature data, machine learning is used to perform hierarchical modeling of the dataset, and the obtained hierarchical prediction model is trained to obtain the prediction model. The hierarchical modeling includes Bagging, Random Forest, Decision Tree, Support Vector Regression, Gradient Boosting Decision Tree, and K-Nearest Neighbor.

[0009] The pre-test model is evaluated using the evaluation metrics and the test set, and a better pre-test model is selected. The evaluation metrics include mean absolute error (MAE), root mean square error (RMSE), root mean square error (MASE), goodness of fit (R2), and mean absolute percentage error (MAPE).

[0010] The interpretability analysis of the predicted results was performed using SHapley Additive exPlanations (SHAP) to verify the reliability of the prediction model and analyze the sensitivity of the input reservoir features.

[0011] This invention provides a method for predicting the thermal conductivity of marine natural gas hydrate reservoirs based on machine learning. The method involves acquiring the characteristics of the natural gas hydrate reservoir whose thermal conductivity is to be predicted, and then inputting these characteristics into a pre-trained prediction model to obtain the thermal conductivity value of the natural gas hydrate reservoir. The prediction model is trained on a pre-built machine learning model using a dataset constructed based on the reservoir characteristics and corresponding thermal conductivity values. The reservoir characteristics include environmental features, reservoir physical properties, reservoir composition features, and reservoir structural features. The corresponding thermal conductivity values ​​are derived from experimental testing and calculation.

[0012] This invention solves the problems of difficult measurement, numerous influencing factors, and difficulty in prediction of the thermal conductivity of natural gas hydrate reservoirs. It provides a method for predicting the thermal conductivity of natural gas hydrate reservoirs based on machine learning for different marine natural gas hydrate reservoirs, and also provides a technical method for real-time estimation of reservoir thermal conductivity during the field exploitation of marine natural gas hydrate reservoirs. Attached Figure Description

[0013] Figure 1 is a flowchart of the method provided in an embodiment of this application; Figure 2 is a flowchart for establishing a reservoir thermal conductivity dataset; Figure 3 shows the fitting effect between the measured and predicted thermal conductivity results of the KNN model in step 2.3 of the present invention. Figure 4 shows the fitting effect between the thermal conductivity measurement results and the prediction results of the DT model in step 2.3 of the embodiment of the present invention; Figure 5 shows the fitting effect between the measured and predicted thermal conductivity results of the Bagging model in step 2.3 of the present invention. Figure 6 shows the fitting effect between the measured and predicted thermal conductivity results of the SVR model in step 2.3 of the present invention. Figure 7 shows the fitting effect between the measured and predicted thermal conductivity results of the RF model in step 2.3 of the present invention; Figure 8 shows the fitting effect between the measured and predicted thermal conductivity results of the GBDT model in step 2.3 of the present invention. Figure 9 shows the interpretability analysis results of the GBDT model SHAP in step 4 of the embodiment of the present invention; Figure 10 shows the sensitivity analysis results of the GBDT model in step 4 of the embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0015] To address the shortcomings and problems of existing technologies, this application provides a method for predicting the thermal conductivity of marine natural gas hydrate reservoirs, including: To obtain the reservoir characteristics of a natural gas hydrate reservoir with a predicted thermal conductivity, the reservoir characteristics include environmental characteristics, reservoir physical property characteristics, reservoir composition characteristics, and reservoir structural characteristics; The reservoir characteristics are input into a pre-trained prediction model to obtain the thermal conductivity of the natural gas hydrate reservoir corresponding to those characteristics.

[0016] The above method is described below in a more detailed embodiment with reference to more accompanying drawings.

[0017] Referring to Figure 1, this invention is a flowchart of a method for predicting the thermal conductivity of natural gas hydrate reservoirs based on machine learning, comprising the following steps: Step 1: Obtain the reservoir characteristics of the natural gas hydrate with the predicted thermal conductivity. These reservoir characteristics include environmental characteristics, reservoir physical properties, reservoir composition, and reservoir structure. Environmental characteristics include ambient temperature (T) and pressure (P). Reservoir physical properties include sediment material (quartz sand, montmorillonite, illite, etc.), sediment density (ρ), sediment porosity (φ), and sediment grain size (r). Reservoir composition includes quartz content, montmorillonite content, and illite content. Reservoir structure includes component phase, water saturation (S_w), hydrate saturation (S_h), salt concentration (c), hydrate occurrence mode (pore-filling / particle-encapsulated), and component contact mode. These reservoir characteristics are obtained through experimental testing using various testing instruments.

[0018] Step 2: Input the characteristics of the natural gas hydrate reservoir with the predicted thermal conductivity into a pre-trained prediction model to obtain the thermal conductivity (λ) of the natural gas hydrate reservoir under these characteristics. The prediction model is trained in a pre-built machine learning model based on the thermal conductivity of the natural gas hydrate reservoir under different reservoir characteristics, specifically including: Step 2.1: Establish a reservoir thermal conductivity dataset. The dataset contains the thermal conductivity values ​​of natural gas hydrate reservoirs under different reservoir characteristics. These values ​​are calculated from physical data obtained through thermal conductivity testing experiments on natural gas hydrate reservoirs under different reservoir characteristics, as shown in Figure 2. The specific steps include: Step 2.1.1: Using the natural gas hydrate sediment thermal conductivity measurement experimental system, the physical data such as temperature (T), pressure (P), real-time voltage (U), and real-time current (I) of the natural gas hydrate reservoir thermal conductivity process under different reservoir characteristics were obtained by steady-state testing.

[0019] The aforementioned experimental system for measuring the thermal conductivity of sediments containing natural gas hydrates comprises a high-pressure reactor module, a constant temperature module, an electric heating module, a gas injection module, and a data acquisition module. The gas injection module includes a methane cylinder and a gas storage tank. During operation, gas at a certain pressure is first injected into an intermediate buffer tank through the methane cylinder, and then the required gas is injected into the reactor through the gas storage tank. The high-pressure reactor module is a stainless steel high-pressure reactor containing a long hollow cylinder, with heating tubes distributed along the axis of the cylinder. Each heating tube consists of a 1 mm diameter outer protective shell and a heating wire. The protective shell is made of stainless steel with good thermal conductivity, and the heating wire is platinum wire with a thermal resistance wire length of (L). The electric heating module is used to heat the heating wire at the center of the high-pressure reactor. The ends of the heating tube are insulated and sealed with epoxy resin, and copper rods are welded to both ends of the heating tube, connected to a DC power supply to provide a stable voltage (U) to the heating wire. Three Pt1000 thermocouples (T1, T2, T3) are installed on the same horizontal plane in the middle of the stainless steel high-pressure reactor, distributed at different radial positions (r1, r2, r3). In steady state, an isothermal region exists along the axial direction in the central region of the cylinder, within which a stable one-dimensional heat flow (I) is generated radially from the cylinder axis to the outer wall of the cylinder. The thermal conductivity experiment of natural gas hydrate reservoirs involves loading the sample into a reaction vessel, heating it with current after the temperature stabilizes, and recording physical data such as pressure (P), temperature (T), real-time voltage (U), and real-time current (I) using a data acquisition module.

[0020] Step 2.1.2: The thermal conductivity (λ) of the natural gas hydrate reservoir is calculated using physical data from experimental tests of the thermal conductivity process. This calculation is performed using a pre-defined module. This module employs the Python program in the appendix to calculate the thermal conductivity (λ) of the natural gas hydrate reservoir under different reservoir characteristics based on the physical data. The steps are as follows: Step 2.1.2.1: Read the physical data from the experiment using a Python program, and extract the parameters required to calculate the thermal conductivity: time t; ambient temperature; ambient pressure; thermocouple temperatures T1, T2, and T3; and experimental voltage U.

[0021] Step 2.1.2.2: Using graphing software, with the logarithm of time (lnt) as the x-axis and temperature (T1, T2, T3) as the y-axis, perform linear fitting of the data segment based on the least squares method and calculate the slope k of the line.

[0022] Suppose the equation of the fitted line is:

[0023] Least squares formula:

[0024] Step 2.1.2.3: Substitute the slope k obtained from the linear fitting into the following formula to calculate the thermal conductivity λ. Repeat this process to calculate the thermal conductivity for other time periods:

[0025] Where T is the experimental test temperature and t is the time. q The power per unit length of the heat source is calculated using the following formula:

[0026] in U For real-time voltage, I For real-time current, L This represents the length of the thermal resistance wire.

[0027] Step 2.2: Divide the dataset to obtain a training set and a test set. Select any set of thermal conductivity values ​​from the reservoir thermal conductivity dataset to form a thermal conductivity sample. Use one thermal conductivity sample and its corresponding parameter information to form a training sample. Select 80% of the training samples from the dataset to form the training set, and the remaining 20% ​​to form the test set.

[0028] Step 2.2.1: Dataset preprocessing, including outlier handling, missing value handling, and normalization. The normalization process unifies the thermal conductivity and corresponding parameter information of different reservoir characteristics in the training set into numerical form. These parameter information refers to reservoir characteristics, including ambient temperature (T), pressure (P), reservoir sediment material (quartz sand, montmorillonite, illite, etc.), reservoir sediment density (ρ), sediment porosity (φ), reservoir sediment grain size (r), quartz content, montmorillonite content, illite content, component phase, water saturation (S_w), hydrate saturation (S_h), salt concentration (c), hydrate occurrence mode (pore-filling / granular encapsulation), and component contact mode. The same reservoir characteristics were standardized numerically, so that different values ​​corresponded to different thermal conductivity values. Specifically, the units of reservoir characteristics were standardized to 1. For example, the unit of water saturation was standardized to g / g, the unit of salt concentration was standardized to mol / L, and the unit of thermal conductivity λ was standardized to W / (m·K). The data was then organized into Excel format to obtain the training data sample set.

[0029] Step 2.2.2: Set the target parameter to be predicted, i.e., the thermal conductivity λ, and calculate the correlation between variables. Calculate the Pearson correlation coefficient for each model according to the following formula. Analyze the correlation between several variables:

[0030] In the formula Let be the thermal conductivity of the i-th sample. The average thermal conductivity is... As a variable related to thermal conductivity, Let n be the average value of the variable and n be the sample size.

[0031] Step 2.2.3: Based on the correlation results of model variables in Step 2.2.1, establish a hierarchical model for predicting thermal conductivity under different reservoir characteristics. The models include Bagging, Random Forest, Decision Tree, Support Vector Regression, Gradient Boosting Decision Tree, and K-Nearest Neighbor algorithms. Input the training set data samples to be processed into the hierarchical ensemble model to generate prediction results. Each thermal conductivity prediction model is independently trained on the training set data samples.

[0032] Step 2.3: The stratified models for predicting thermal conductivity of different reservoir characteristics established in Step 2.2.3 are evaluated using the test set partitioned in Step 2.2. The prediction results are cross-validated using evaluation metrics to select the superior prediction model. The evaluation metrics include Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Root Mean Square Error (MASE), Goodness of Fit (R²), and Mean Absolute Percentage Error (MAPE), expressed as follows: The goodness of fit R 2 The formula for calculating the evaluation indicators is as follows:

[0033] The formula for calculating the Mean Absolute Error (MAE) evaluation index is as follows:

[0034] The formula for calculating the Mean Absolute Percentage Error (MAPE) evaluation index is as follows:

[0035] The formula for calculating the root mean square error (RMSE) evaluation index is as follows:

[0036] in, To test the thermal conductivity of reservoir samples obtained from indoor experiments using a concentrated sample set, The thermal conductivity of the reservoir sample is predicted by the model. This represents the average thermal conductivity of the reservoir sample predicted by the model.

[0037] Table 1 Comparison of evaluation metrics for prediction models obtained from training six hierarchical models

[0038] Step 3: Based on the selection in Step 2.3, a better pre-testing model is obtained. The characteristics of the natural gas hydrate reservoir whose thermal conductivity is to be predicted are input into the pre-testing model to generate the predicted thermal conductivity value, thus obtaining the thermal conductivity of the natural gas hydrate reservoir under this reservoir characteristic. This invention is based on 103 sets of laboratory measured data and compares the prediction results of the thermal conductivity of marine natural gas hydrate reservoirs using six stratified models. The prediction results of several models are shown in Table 1 and Figures 3-8. (Figure 3-8 is an example of this.) Figure 8 The figures show the predicted thermal conductivity of marine natural gas hydrate reservoirs under different reservoir characteristics using the KNN, DT, Bagging, SVR, RF, and GBDT models, respectively. The analysis shows that GBDT and RF outperform other models in terms of accuracy, fitting effect, and prediction correctness, indicating that the model has good generalization ability.

[0039] Step 4: Based on the prediction results generated in Step 3, perform SHapley Additive exPlanations (SHAP) interpretability analysis to verify the model's reliability and analyze the sensitivity of input parameter information (reservoir characteristics). This method quantifies the intervention effect of each feature in the model construction process, establishes the interpretative relationship between input features and output results based on marginal contribution theory, and derives the contribution degree of different factors. Its principle is as follows:

[0040]

[0041] In the formula This is for post-hoc interpretation of model predictions; The predicted value is from a machine learning algorithm. S represents the mean of the predicted values ​​of the machine learning algorithm on the training set; S is a subset of the features. is the contribution value of the j-th feature (S value); M is the total number of features; X is the set of all instances.

[0042] The results, as shown in Figure 9, show the magnitude and criticality of each factor's influence on the system's thermal conductivity by comparing the distribution range of the SHAP values ​​for each factor. Figure 10 shows the order of criticality of the influencing factors as follows: salt concentration > reservoir sediment material > component phase > component contact mode > water saturation.

[0043] The present invention also provides an electronic device, which can be a mobile terminal, desktop computer, laptop, handheld computer, server, or other computing device. The electronic device includes a processor, a memory, and a display.

[0044] In some embodiments, the memory can be an internal storage unit of a computer device, such as a hard drive or RAM. In other embodiments, the memory can be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory can include both internal and external storage units. The memory is used to store application software and various types of data installed on the computer device, such as program code installed on the computer device. The memory can also be used to temporarily store data that has been output or will be output. In one embodiment, a prediction method program is stored on the memory.

[0045] In some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display is used to show information on a computer device and to display a visual user interface. Components of the computer device communicate with each other via a system bus.

[0046] This embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the prediction methods described above.

[0047] In the several embodiments provided by this invention, it should be understood that the disclosed methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of predicting thermal conductivity of a marine natural gas hydrate reservoir, characterized by, The application relates to a method for predicting the thermal conductivity of a natural gas hydrate reservoir. The method comprises the following steps: acquiring reservoir characteristics of a natural gas hydrate reservoir to be predicted, wherein the reservoir characteristics comprise environmental characteristics, reservoir physical characteristics, reservoir composition characteristics and reservoir structure characteristics; inputting the reservoir characteristics into a pre-trained prediction model to obtain the thermal conductivity of the natural gas hydrate reservoir corresponding to the reservoir characteristics; wherein the prediction model is obtained by training the thermal conductivity of the natural gas hydrate reservoir under different reservoir characteristics in a pre-constructed machine learning model, and the method comprises the following steps: acquiring thermal conductivity values of the natural gas hydrate reservoir under a preset large number of different reservoir characteristics; constructing a data set by using the thermal conductivity values and the corresponding reservoir characteristics, pre-processing the data set, and dividing the data set into a training set and a test set; extracting feature data of all the reservoir characteristics and the thermal conductivity values in the training set; based on the feature data, performing hierarchical modeling on the data set by using machine learning calculation, training the obtained hierarchical model, and obtaining the prediction model; using evaluation indexes to evaluate the prediction model by using the test set, and screening the prediction model; 2. The method of predicting the thermal conductivity of a marine natural gas hydrate reservoir according to claim 1, wherein, performing explainability analysis on the predicted results, verifying the reliability of the prediction model, and analyzing the sensitivity of the input reservoir characteristics.

3. The method of predicting the thermal conductivity of a marine natural gas hydrate reservoir according to claim 2, wherein, The environmental characteristics comprise environmental temperature (T) and pressure (P); the reservoir physical characteristics comprise reservoir sediment material, reservoir sediment density, sediment porosity and reservoir sediment particle size; the reservoir composition characteristics comprise the content of various reservoir sediment materials; and the reservoir structure characteristics comprise component phase state, water saturation, hydrate saturation, salt concentration, hydrate occurrence mode and component contact mode. The data set is constructed, and the method comprises the following steps: obtaining the temperature (T), pressure (P), real-time voltage (U) and real-time current (I) of the thermal conduction process of the natural gas hydrate reservoir under different reservoir characteristics by using a steady-state method test of a natural gas hydrate-containing sediment thermal conduction measurement experimental system; calculating the thermal conductivity (lambda) value of the natural gas hydrate reservoir under different reservoir characteristics, and the calculation steps are as follows: taking the logarithm of time lnt as an x axis and temperature T as a y axis, linearly fitting the data based on a least square method, and calculating the slope k of the straight line; assuming that the linear equation of fitting is: the least square method formula is: where T is the experimental test temperature, t is time, q The heat source power per unit length is calculated from the following equation: wherein U is the real-time voltage, I is the real-time current, L is the length of the heat resistance wire.

4. The method of predicting thermal conductivity of a marine natural gas hydrate reservoir according to claim 3, wherein, the linearly fitted slope k is substituted into the following formula to obtain the thermal conductivity lambda, and the thermal conductivity is calculated: the natural gas hydrate-containing sediment thermal conduction measurement experimental system comprises a high-pressure reaction kettle module, a constant temperature module, an electric heating module, a gas injection module and a data acquisition module. The high-pressure reaction kettle module is a stainless steel high-pressure reaction kettle comprising a long hollow cylinder, wherein heating pipes are arranged along the axis for heating, the heating pipe is composed of a 1mm diameter protective shell and a heating wire, the protective shell is made of stainless steel with good heat conduction performance, the heating wire is a platinum wire, the length of the heat resistance wire is (L), the end of the heating pipe is insulated and sealed by epoxy resin, copper rods are welded at both ends of the heating pipe and connected with a direct current power supply to provide a stable voltage (U) to both ends of the heating wire, three Pt1000 thermocouples (T1, T2, T3) are arranged at different radial positions (r1, r2, r3) at the same horizontal plane in the middle of the stainless steel high-pressure reaction kettle, and in the steady state, there is an isothermal zone in the central region of the cylinder along the axial direction, and a stable one-dimensional heat flow (I) is generated along the radial direction from the axis of the cylinder to the outer wall of the cylinder in the isothermal zone; a data acquisition module is used to record the physical data of pressure (P), temperature (T), real-time voltage (U) and real-time current (I) during the experiment.

5. The method of predicting thermal conductivity of a marine natural gas hydrate reservoir according to claim 3, wherein, The evaluation indicators of the screening prediction model include mean absolute error (MAE), root mean square error (RMSE), mean absolute scaled error (MASE), goodness of fit (R2), and mean absolute percentage error (MAPE).

6. The method of predicting the thermal conductivity of a marine natural gas hydrate reservoir according to claim 5, wherein, The reliability of the prediction model is verified and the sensitivity of the reservoir characteristics is analyzed by using SHapley Additive exPlanations (SHAP) explainability analysis.

7. A computer readable storage medium characterized by The computer program is stored, and the computer program is executed by the processor to realize the method of any one of claims 1-6.

8. An electronic device, comprising: The computer program is stored, and the computer program is executed by the processor to realize the method of any one of claims 1-6.

Images