AMHP Dataset

The AMHP (AI-based metamodel for hydrate production) dataset generated in this paper is provided in this dataset depository. The AMDHP dataset is divided into three Microsoft excel files, namely, TrainingData.xlsx, BackAnalysis.xlsx, and Sensitivity. xlsx. The arrangement of data in those three files is as follows: • TrainingTestingData.xlsx contains data obtained from the FLAC THM simulation used for the ANN training and testing. There are four sheets in this file. The first sheet (“input” sheet) provides 8,000 samples of 18 input variables. Sheet 2 (“Gas” sheet), sheet 3 (“Water” sheet), and sheet 4 (“Disp” sheet) give the time history output of gas, water, and vertical displacement corresponding to each input sample in sheet 1. Since the time histories of the three reservoir responses are calculated in a time step of 0.2 days within six days, each output sample contains 30 values and is denoted as t1 to t30 in the column headings. • Backanalysis.xlsx displays data related to back analysis. There are three sheets in this document. The first sheet reports 100,000 samples of 18 input variables. The second and third sheets, respectively, show the time history of gas and water productions corresponding to the samples shown in sheet 1. Similar to the TrainingData.xlsx file, the outputs are reported for six days with a time step of 0.2, each output sample contains 30 values and is denoted as t1 to t30 in the column headings. • Sensitivity. xlsx provides data related to sensitivity analysis. This file contains four sheets. The first sheet shows 54,000 samples of 18 input variables. The second, third, and fourth sheets, respectively, provide the time histories of gas, water, and displacement corresponding to the sample shown on sheet 1. Similar to the TrainingData.xlsx file, the outputs are reported for six days with a time step of 0.2. Each output sample contains 30 values and is denoted as t1 to t30 in the column headings.本文生成的AMHP（AI-based metamodel for hydrate production）数据集已上传至本数据集仓库。AMDHP数据集分为三个Microsoft Excel文件，分别为TrainingData.xlsx、BackAnalysis.xlsx及Sensitivity.xlsx。三个文件内的数据编排如下： • TrainingTestingData.xlsx包含用于人工神经网络（Artificial Neural Network, ANN）训练与测试的FLAC THM数值模拟数据。该文件包含4个工作表：第1个工作表（"input"工作表）提供18个输入变量的8000组样本；第2个工作表（"Gas"工作表）、第3个工作表（"Water"工作表）及第4个工作表（"Disp"工作表）分别给出与第1个工作表中每组输入样本对应的气体、水与垂直位移的时程响应。由于三类储层响应的时程以0.2天为时间步长计算，总时长为6天，因此每组输出样本包含30个数值，列标题记为t1至t30。 • Backanalysis.xlsx展示与反演分析相关的数据。该文件包含3个工作表：第1个工作表包含18个输入变量的100000组样本；第2、3个工作表分别对应第1个工作表中样本的产气与产水的时程响应。与TrainingData.xlsx文件类似，输出结果以0.2天为时间步长、总时长6天的条件下生成，每组输出样本包含30个数值，列标题记为t1至t30。 • Sensitivity.xlsx提供与敏感性分析相关的数据。该文件包含4个工作表：第1个工作表包含18个输入变量的54000组样本；第2、3、4个工作表分别对应第1个工作表中样本的气体、水与位移的时程响应。与TrainingData.xlsx文件类似，输出结果以0.2天为时间步长、总时长6天的条件下生成，每组输出样本包含30个数值，列标题记为t1至t30。