基于机器学习方法的海洋天然气水合物水平井降压开采模拟—优化耦合模型  

Simulation-optimization coupling model for the depressurization production of marine natural gas hydrate in horizontal wells based on machine learning method

作　　者：辛欣[1] 王海彬 罗建男[1] 于涵[1] 袁益龙 夏盈莉 朱慧星[1] 陈强 XIN Xin;WANG Haibin;LUO Jiannan;YU Han;YUAN Yilong;XIA Yingli;ZHU Huixing;CHEN Qiang(Key Laboratory for Underground Water Resources and Environment,Ministry of Education//Jilin University,Changchun,Jilin 130012,China;Key Laboratory of Gas Hydrate,Ministry of Natural Resources//Qingdao Institute of Marine Geology,China Geological Survey,Qingdao,Shandong,266071,China;Laboratory for Marine Mineral Resources,Pilot National Laboratory for Marine Science and Technology-Qingdao,Qingdao,Shandong 266071,China)

机构地区：[1]吉林大学地下水资源与环境教育部重点实验室 [2]自然资源部天然气水合物重点实验室·中国地质调查局青岛海洋地质研究所 [3]青岛海洋科学与技术试点国家实验室海洋矿产资源评价与探测技术功能实验室

出　　处：《天然气工业》2020年第8期149-158,共10页Natural Gas Industry

基　　金：自然资源部中国地质调查局地质调查项目(编号:DD20190231);国家自然科学基金项目“天然气水合物形成和分解过程的多相流数值模拟研究”(编号:41202166)。

摘　　要：由天然气水合物(以下简称水合物)开采引起的含水合物沉积层力学变形问题不能被忽视,因其直接威胁到海域水合物的安全开采。为了找到甲烷累计产气量最优值与地层稳定性的关系,基于机器学习方法形成模拟-优化耦合技术,构建起水合物降压开采传热-流动-力学数值模拟模型、可以替代数值模拟模型的机器学习模型和以甲烷累计产气量最优为目标的混合整数非线性规划优化模型;在此基础上,选取南海北部神狐海域厚层Ⅱ类水合物藏W11站位为研究对象,获得了海底面沉降量约束下的水合物储层甲烷累计产气量及相对应的最优开采方案参数。研究结果表明:①模拟-优化耦合技术的关键是机器学习方法的运用,基于径向基函数人工神经网络方法而建立的替代模型计算精度较高,可以替代模拟模型来确定输入输出变量的关系,从而摆脱既定方案的限制,找到全局最优解;②模拟-优化耦合技术可以解决受含水合物沉积层力学响应特征影响的水合物开采方案优选问题,根据试采工程安全要求改变海底面沉降量最大允许值,可以计算得到相应的甲烷累计产气量,以及降压幅度、开采时间、井位布置、水平井段长度等最优开采方案参数;③随着最大允许沉降量增大,甲烷累计产气量增大,二者满足正相关关系;④海底面沉降量随着开采时间增长而增大,也随降压幅度增大而增大;⑤水合物开采引起海底面沉降主要发生在开采初期,为了获得较高甲烷累计产气量及较小海底面沉降量最大允许值,在开采初期必须减小降压幅度。结论认为,所形成的模拟-优化耦合技术适用性强,可以为水合物安全、高效规模化开采方案的制订提供支撑。The mechanical deformation of hydrate bearing deposits caused by the production of natural gas hydrate(NGH)directly threatens the safe production of marine NGH,so it shall not be ignored.In order to determine the relationship between optimal cumulative methane production and formation stability,this paper developed the simulation-optimization coupling technology on the basis of machine learning method.Then,a numerical simulation model of thermal transfer-hydrodynamic-mechanical process(THM)for NGH depressurization production,a machine learning model that can replace numerical simulation model and a mixed integer nonlinear programming optimization model aiming at the optimal cumulative methane production were established.On this basis,the Site W11 of TypeⅡthick hydrate reservoir in the Shenhu sea area of the South China Sea was selected as the research object to calculate the cumulative methane production of NGH reservoirs under the constraint of seafloor subsidence and the parameters of corresponding optimal production scheme.And the following research results were obtained.First,the key to the simulation-optimization coupling technology is the application of the machine learning method.The substitution model established based on the RBF(radial basis function)artificial neural network method has a higher calculation accuracy and it can replace the simulation model to determine the relationship between input and output variables,so as to avoid the limitation of the established scheme and work out the global optimal solution.Second,the simulation-optimization coupling technology can solve the optimization problem of hydrate production scheme affected by the mechanical response characteristics of hydrate bearing deposits.By changing the maximum allowable seafloor subsidence according to the safety requirements of production test engineering,it can calculate the corresponding cumulative methane production,as well as the parameters of optimal production scheme,such as depressurization amplitude,production duration,w

关 键 词：海洋天然气水合物 地层稳定性 数值模拟 机器学习 混合整数 优化模型 开采方案制订 水平井 

分 类 号：TE53[石油与天然气工程—油气田开发工程] TP181[自动化与计算机技术—控制理论与控制工程]