摘要
油气田数字化、智能化是未来发展的必然趋势,目前数据驱动软测量模型已逐步应用于海上天然气生产系统流量、压力等关键状态变量的监测。为进一步提高数据驱动软测量模型准确度和生产监测水平,开展基于非线性滤波技术的软测量研究,结合压力传感器观测数据与所建外源输入非线性自回归模型库,建立了单个无迹卡尔曼滤波器(Unscented Kalman Filter,UKF)以及“先融合后滤波”(FF1)和“先滤波后融合”(FF2)等2种融合UKF,实现单井流量和井口压力的在线校正估计。研究结果表明:(1)基于UKF的软测量在纯模型软测量的基础上提高了准确度,且单个UKF、FF1、FF2等3种软测量方案的准确度和计算时间都满足在线估算需求;(2)所建模型库中,基于深度前馈网络灰箱模型构造的UKF准确性和稳定性较高,满足了生产的需要;(3)对比了3种软测量方案的优劣,认为FF1的全局准确性更高、耗时更少,推荐作为首选软测量方案。结论认为,研究结果对智能化海上油气田建设具有引导性作用,所推荐的软测量方案具有较强的工程适用性,为未来构建智慧油气田奠定了技术基础。
Digital and intelligent oil and gas field is an inevitable trend for the industry. Currently, the data-driven soft sensing model has been gradually applied to monitor the key state variables(e.g. flow rate and pressure)of the offshore natural gas production system.In order to improve the accuracy of the model and ensure the production safety, this paper researches the nonlinear-filtering-based soft sensing, and then combined with the measurement data of pressure sensors and the nonlinear auto-regressive model base with exogenous inputs(NARX), establishes a single unscented Kalman filter(UKF) and two fusion UKFs( "fusion before filtering"(FF1) and "filtering before fusion"(FF2), achieving online correction estimation of single-well flow rate and wellhead pressure. And the following research results are obtained. First, the estimation accuracy of the UKF-based soft sensing is higher than that of the model-based soft sensing;and the three soft sensing schemes of single UKF, FF1 and FF2 all meet the demand of online estimation in accuracy and calculation time.Second, the UKF constructed on the basis of grey-box Dense Neural Network NARX model has better accuracy and stability than the other constructed model bases and satisfies the need of production. Third, the comparison between three soft sensing schemes indicates that FF1 is the highest in global accuracy and least in time consumption, so it is recommended as the preferred soft sensing scheme. It is concluded that the research results have a guiding role in the construction of intelligent offshore oil and gas fields, and the recommended soft-sensing scheme is of greater engineering applicability, which lays a technical foundation for the construction of intelligent oil and gas fields in the future.
作者
王丹
康琦
杨居衡
宫敬
张奇
WANG Dan;KANG Qi;YANG Juheng;GONG Jing;ZHANG Qi(School of Economics and Management,China University of Petroleum(Beijing),Beijing 102249,China;CNPC Economics and Technology Research Institute,Beijing 100724,China;Research Institute of Tsinghua University in Shenzhen,Shenzhen,Guangdong 518057,China;PetroChina International Co.,Ltd.,Beijing 100033,China;National Engineering Laboratory for Pipeline Safety,China University of Petroleum(Beijing),Beijing 102249,China)
出处
《天然气工业》
EI
CAS
CSCD
北大核心
2022年第9期84-92,共9页
Natural Gas Industry
基金
国家科技重大专项“海上管道降凝输送及流动管理技术研究”(编号:2016ZX05028-004-001)
国家自然科学基金面上项目“可燃冰固态流化开采中天然气水合物浆体分解机理与输送规律研究”(编号:51874323)
中海石油(中国)有限公司科研课题“渤海浅水气田水下生产流动管理系统”(编号:CCL2020RCPG0318RSN)。
关键词
天然气生产系统
软测量
数据驱动
数据融合
最优估计
融合滤波
智慧油气田
Natural gas production system
Soft sensing
Data-driven
Data fusion
Optimal estimation
Fusion filter
Intelligent oil and gas field