由于页岩气渗流机理复杂,赋存方式多样,压裂后对裂缝网络的精确识别和表征存在较大困难,现有方法难以准确预测页岩气井产量。为此,提出了机理—数据融合建模的思路,结合连续拟稳态假设、物质平衡方程、产量递减分析方法和递推原理,建立...由于页岩气渗流机理复杂,赋存方式多样,压裂后对裂缝网络的精确识别和表征存在较大困难,现有方法难以准确预测页岩气井产量。为此,提出了机理—数据融合建模的思路,结合连续拟稳态假设、物质平衡方程、产量递减分析方法和递推原理,建立了物理—数据协同驱动的产量预测方法,进而以中国某区块页岩气井现场生产数据为例,对该方法的准确性、可靠性进行了测试,并与经验产量递减分析和时间序列分析方法进行了对比分析。研究结果表明:(1)建立的产能模型采用拟压力代替压力,采用物质平衡拟时间代替时间,弱化了产量、流压和甲烷物性变化带来的影响;(2)以累计产量误差最小为目标开展历史拟合,弱化了生产制度变化带来的影响,使得建立的产能模型能够自动适应流压—产量变化;(3)应用该方法的关键在于采气指数—物质平衡拟时间双对数图中的特征直线,若图中出现特征直线,则可以开展产量预测,反之,则不能预测。结论认为:(1)建立的产量预测方法将不稳定流动问题转化为拟稳态流动问题求解,简化了对储层非均质性的描述,避开了裂缝网络精确识别和定量表征的难题,计算效率高,可解释性强;(2)生产数据测试结果表明该产量预测方法精度高,长期预测结果稳定,并优于Logistic Growth Model、Duong和StretchedExponential Production Decline经验产量递减分析方法,也优于非线性自回归神经网络、长短记忆神经网络时间序列分析方法。展开更多
针对电力系统扰动后频率响应计算问题,该文基于门控循环单元神经网络提出一种融合物理与数据知识的频率在线计算方法,以实现频率快速精准计算。该方法以同步电源惯性时间常数等影响频率响应的主导因素作为经典“黑箱”机器学习方法的基...针对电力系统扰动后频率响应计算问题,该文基于门控循环单元神经网络提出一种融合物理与数据知识的频率在线计算方法,以实现频率快速精准计算。该方法以同步电源惯性时间常数等影响频率响应的主导因素作为经典“黑箱”机器学习方法的基本输入特征量,并进一步在“黑箱”方法中嵌入频率响应相关物理知识,通过基本输入特征量和所嵌入物理知识形成新的输入特征量并用于模型训练。该方法能够提高小样本场景下的模型泛化能力和抗噪能力,并且增强其可解释性。采用新英格兰10机39节点系统作为仿真算例,通过与电力系统仿真器(power system simulator for engineering,PSS/E)中的仿真结果相对比,证明所提方法能够快速、准确地计算电力系统扰动后频率响应曲线。展开更多
Seismic fluid identification works as an effective approach to characterize the fluid feature and distribution of the reservoir underground with seismic data. Rock physics which builds bridge between the elastic param...Seismic fluid identification works as an effective approach to characterize the fluid feature and distribution of the reservoir underground with seismic data. Rock physics which builds bridge between the elastic parameters and reservoir parameters sets the foundation of seismic fluid identification, which is also a hot topic on the study of quantitative characterization of oil/gas reservoirs. Study on seismic fluid identification driven by rock physics has proved to be rewarding in recognizing the fluid feature and distributed regularity of the oil/gas reservoirs. This paper summarizes the key scientific problems immersed in seismic fluid identification, and emphatically reviews the main progress of seismic fluid identification driven by rock physics domestic and overseas, as well as discusses the opportunities, challenges and future research direction related to seismic fluid identification. Theoretical study and practical application indicate that we should incorporate rock physics, numerical simulation, seismic data processing and seismic inversion together to enhance the precision of seismic fluid identification.展开更多
文摘由于页岩气渗流机理复杂,赋存方式多样,压裂后对裂缝网络的精确识别和表征存在较大困难,现有方法难以准确预测页岩气井产量。为此,提出了机理—数据融合建模的思路,结合连续拟稳态假设、物质平衡方程、产量递减分析方法和递推原理,建立了物理—数据协同驱动的产量预测方法,进而以中国某区块页岩气井现场生产数据为例,对该方法的准确性、可靠性进行了测试,并与经验产量递减分析和时间序列分析方法进行了对比分析。研究结果表明:(1)建立的产能模型采用拟压力代替压力,采用物质平衡拟时间代替时间,弱化了产量、流压和甲烷物性变化带来的影响;(2)以累计产量误差最小为目标开展历史拟合,弱化了生产制度变化带来的影响,使得建立的产能模型能够自动适应流压—产量变化;(3)应用该方法的关键在于采气指数—物质平衡拟时间双对数图中的特征直线,若图中出现特征直线,则可以开展产量预测,反之,则不能预测。结论认为:(1)建立的产量预测方法将不稳定流动问题转化为拟稳态流动问题求解,简化了对储层非均质性的描述,避开了裂缝网络精确识别和定量表征的难题,计算效率高,可解释性强;(2)生产数据测试结果表明该产量预测方法精度高,长期预测结果稳定,并优于Logistic Growth Model、Duong和StretchedExponential Production Decline经验产量递减分析方法,也优于非线性自回归神经网络、长短记忆神经网络时间序列分析方法。
文摘针对电力系统扰动后频率响应计算问题,该文基于门控循环单元神经网络提出一种融合物理与数据知识的频率在线计算方法,以实现频率快速精准计算。该方法以同步电源惯性时间常数等影响频率响应的主导因素作为经典“黑箱”机器学习方法的基本输入特征量,并进一步在“黑箱”方法中嵌入频率响应相关物理知识,通过基本输入特征量和所嵌入物理知识形成新的输入特征量并用于模型训练。该方法能够提高小样本场景下的模型泛化能力和抗噪能力,并且增强其可解释性。采用新英格兰10机39节点系统作为仿真算例,通过与电力系统仿真器(power system simulator for engineering,PSS/E)中的仿真结果相对比,证明所提方法能够快速、准确地计算电力系统扰动后频率响应曲线。
基金supported by the National Basic Research Program of China(Grant No.2013CB228604)the National Grand Project for Science and Technology(Grant Nos.2011ZX05030-004-002,2011ZX05019-003,2011ZX05006-002)SINOPEC Key Laboratory of Geophysics+2 种基金Science Foundation for Post-doctoral Scientists of ChinaScience Foundation for Post-doctoral Scientists of Shandongthe Western Australian Energy Research Alliance(WA:ERA)
文摘Seismic fluid identification works as an effective approach to characterize the fluid feature and distribution of the reservoir underground with seismic data. Rock physics which builds bridge between the elastic parameters and reservoir parameters sets the foundation of seismic fluid identification, which is also a hot topic on the study of quantitative characterization of oil/gas reservoirs. Study on seismic fluid identification driven by rock physics has proved to be rewarding in recognizing the fluid feature and distributed regularity of the oil/gas reservoirs. This paper summarizes the key scientific problems immersed in seismic fluid identification, and emphatically reviews the main progress of seismic fluid identification driven by rock physics domestic and overseas, as well as discusses the opportunities, challenges and future research direction related to seismic fluid identification. Theoretical study and practical application indicate that we should incorporate rock physics, numerical simulation, seismic data processing and seismic inversion together to enhance the precision of seismic fluid identification.
