三维初至波旅行时层析速度反演算法优化

Optimization of 3D first-arrival traveltime tomography inversion

通过机器学习拾取初至波、建立微测井约束的初始速度场、多模板快速推进算法与窄带延拓算法求取全局旅行时、射线追踪技术求取射线传播路径、视慢度与正则化对目标反演方程进行约束等方法对三维初至波旅行时层析速度反演算法进行了优化。研究结果表明:(1)采用基于图像分割的机器学习算法,将地震记录分成初至前、初至后、包含初至窄带等3类,并赋予不同权重,对576 000道地震数据进行初至拾取,拾取精度超过99%,拾取效率较人工逐道拾取提升了80余倍。(2)采用多模板快速推进算法,6个差分模板可覆盖网格周围26个节点的旅行时计算,可提升网格对角方向的旅行时计算精度;三维窄带延拓技术可完成全局网格节点正演旅行时计算;基于Runge-Kutta算法的三维射线追踪技术可完成射线路径的求取。(3)使用微测井信息构建初始速度场,同时使用视慢度信息与正则化方法对层析反演方程进行约束,可有效提高地下1 000 m以内的速度反演精度。(4)采用优化后的层析反演技术对塔里木油田甫沙4线束与英买某区块的实际资料进行处理,消除了“牛眼”假象与边界伪影假象,且井点反演速度曲线与实际测井数据高度吻合。

The 3D first-arrival traveltime tomography inversion was optimized by picking up the first-arrival wave using machine learning,establishing the initial velocity constrained by micro logs,obtaining the global traveltime by multi-stencils fast marching method and narrow-band continuation algorithm,gaining the ray propagation path by ray tracing technology,and restricting the target inversion equation by apparent slowness and regularization.The results show that:(1)The machine learning based on image segmentation was used to divide the seismic records into three categories,including the early,the later and the narrow band of first arrival.The classification was weighted to pick up 576000 channels of seismic data.The accuracy of picking up was greater than 99%,and the efficiency of picking up was more than 80 times higher than that of manual pick up.(2)Using the multi-stencils fast marching method which includes six difference templates can cover the traveltime calculation of 26 nodes around the grid,which can improve the traveltime calculation accuracy of grid diagonal direction.The 3D narrowband extension technique could be used to calculate the forward traveltime of global grid nodes.The 3D ray tracing technique based on Runge-Kutta algorithm could obtain the ray path.(3)Using micro logs information to construct the initial velocity and using apparent slowness information and regularization technology to constrain the tomography inversion equation can effectively improve the accuracy of velocity inversion within 1000 m underground.(4)The optimized tomography inversion technology has been used to process the actual seismic data of Fusha 4 wiring bundle and Yingmai block in Tarim Oilfield,which could eliminate the“bull’s eye”illusion and boundary artifact.The velocity curve of well point is highly consistent with the actual logs data.