最小二乘偏移研究现状及发展趋势

Research status and development trend of least square migration

地震勘探的核心目标是尽可能定量地、精确地描述油气藏,地震波成像由定位反射(散射)点位置发展到当前的估计(角度)反射系数是地震勘探的核心需求。一般地,逆时偏移是复杂介质成像最精确的方法,最小二乘偏移成像是估计(角度)反射系数的理想选择。最小二乘偏移成像基于线性反演理论框架,理论上能够消除采集照明不佳的影响、均衡成像振幅以及提高成像分辨率。然而,该理论优势并没有被转化成预期的实用效果,最小二乘偏移技术的生产应用仍然处于试验探索阶段,不能大规模推广应用。在对国内外最小二乘偏移成像技术进行全面调研的基础上,介绍了该技术的方法原理,指明了该技术的理论优势,分析了数据域迭代反演算法和成像域非迭代反演算法两种最小二乘偏移成像技术的特点,认为最小二乘偏移成像技术至今尚未规模化应用于生产的原因在于:(1)背景速度的精度不能满足线性反演成像问题的假设条件;(2)Born近似正演算子不能很好地模拟实际观测数据中的一次反(散)射波;(3)噪声不满足高斯假设条件;(4)子波未知增加了模拟数据的误差;(5)计算量大等。最后指出,合理的数据匹配技巧、合适的正则化技术及近似计算Hessian逆矩阵是未来最小二乘偏移技术应用研究的方向,长期看应该将最小二乘偏移成像融入到全波形反演(FWI)中。

Quantitative and accurate description of oil and gas reservoirs is the goal of seismic exploration.In view of that,seismic imaging has advanced from locating reflection/scattering point to estimating angle reflection coefficients.In general,reverse time migration(RTM)is the most accurate method for imaging complex structure,and least squares migration(LSM)is the best choice for estimating angle reflection coefficients.LSM is based on the linearized inversion theoretical framework.In theory,LSM can remove the influence of poor illumination and balance amplitude,and can improve image resolution.However,these theoretical advantages have not been realized in practical applications and the application of LSM is still in experimental exploration.This paper briefly introduces the theory and advantages,then compares the application effects of data-domain iterative inversion and the image-domain non-iterative inversion algorithm.Some of the observed major challenges of LSM in practical application were as follows:the background velocity cannot meet the requirement of linearized inversion,Born modeling operator cannot simulate primary reflected/scattered wave in observed data well,noise does not satisfy the Gaussian distribution assumption,the wavelet is unknown,and the cost of computation is high.In short term,the application of LSM depends on reasonable data matching techniques,an appropriate regularization technique,and the approximation of the inverse Hessian matrix.Meanwhile,integrating LSM into full waveform inversion(FWI)is essential in the long term.