提高深层地震资料分辨率方法研究

发布时间：2018-05-30 00:40

本文选题：高分辨率 + SDRNAR　；参考：《中国石油大学(北京)》2017年博士论文

【摘要】：随着油气勘探的深入发展,深层地震勘探是今后油气勘探的重要领域,是实现并完成油气储量任务的重要手段,对油气勘探与开发具有重要的现实意义。与中浅层地震勘探不同的是,由于埋藏深度较深以及地下地质条件复杂,深层地震资料信噪比更低、层间多次波淹没一次波、叠加速度估计不准确以及地震信号吸收衰减更为严重。为了提高深层勘探精度,提高深层地震资料的分辨率有着重要意义。本文联合正则化非平稳自回归谱分解技术(spectral decomposition using regularized non-stationary autoregression简称SDRNAR技术)、Gabor反褶积方法提高深层资料分辨率。两种方法联合使用目前尚属首例。SDRNAR技术将地震信号分解为具有不同频率和振幅的信号组合,分解得到的这些信号具有平滑变化的频率和振幅。该方法在分解、重构原始信号时,压制噪声。在此基础上利用Gabor反褶积技术拓宽目标信号的频带宽度。模型数据以及实际数据应用效果来看,深层地震信号分辨率得到提高。深层资料中广泛发育的层间多次波是影响一次波有效成像的重要因素。常规方法由于对模型或者数据的依赖性,导致最后的预测相减不够彻底,残留的层间多次波造成构造假象。本文利用数学形态学在信噪分离上的优势,对一次波和层间多次进行分离。模型数据和实际地震数据的应用效果显示,压制近道层间多次波取得满意的结果。由于地质条件复杂,深层速度信息估计不准确,同相轴不能满足常规叠加所需要的双曲线形态,同相叠加难度大,远偏移距信息在叠加过程中被弱化,局限性较为突出。CRS(the common reflection surface)方法不依赖于宏观速度信息,由反射面元的位置、倾角、曲率这三个参数来代替叠加速度信息,实现共反射面元叠加。本文改进CRS应用思路,首先在叠前道集上偏移距方向进行相似性加权叠加,再沿着叠加剖面同相轴走向利用平面波校平技术进行叠加。模拟数据和实际深层地震数据的应用效果显示,深层弱同相轴得到了加强。地震信号在传播过程中由于几何扩散、介质的非完全弹性等因素会发生频率、振幅上的衰减,尤其深层地震资料的吸收衰减现象更为严重。地层吸收衰减作用常用品质因子Q值来表示,通过估算Q值来实现补偿的方法很多,需要各种各样的假设,本文应用形态学多尺度算法,对深层地震资料的能量进行补偿,再利用反褶积方法补偿频率,模型数据以及实际数据的应用效果显示,该方法适用性强,取得了较好的补偿效果。
[Abstract]:With the further development of oil and gas exploration, deep seismic exploration is an important field of oil and gas exploration in the future, is an important means to realize and complete the task of oil and gas reserves, and has important practical significance for oil and gas exploration and development. Different from the middle and shallow seismic exploration, because of the deep buried depth and complicated underground geological conditions, the signal-to-noise ratio of the deep seismic data is lower, and the interlayer multiple wave submerges the first wave. The estimation of stacking velocity is inaccurate and the attenuation of seismic signal is more serious. In order to improve the precision of deep exploration and improve the resolution of deep seismic data, it is of great significance. In this paper, a regularized non-stationary autoregressive spectral decomposition technique, SDRNAR decomposition using regularized non-stationary autoregression, is used to improve the resolution of deep data. The two methods are still the first. SDRNAR technology to decompose the seismic signal into different frequency and amplitude signal combination. These signals have smooth variation frequency and amplitude. The noise is suppressed when the original signal is decomposed and reconstructed. On this basis, the Gabor deconvolution technique is used to widen the bandwidth of target signal. The resolution of deep seismic signal is improved by using model data and actual data. Interlaminar multiple, which is widely developed in deep layer data, is an important factor affecting primary wave effective imaging. Because of the dependence on the model or data, the subtraction of the final prediction is not complete enough, and the residual interlayer multiples make the structure false. In this paper, the advantage of mathematical morphology in the separation of signal and noise is used to separate the primary wave from the interlayer multiple times. The application results of model data and actual seismic data show that satisfactory results can be obtained by suppressing near-channel interlayer multiple. Because of the complex geological conditions and inaccurate estimation of deep velocity information, the cophase axis can not meet the hyperbolic shape required by conventional stacking, and the in-phase stacking is difficult, and the far-offset information is weakened in the stack process. The limitation of the common reflection surface) method is not dependent on the macroscopic velocity information, but is replaced by the position, dip angle and curvature of the reflecting surface element instead of the superposition velocity information, and the common reflection surface element superposition is realized. In this paper, the idea of CRS application is improved. Firstly, similarity weighted stacking is carried out in the offset direction of prestack gathers, and then the plane wave leveling technique is used to stack along the same phase axis strike of the stack profile. The application results of simulated data and actual deep seismic data show that the weak cophase axis of deep layer has been strengthened. In the process of seismic signal propagation, due to geometric diffusion and incomplete elasticity of the medium, frequency and amplitude decay, especially the absorption and attenuation of deep seismic data, are more serious. The Q value of the quality factor is commonly used to express the absorption and attenuation of strata. There are many methods to realize compensation by estimating the Q value, which requires a variety of assumptions. In this paper, the energy of deep seismic data is compensated by morphological multi-scale algorithm. The application of deconvolution method to compensate frequency, model data and actual data shows that the method has strong applicability and good compensation effect.
【学位授予单位】：中国石油大学(北京)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：P631.44

【相似文献】