High-Resolution Seismic Stochastic Inversion as a Direct Input for Reservoir Modeling

Abstract

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper OTC 26404, “High-Resolution Seismic Stochastic Inversion as a Direct Input for Reservoir Modeling,” by Chen Xin, Wei Xiao-Dong, Li Yan-Jing, Cui Yi, Ma Yingzhe, Yan Xiao-Huan, and Xia Yaliang, CNPC, and Wang Guan and Wang Xiaotian, China University of Petroleum, prepared for the 2016 Offshore Technology Conference Asia, Kuala Lumpur, 22–25 March. The paper has not been peer reviewed. Copyright 2016 Offshore Technology Conference. Reproduced by permission. Three-dimensional reservoir models are best created with a combination of well logs and 3D-seismic data. However, the effective integration of these results is not easy because of limited seismic resolution. With the increasing quality of seismic data and wide application of new methods, high-resolution seismic-stochastic-inversion volume was used as a direct input to reduce the uncertainty of the reservoir model. Used as a direct input for reservoir modeling, this method reduces the uncertainty of the model greatly. Introduction A common method of reservoir modeling is the stochastic modeling method. The deficiency of this method is that it can be difficult to determine the changes between wells. In order to describe the reservoir changes between wells better, sedimentary facies and 2D-seismic attribute tendencies were used as constraints in the process of modeling. However, because every location has only one tendency value, it is difficult to describe the reservoir overlay. Method The work flows of this method, described in greater detail in the complete paper, essentially include three steps. The first step is target processing. Wavelet transform is applied to achieve noise elimination and resolution improvement. On the basis of the high-resolution seismic data, the second step is seismic stochastic in-version. After the process of time/depth conversion, the high-resolution 3D data from seismic stochastic inversion and well-log data were used as a direct input. Target Processing. In the first step, a wavelet-transform method was applied to achieve noise elimination and resolution improvement. The principle of this method includes the Hilbert spectrum for nonlinear and nonstationary time-series analysis. Seismic Stochastic Inversion. The algorithm of seismic stochastic inversion used in this study has the following stages: Initially, a sequential Gaussian simulation is performed on the well data of acoustic impedance, aimed at filling the volume to be inverted. The vertical and horizontal variograms are generated and modeled. Each knot of the grid is revisited randomly and resimulated. For each knot, a synthetic trace is calculated; for this, knowledge of the wavelet is needed. The synthetic trace is generated and compared with the original trace, where the residuals are calculated on the basis of the squares sum of the difference between the original trace and the synthetic trace.