عنوان مقاله [English]
نویسندگان [English]چکیده [English]
During the last decade, there has been an increasing interest in the use of attributes derived from 3-D seismic data to define reservoir properties, such as the presence and amount of porosity and fluid content. Therefore, it is worthwhile to continue the advances in the study and application of expert systems in the petroleum industry so that it is possible to use the attributes in reservoir characterization more effectively. The establishment of the existence of an intelligent formulation between two sets of data (inputs/outputs) has been the main topic of such studies. One such topic of great interest was the characterization of 3D seismic data with relation to lithology, rock type, fluid content, porosity, shear wave velocity, and other reservoir properties. Petrophysical parameters, such as water saturation and porosity, are very important data for hydrocarbon reservoir characterization. Hitherto, several researchers endeavored to predict them from seismic data using statistical methods and intelligent systems (Russell et al., 2002; Russell et al., 2003; Chopra and Marfurt, 2006). Correct recognition of porosity model and estimation of petrophysical parameters in reservoirs is a key issue in any oil project. The correct estimation of porosity as a petrophysical parameter can inform decisions that have high financial risk, such as drilling. By determining reservoir characterizations and assessing petrophisical parameters with a adequate accuracy during the first steps of studies, researchers would be able to produce optimum exploitation with a minimum number of wells.
This paper focuses on the link between seismic attributes and reservoir properties such as lithology, porosity, and pore-fluid saturation. Typically, seismic attributes have been the only information obtainable from seismic data. Using statistical rock-physics, the type of seismic attributes that are direct functions (analytically defined) of the elastic properties can be probabilistically transformed, sample-by-sample and independently one of each other, into reservoir properties. In this paper, we combine the methods of geostatistics and multiattribute prediction for the integration of seismic and well-log data, and illustrate this new procedure with a case study. A number of new ideas are developed for the statistical determination of reservoir parameters using seismic attributes, combining the classical techniques of multivariate statistics and the more recent methods of neural network analysis. We first extract average porosity values at the zone of interest, and then compare these values to average seismic attributes over the same zone. The technique of cross-validation is subequently used to show which attributes are significant. We then apply the results of the training and cross-validation to data slices derived from both the seismic data cube and the inverted cube to produce an initial porosity map. Finally, we improve the fit between the well log values and the porosity map using co-kriging.
The main purpose of this paper is to present a quantitative assessment of porosity as a petrophisical parameter in an offshore oil field in Iran using the newly proposed method of reservoirs parameter estimation. This paper shows that by using both seismic data and well logging data it is possible to obtain a more accurate model of porosity in a given reservoir. Specifically, the study determines the relationship between a set of seismic attributes and a reservoir parameter such as porosity at well locations, and then uses this relationship to compute reservoir parameters from sets of seismic attributes throughout a seismic volume. Therefore, a primary plan of porosity is available for the area of study. In the next step, by using geostatistics and, according to the initial plan, as a secondary variable in collocated cokriging, we can approach a more accurate plan to show the distribution of porosity. In effect, the proposed method combines geostatistics with multiattribute transforms. This technique uses multivariate statistics and neural networks to improve the secondary dataset used in the collocated cokriging technique.