Publications

Proper management of a hydrocarbon reservoir requires detailed knowledge about critical reservoir properties, such as the distribution of effective reservoir rocks. Several state-of-the-art geostatistical techniques (e.g. Block Kriging and Collocated- CoKriging) have recently been used to construct a full-field, three dimensional (3-D) model of Pedada Field in Central Sumatra. The successful application of these reservoir characterization techniques has led to significant improvements in the design of a waterflood for the field. The model has also been an important tool to guide other reservoir management activities, such as drilling attic and infill wells.Construction of an accurate 3-D reservoir model requires the analysis and integration of many types of data. These data include wireline measurements, core measurements, engineering and 3-D seismic data. An overview of the integration scheme discussed in this paper is shown in Figure 1. The implementation of this process is challenging but important. Statistical analysis of the core and log data has been integrated with geostatistical analysis and modelling of the seismic/data. This combination has lead to a reservoir characterization model in which the large aerial distribution of the 3-D seismic data is effectively combined with the high vertical resolution inherent in log and core data. Well log (and particularly core) data are much more sparse and spatially limited than seismic data.Successful reservoir characterization is critically dependent upon the quality of the available input data. Log and core measurements of reservoir parameters are traditionally considered to be more accurate and reliable than seismic measurements. However, a key observation from Pedada is that ",hard data", from logs contains uncertainties that should be modelled, and these data may not be more certain than so called ",soft data", from seismic. A comparison between an algofithmically-guided analysis of log data and a human analysis demonstrates some uncertainties inherent in log interpretation. Therefore, the objective is the construction of a statistically accurate reservoir model, integrating the most significant data from various sources and honoring the modelled data uncertainties.