The Caddo Sequence of the Boonsville Field (Texas, USA) has been characterized combining Similarity Analysis and unsupervised Neural Networks. The main goal was to identify and delimit patterns, which could be associated with prospective hydrocarbon areas and other petrophysical properties of the zone. For the Similarity Analysis, twenty wells with different production characteristics were selected as reference points. Interval attribute maps were used to calculate the mean and standard deviation around these reference points, taking several attribute traces around them. The comparison of these values with the rest of the traces of the seismic attribute maps leads to the generation of binary maps. These were added to obtain the similarity maps for each reference point. Unsupervised Neural Network maps were obtained after classifying the seismic data using both waveform and seismic attributes, as inputs. Although some similarity areas coincide with various seismic facies identified by Neural Networks, it seems that the methods are delimiting some zones that respond to different factors. In fact, the integration of well log analyses, as well as seismic attributes interpretation, supports this and indicates that both techniques give complementary information. Both methodologies respond to four factors that include the presence of fluid, the type of fluid (gas or oil), the type of reservoir rock (sandstone or limestone) and the sedimentary environment. These results points to the integration of different techniques and data for a better reservoir characterization.
Presentation Date: Tuesday, October 13, 2020
Session Start Time: 9:20 AM
Presentation Time: 11:25 AM
Location: Poster Station 9
Presentation Type: Poster