Integrated Sedimentary Analysis and Facies Classification Using Artificial Neural Network: Case Study from Fluvio-Deltaic to Nearshore Clastic Sequences of Central Sumatra

Eko Rukmono², Doni Hernadi², Indrajit Bandyopadhyay¹, Dedi Juandi¹, and Sangeeta Singhal^1
1DCS, Schlumberger, Beijing, China
²Development, BOB PT. Bumi Siak Pusako - Pertamina Hulu, Jakarta, Indonesia

Growing demand of hydrocarbons and absence of new giant discovery is continuously focusing exploration and development towards complex stratigraphic traps, which is enhancing requirement of detailed knowledge of morphology and internal characteristics of stratigraphic bodies. High resolution borehole images provide detailed insight to rock fabric enabling identification of minor features, which help to position a unit within specific depositional system. But to achieve regional consistency a classification scheme is to be established. Artificial neural networks (ANN) enable development of quantified classification schemes. ANN can classify units based on features and assign them into appropriate depositional system. This technique solely based on numerical computation may fail if not properly calibrated with real rocks. So, right process is to generate a first pass of classification using real rocks, which is then modified to maintain coherency with log signatures. ANN generated with this modified classification will have benefits of numerical computations with real rock signatures still preserved in it. Similar process was used in fluvio-deltaic to nearshore clastic sequences in Riau Province in Central Sumatra Basin. Borehole images from six wells were analyzed to identify paleocurrent direction, sedimentary features and depositional processes. From these depositional facies were derived, which were calibrated with core and mudlog. Special algorithms were used to convert morphological attributes on images to depth indexed logs. Depositional facies classification was modified by comparing with standard and morphological attribute logs. Modified classification was used to build ANN which in turn successfully derived facies classification in all wells. The paper illustrates the entire process highlighting both classification schemes and evolution of one from other.

Presentation GEO India Expo XXI, Noida, New Delhi, India 2008©AAPG Search and Discovery