Carbonate reservoir characterization using sequential hybrid seismic rock physics and artificial neural-network: a case study of North Tiaka Field

摘要

Tiaka field is located in Senoro-Toili block at eastern arm of Sulawesi, Indonesia. The mainrnhydrocarbon bearing reservoir of this field is lower Miocene carbonate sequences which has dualrnporosity system (matrix and fracture). Actually, the carbonate characterization is complex. It is notrnonly caused by the complexity of pore system but also because of the influence of chemical reactionrnproduced from fluid interaction in interior wall of their pores space. It makes wave propagationrnsystem in carbonate becomes complex. Therefore, special treatment is required to preciselyrncharacterize this complexity.rnThis paper presents the latest technology for carbonate complex reservoir characterization usingrnhybrid seismic rock physics, statistic and artificial neural network. This methodology makesrnintegrating a huge size of various data to produce “coherence correlation” among input (seismic, wellrnlog data) and their target properties (porosity, saturated water, fracture) possible. This method isrnapplied in North Tiaka Field to predict the lateral lithofacies, fracture, porosity, and their fluid orrnhydrocarbon distribution. By using this approach, high accuracy on the reservoir parameter predictionrncan be produced.rnThe blind test well shows that predicted properties on an average 90 percent match with reservoirrnparameter in the existing wells.rnThe available data of this field consist of core information (i.e: lithology, lithofacies, fracturernintensity, fracture width, porosity), well log data (i.e. gamma ray, density, saturation of water,rnporosity, resistivity etc.), multi-attribute seismic either pre-stack or post-stack 2 D seismic lines andrnseismic rock physics. The whole input data was trained together and reservoir parameter wasrnpredicted by using natural algorithm based on lithofacies prediction result which is combined withrnstatistic and artificial neural network. Therefore the lithofacies prediction is the first task which shouldrnbe done before characterizing the other properties of reservoir. And afterward, it is used to predictrnseveral reservoir parameters.