Well placement in tight carbonate reservoir with targeting thin karst is very challenging, especially to
locate and map karst distribution and to identify the good spot. The nature of heterogeneous carbonate
reservoir and limitation of seismic resolution are the subjects to be understood and resolved prior to
Seismic ML is a robust method and fast to generate new insight of seismic resolution for interpretation,
maximizing and extracting tens of seismic attributes to be used simaultaneously in predicting and
determining porous zones which are associated with karst.
Seimic ML has been tested in AA field with the objective to optimize well placement into sweet spots
associated with porous karst. The results show that ML has capability to predict the porous zone target
which is reflected by losses zone during drilling operation of development well AAD-1.
In the development phase, detailed carbonate reservoir characterization is the key to understanding the
properties of carbonate and its origins. Carbonate rocks are categorized as unconventional reservoirs that
have heterogeneous porosity and permeability. The heterogeneity is caused by depositional and
diagenetic environments (Jardine, D. and Wilshart, J. W., 1982).
Karst is a target in oil and gas development well which have excellent porosity and permeability. The
geometry of karst is thin, spatially uncertain, and below seismic vertical resolution.
The pilot study of seismic machine learning was applied to resolve those problems. The objectives of this
study are to locate and map porous zones associated with karst and to improve, optimize, de-risking well
placement of infill well. The result showing that seismic ML is enhancing the sweet spot of target zone to