Application of Seismic Inversion and Machine Learning Techniques In Reservoir Evaluation and Optimization solution in Dimo Field, Western Niger Delta, Nigeria

Abstract

Characterisation of litho-facies and fluid types for well planning and development in heterogeneous reservoirs is  a very challenging task with no direct solution. However, in this study, we integrate various geophysical tools, namely rock-physic models, seismic inversion and machine learning. The aim among others is to characterise complex reservoir lithogical properties, static and dynamic properties of the reservoir, and to identify bypassed hydrocarbon accumulation channels for further well placement and development. The results of the rock-physics analysis via attribute cross-plotting show the P- impedance and density discriminated four major rock types in the reservoirs namely the sand, the shale, the sandy-shale and the shaly-sand. Sand shows values less than 2.09g/cc - 20000ft/s*g/cc while values greater than 20000ft/s*g/cc are indication shale. The sandy-shale and shaly-sand show density range of 2.09g/cc – 2.28g/cc but with different acoustic impedance ranges. Furthermore, the results of the seismic inversions and density derived probabilistic neural network (PNN) revealed a spatial distribution of reservoir lithofacies and fluid types with an anomalous increase in acoustic impedance and density values around the current well locations. This is probably an indication of brine, or CO₂ gas replacing oil in the reservoirs. The increase in brine saturations esecially around the well locations probably arose from production a related effect, which is a primary indicator of reservoir depletion. Hydrocarbon-charged sands with very low acoustic impedance and density were identified in the northwestern-southeastern trending channel and touted for further exploration consideration. The acoustic impedance attributes and PNN derived density attributes were compared and the results provide an excellent front end but PNN computed density appears more robust, definite and continuous in lithofacies and fluid discriminations.