Reservoir Quality Evaluation based on Integration of Artificial Intelligence and NMR-derived Electrofacies

Logarithmic Mean of Transverse relaxation time (T۲LM) and total porosity of the Combinable Magnetic Resonance tool (TCMR) are the main parameters of the Nuclear Magnetic Resonance (NMR) log which provide very substantial information for reservoir evaluation and characterization.  Reservoir properties, for example, porosity and permeability, free and bound fluid volumes, and clay-bound water, could be calculated through the interpretation of T۲LM and TCMR. In this manuscript, an intelligent approach has been used by us to predict NMR log parameters and their corresponding electrofacies from well log data. We define NMR electrofacies as classes of NMR log parameters representing reservoir quality are defined by us. For this purpose, NMR logs and petrophysical data are available for two different formations situated in the Ahvaz field. Data from Ilam formation were applied in order to construct the intelligent models, the same as Asmari formation, data were applied for reliability evaluation of the created models.  The outcome results reveal higher performance levels of the Neural Network (NN) technique compared to the neuro-fuzzy (NF) model. The synthetically generated T۲LM and TCMR logs are then calculated for the four logged wells from the Ahvaz oilfield using a mathematical function, and they are named Virtual Nuclear Magnetic Resonance (VNMR) logs. Finally, VNMR logs were classified into a set of reservoir electrofacies by cluster analysis approach.  Correlations between the VNMR electrofacies and reservoir quality based on porosity and permeability data helped evaluate the reservoir quality quickly, cost-effectively, and accurately.

Keywords:

Virtual NMR log , neural network , neuro-fuzzy , Back Propagation , Conventional Logs , NMR log parameters , Electrofacies