Using NMR logging and ANN to estimate permeability in one of heavy oil reservoirs in the south of Iran

Authors

Abstract

Permeability is a property of porous medium that quantifies rock capacity to transmit fluids. Frequently, core based permeability data are not available either because of the borehole conditions or due to the high cost of coring. For these reasons, over the years attempts have been made to estimate permeability by alternative ways. Permeability is an elusive parameter in hydrocarbon reservoirs as it is very difficult, if not impossible to determine precisely and directly from current subsurface logging technologies. In this research, an attempt is made to test some methods for estimating permeability as a function of depth from Nuclear Magnetic Resonance (NMR) logging in one of the carbonate reservoirs, bearing heavy oil, in the south of Iran.
NMR uses hydrogen protons as an indicator of fluid presence. Not all nuclei possess the ability to interact with magnetic fields (magnetic moment); only those with an odd number of protons, such as hydrogen, possess magnetic moment. In calculating permeability from NMR logging in the upper Sarvak formation, three models such as average-T2, free-fluid and Swanson model have been used. Permeabilities obtained by these models are compared with core permeability and correlation coefficients are calculated which give poor results. It can be considered that the trends of permeabilities obtained by NMR models have good compatibility with core permeability, so they can be used for in-situ permeability estimation.
A hydraulic unit (HU) is a reservoir layer or zone that has similar average rock properties affecting fluid flow. By using hydraulic units, the samples are grouped into distinct units with clear porosity and permeability properties. We are trying to consider the effect of using hydraulic units in obtained correlation coefficients. Hydraulic units can be determined from core porosity and permeability. This technique calculating the flow zone indicator (FZI) from the pore volume to solid volume ratio (?z) and reservoir quality index (RQI). Based on core porosity and permeability 3 hydraulic flow units for the desired interval were distinguished. The results obtained from hydraulic unit 3 cannot be reliable because the small amount of samples distinguished in this unit.
For accurate permeability estimation, an Artificial Neural Network (ANN) model with two different sets of inputs is applied. In the first case, porosity obtained by NMR logging has been used as one of the input data but in the second case there is no NMR data as an input and core porosity has been used as one of the input data. The predicted permeability by ANN with both sets of input data is then compared with the core permeability. The results show that the correlation between predicted and core permeabilities is very good when the porosity obtained by NMR logging is used as one of the inputs of the ANN model. Using hydraulic units results in increasing obtained correlation coefficients for both NMR models and ANN model.
Carbonate formations are more complex than sandstone formations. This is due to the broad range of pore sizes, complex pore structures and low surface relaxitivity values found in carbonate formations. Using ANN models will lead to better results compared with traditional NMR models because ANN models can consider complex relations between permeability and NMR parameters.

Keywords