NMR Measurements in Carbonates Samples and Core-Logs Correlations Using Artificial Neural Nets

Abstract

Abstract The Mara West Field reservoir in the Maracaibo Lake presents some challenges in petrophysical evaluation, which are related to the determination of the water saturations, porosities and permeabilities. The information extracted from conventional log data is not sufficient to fully evaluate the formation. This is due to the fact that these logs are affected by the presence of carbonates, difficulting the evaluation of porosities and other properties. In order to overcome this problem, Nuclear Magnetic Resonance technique has been used in the evaluation of core plug samples and further core logs correlations. Its advantage lies in the measurement of porosity independently of lithology and pore distribution and in the calculation of permeability using parameters such as Nuclear Magnetic Resonance porosity, T2 cutoff, free fluid index and bound water volume. The present work shows a very good correlation between Nuclear Magnetic Resonance data and conventional data extracted from core plug samples. It also shows how the implementation of a new equation for the estimation of permeability, provides an increment of the correlation coefficient from 0.36 to 0.81 obtained from the Nuclear Magnetic Resonance permeability Vs Klinkemberg permeability plot. In order to complete the log information in the cored well and in a neighbor well, Artificial Neural Nets based on the backpropagation algorithm have been applied, generating the so called pseudologs (synthetic logs). To obtain the pseudologs of NMR porosity, gamma ray, resistivity and neutron logs have been used as input, while the core plugs value for NMR porosity and permeability as output. The results show a very good match between the existing data and the artificial neural net output in the validating data set. Introduction The permeability and the irreducible water saturation are fundamental properties used in the reservoir evaluation. It cannot be measured with well logging; therefore their determination is made from conventional logs, using equations and correlations that are not very satisfactory in many cases. A new method to determinate the permeability have been proposed using porosity and NMR parameters as the Free Fluid Index, Bound Fluid Volume [1,2,3], or mean T2 [4]. Although these are empirical equations, their relationship with the permeability measured with conventional methods in the laboratory is acceptable. This paper presents a novel equation for carbonates that increment the correlation coefficient from 0.36 to 0.81. This equation has been proved also in sandstones, yielding good results too. However, all the permeability equations include parameters which need to be adjusted in the laboratory according to the reservoir characteristics. Another important step is the core-log correlations and the extrapolations of the core properties to the neighbor wells. Recent applications of the techniques of Artificial Neural Nets have given a reasonable solution in the present case study of the Mara West Field. The Artificial Neural Nets used in this work is based on the backpropagation algorithm described frequently in the literature [5,6]. The input data is taken from well logs and the output data from NMR experiments on core samples. Once the Artificial Neural Net is trained in a cored well, the algorithm is used in estimating pseudologs of NMR porosity, BFV and NMR permeability in the uncored well.