The Use of Machine Learning to Estimate Key Properties in Vaca Muerta Play

Abstract

Abstract The evaluation of shale reservoir requires solving key properties: total organic carbon (TOC), porosity (PHIT) and water saturation (SW). The determination of these properties with precision, requires laboratory data and high-tech logs. In Vaca Muerta play, the most legacy wells do not have this data. This work presents the Machine Learning as a methodology to estimate TOC, PHIT, cementation factor (M) and saturation exponent (N) with basic logs. Synthetic logs are generated with different predictive models, taking readily available conventional wire logs as input data (Resistivity, Density, Sonic and Neutron). Regarding of PHIT and saturation parameters (M and N), the models are trained with wells in which these logs are available; in the case of TOC, measures in core are used. Once the target log has been defined, an exploratory analysis is carried out. These results feed the machine learning models. The different models are trained and validated, to obtain the best result. For each synthetic log, a 90% confidence interval is also calculated. Linear and non-linear models are developed, and their effectiveness is measured by dividing the data randomly, using 80 % of the data as a training set and the remaining 20% as validation set. Moreover, the validation "leave one out" is also performed. The models are applied to more than 170 vertical wells, in all cases, the synthetic curve is accompanied by a confidence interval. The results of these confidence intervals show that the synthetic logs have a precision that makes them reliable and relevant for decision-making by the specialist. The implementation of this technique provides relevant information for landing zone definition, making maps and volumetrics of greater areal detail. The application of machine learning techniques for the generation of synthetic logs is a procedure that has recently begun to be used in the oil and gas industry. In the particular case of the Vaca Muerta oil field, the series of synthetic logs developed and tested in this work is a complete novelty.