Methodological Development for the Comparison of Multiple Machine Learning Algorithms in the Reconstruction of Well Loggings in the Middle Magdalena Valley Basin in Colombia

Abstract

In the Middle Magdalena Valley Basin (MMVB) in Colombia, there is the Mugrosa formation of high oil interest, since large hydrocarbon reserves accumulated in it. of the various wells that were drilled in the area, the three of main interest for this research are Tenerife 1 (T1), Tenerife 2 (T2) Of the various wells that were drilled in the area, the three of main interest for this research are Tenerife 1 (T1), Tenerife 2 (T2) where the dipolar P wave (DTCO), dipolar S wave (DTSM), and Gamma-Ray (GR) well logs were partially registered, and Tenerife 3 (T3) that does not have such logs. The lack of this information becomes an obstacle for present and future research and industrial projects in this and other areas of the country, since completing them by traditional methods is technically and economically unfeasible.The present research proposes the implementation of a range of Machine Learning (ML) methods such as Random Forest, KNN, Gradient Boosting, AdaBoost, Multi-linear Regression, and Artificial Neural Networks, to provide an economically viable solution. to this problem, since the computational alternatives proposed in this work present a much lower cost than the entire technical and logistical process of obtaining records traditionally.