IBP - Rio Oil & Gas Expo and Conference 2020
Despite the maturity of geological modeling techniques using geostatistics, there are situations, as in the Brazilian pre- salt, where these traditional methods do not produce a good representation of the reservoir properties due to the high degree of heterogeneity of the rocks. These limitations, however, can be overcome using modern Machine Learning techniques widely used in other industries. Unfortunately, due to the more recent development of such techniques, there are still few geoscientists with the necessary computer and mathematical skills to apply them in the E&P workflow. To overcome this challenge, we developed a platform – MachLee – to facilitate access to the most modern machine learning technologies and assess the quality of predictions. MachLee enables loading and preparing the subsurface data and automating the selection, classification, and parameterization of machine learning algorithms, generating a prediction solution ready to apply. The use of this platform is illustrated through an example in a Santos basin pre-salt field where well properties were predicted based on learning about the available data.
