Prediction of Dead Oil Viscosity: Machine Learning vs. Classical Correlations

Dead oil viscosity is a critical parameter to solve numerous reservoir engineering problems and one of the most unreliable properties predict with classical black correlations. Determination dead by experiments expensive time-consuming, which means developing an accurate quick prediction model required. This paper implements six machine learning models: random forest (RF), lightgbm, XGBoost, multilayer perceptron (MLP) neural network, stochastic real-valued (SRV) SuperLearner viscosity. More than 2000 pressure–volume–temperature (PVT) data were used for testing these models. A huge range used, from light intermediate heavy oil. In this study, we give insight into performance different functional forms that have been in literature formulate The results show form f(γAPI,T), has best performance, additional correlating parameters might be unnecessary. Furthermore, outperformed other (ML) algorithms as well common correlations are based on metric analysis. can potentially replace empirical models predictions wide viscosities (any type). Ultimately, proposed capable simulating true physical trend variations API gravity, temperature shear rate.