Perpustakaan Digital ITB

The focus on this thesis is the application of machine learning to predict porosity and permeability using core and well log data. This study analyses and describes a workflow that aims to develop a Machine Learning model for predicting permeability and porosity in uncored wells utilizing core measurements integrated with available log data as a feature to train. Different machine learning models including random forest, GradientBoost, artificial neural network, AdaBoost, SVM, XGBoost, and lasso regression are employed to have a comprehensive comparison. The prediction results gained from the machine learning models used in this paper are compared to the relevant real petrophysical data and empirical method to demonstrate the advantages and disadvantages of using artificial intelligence to predict reservoir characteristics. Both result for porosity and permeability prediction obtain machine learning as the best method with the best algorithm for porosity prediction is GradientBoost that gives R2 score of 0.847, while permeability prediction obtain the best R2 result of 0.771 by utilizing Support Vector Machine algorithm and transforming some of feature and target variable to logarithmic values. The results reported in this thesis indicate that implication of machine learning methods in porosity and permeability estimations can lead to the construction of more reliable reservoir characteristics prediction.