Reservoir Characterization of Zamzama Gas Field by using Machine Learning and Artificial Intelligence

Abstract

The reservoir characterization is an important step in the exploration of oil and gas. Reservoir characterization enables the exploration scientists to estimate the properties of reservoir. This study is conventionally performed by using the seismic data along with well log data. These datasets provides the inFormation about the subsurface geology and the rock Formations. The advancement in computational techniques and increasing complexity of the hydrocarbon reservoirs motivated the geoscientists to introduce the advanced computational techniques like artificial intelligence and machine learning in the workflow of hydrocarbon exploration workflow. The machine learning techniques provides the efficient tool for the intelligent analysis of huge datasets of subsurface. It also automates the analysis process that greatly reduces the chances of personal error in seismic and petrophysical interpretation. The present study involves the seismic interpretation using the machine learning techniques. The prediction of missing well curves were also performed by using the available dataset. Random forest, Support Vector Machine, Decision Tree and Extreme Gradient Boosting (Xgboost) are implemented with Python programming language and reservoir characterization is done with Python programming language 3.9.0. This study employed a total of five wells, with extensive petrophysical interpretation performed first. The machine learning random forest system was trained on two selected wells in order to forecast the essential petrophysical parameters across the cube, resulting in an 80% match. Facies modelling was conducted out using these petrophysical volumes as input to the K-Mean clustering technique