Modélisation Géostatistique d’un Réservoir Pétrolier sous R

Abstract

This thesis focuses on the geostatistical study of an oil reservoir located in the Berkine Basin (Algeria), based on data collected from 62 wells. The objective is to model the spatial variability of subsurface petrophysical properties (effective porosity, permeability, water and hydrocarbon saturation, clay volume, thickness, and net volume) in order to optimize development decisions. The methodology is based on three main steps: - An exploratory analysis in R (descriptive statistics, visualization, PCA) is used to characterize the distribution of variables and their correlations. - A modelling of the spatial structure using experimental variograms, fitted with theoretical models (spherical, exponential), for each variable. - An interpolation using ordinary kriging and conditional simulations produces predictive maps and uncertainty maps, identifying high-potential zones. R software proved to be an essential tool: free, powerful, reproducible, and rich in specialized libraries (gstat, geoR, raster, tmap). It enabled efficient data processing, model fitting, and results visualization. This work leads to practical recommendations to guide future drilling campaigns and improve reservoir management.