Abstract

This thesis investigates the development and application of a geographic information system (GIS) and data mining methods for reservoir characterization and horizontal well placement guidance acquisition. Reservoir characterization is a process of quantitatively assigning reservoir and fluid properties while recognizing geologic uncertainties in spatial variability. To identify reservoir properties with spatial correlation, a new density-based spatial clustering method, SEClu, is presented to group core analysis data. Further, a novel fuzzy ranking artificial neural network (FR-Neural) framework is introduced for accurately characterizing reservoir properties from well log data. SEClu and the FR-Neural framework are evaluated with synthetic and real datasets.

Horizontal well placement guidance acquisition (HWPGA) analyzes the real field data and collects guidelines for placing horizontal wells into a reservoir. In this thesis, a group of horizontal well placement attributes are defined to capture the location of horizontal wells in a heterogeneous reservoir. A customized association rule mining method, named SE-Apriori, is introduced to analyze the influences of the horizontal well placement attributes on the oil production. The SE-Apriori considers two predefined constraints from the HWPGA problem and, thus, can generate fewer association rules with less execution time. A GIS prototype containing the SE-Apriori tool was developed to help in efficiently managing petroleum field data and visualizing the association rule mining results on a map. Finally, the proposed SE-Apriori method is evaluated using a real dataset from a steam assisted gravity drainage (SAGD) project in Alberta, Canada.