Rong Lu

In this work, a data-driven approach is taken to tackle problems in Petroleum Engineering domain, for both conventional and unconventional reservoirs. Conventional reservoirs face the problem of losing energy for flowing after a few years of production, thus operators choose to inject water and inject CO$_2$ as a secondary and tertiary recovery method. The question of interest is that how injection scheme correlates with production responses. As shown from this work, supervised learning (support vector machines) can answer the question and come up with predictive models. On the other hand, unconventional oil and gas resources development has gained much more attention since the last decade, due to the advancement in hydraulic fracturing (HF, or ``frac'') technology. In order to develop shale gas reservoirs, which have extremely low permeability, HF has to be applied. In the process fluids and solids under high pressure are pumped into the formation to break the rock. As fractures are created, more reservoir contact are obtained and the shale gas would flow through the fractures to the wellbore. Two questions the industry are interested in are, where to frac the wells in unconventional shale plays, and with so many completion and stimulation parameters whether there exists any hidden patterns. The two aspects are approached by both supervised (linear regression) and unsupervised learning (cluster analysis) in the following.