Abstract

Just over a decade after the 1st shale revolution, the Permian Basin sits on the verge of another revolution. Laboratory results are suggesting that recovery factors could increase from roughly 12% to >50% with EOR. The implications of this value creating revolution are huge and stand to further tip US energy production in America’s favour. However, the Permian Basin is still wrestling with trying to find the optimum spacing and completion method that maximizes value. Optimal well spacing to maximize value at the granular level is a unique solution for each well and is dependent upon geology, geophysics, geomechanics, well bore geometry, proximity to other well bores, drilling engineering, and completion engineering, price of oil and cost of inputs. Within these typical datasets there are literally thousands of variables yielding unwieldy sized solution spaces with sparse datasets. Many companies are turning to non-parametric modeling to make predictions in data where parametric models have failed. These non-parametric methods are very quickly hitting the limits of the data which are thousands of variables with only hundreds of samples. This ill-posed problem results in poorly constrained endpoints that increase the probability of having failed to realize maximum value. This talk is about solving these problems together and what that framework is starting to look like.