A New Look at Well-to-Well Correlations using Nonparametric, Nonlinear Regression

Abstract

In this paper a nonparametric regression method, Alternating Conditional Expectation (ACE), was applied to production data from the Palinpinon field in the Philippines. The method reveals an interesting nonlinear correlation between the injected flow rate and produced concentration for a number of injector-producer pairs. In order to evaluate the ACE approach, we applied it to a subset of the Palinpinon data set and checked the results by using cross-validation. The nonparametric transformations produced by ACE were used to predict future concentration values. The predictions were compared to measured values with satisfactory results in some cases - for other cases the predictions were not good.

The approach presented here takes a simplified view of the physical model describing flow through fractures with time-varying flow rate and concentration. The shortcomings of the approach are discussed and alternative ways of using ACE to reveal the well-to-well connectivity are suggested.