Analysis of Well Production Data Using Functional Data Analysis

This study employs novel ensemble-based statistical techniques for type-well analysis to quickly analyze the production data from multiple wells in a reservoir. The method is based on functional principal component analysis (FPCA) that accounts for measurement noise and the sparsity of the production timeseries. In particular, the sparse FPCA method is used, which can extract the underlying random process from an ensemble of sparse production timeseries to predict the smooth trend of the well production data. The production from wells with short histories is also extrapolated using the stochastic information extracted from the wells with longer production. This approach is applied to analyze production data from 500 wells in an unconventional tight oil reservoir. In addition, the multivariate FPCA is utilized herein, for the first time, to jointly project the simulated surface condensate and gas rates from an unconventional gas condensate model by accounting for the correlations between the production phases. This approach is utilized to effectively replace the full numerical simulator by alternatively using an efficient timeseries proxy that can generate the full simulation output for any given set of input variables almost instantaneously.