Analysis of Variance Application in Reservoir Simulation: A Case Study

In petroleum engineering, reservoir systems are described by highly nonlinear parameters. Accurate and fully-detailed simulations are very expensive and time consuming specially when uncertainty in material properties makes such systems even more complex. In order to account for uncertainty analysis, adequate high quality data sets are needed. Lacking these or a full reservoir simulation, proxy models are employed to work instead of a numerical simulator. They are employed in a wide range of workflows in petroleum engineering such as risk analysis, history matching, sensitivity analysis, reservoir characterization, field development planning, and process optimization. However, one important step in applying the proxy models is proper selection of effective parameters that have a major impact on reservoir performance. Analysis of Variance (ANOVA) can be applied to determine the effective parameters in proxy models. This paper aims to represents a novel method to choose effective parameters among various existing parameters as an input to proxy models. We use a design of experiment technique i.e. fractional factorial design to determine the effective parameters out of different existing parameters. Effective parameters chosen by this method may be further used as the only the input parameters into proxy models. Result show that three parameters become effective among eight. It is also observed that effectiveness of parameters is time independent for the studied reservoir.