EnSURE: Ensemble Subsurface Understanding - Robustness and Efficiency for increased efficiency in the energy transition

Knowledge of the subsurface is critical to successful field development and reservoir management, including improved reservoir drainage and water management, reduced energy use, and better decision-making in general. A computer model of the reservoir allows one to perform numerical experiments that may be essential for estimating the result of drilling a new well or converting an oil producer to a water injector. Because it is impossible to understand the subsurface completely, a collection of different reservoir models can be used to provide an estimate of uncertainty in reservoir properties and in reservoir behavior.

In general, reservoir models that are capable of reproducing historical data are more reliable for forecasting future behavior than models that are inconsistent with previous observations. Typical observations include items such as amplitudes of reflected seismic waves, well water production rates, and well shut-in pressures. Integration of all relevant types of measurement provides maximum information for subsurface understanding and generally reduces uncertainty. However, when inappropriate assumptions are made, or when important features of a model are omitted, there will be a loss of information and the resulting forecasts will be both biased and overconfident.

This project addresses the need for calibrated reservoir models that are able to provide robust reservoir forecasts and uncertainty quantification. We target the challenge of efficiency of the algorithms so that calibration of model parameters can be achieved faster when new data are acquired. We target the challenge of robustness so that bias in forecasts is reduced and a more comprehensive spectrum of potential outcomes is obtained. We achieve these targets through the development of a workflow that includes model and data checking and provides greater insight. Finally, we apply the methodologies developed in this project to real field cases to ensure that the methods will be useful.