History-matching with Ensemble-based Methods - Application to an Underground Gas Storage Site (SPE 154475)

This paper shows the application of two ensemble-based assimilation methods to constrain an underground gas storage site operated by GDF-Suez to well pressure data. The methods considered here are the Ensemble Kalman filter (EnKF) and the Ensemble Smoother (ES). The EnKF is a sequential data asssimilation method that provides an ensemble of models constrained to dynamic data. It entails a two-step process applied any time data are collected. First, the production response is computed for each model of the ensemble at the following acquisition time. Second, models are updated using the Kalman filter to reproduce the data measured at that time. The EnKF has been widely applied in petroleum industry. More recently, the ES was used successfully on real field cases. This method is also based on the Kalman filter, but the update is performed globally over the entire history-matching period: values simulated at each assimilation time are considered simultaneously in the update step. The multiple restarts necessary with EnKF are thus avoided. We present here an application of these methodologies for constraining an underground gas storage site to well pressure data. The uncertain parameters are the porosity and horizontal permeability values populating several layers of the geological model. Both methods yield a good match of pressure data in the history-matching and prediction periods. For ES, this can require two successive applications. Considering the same initial ensemble, the ES leads to a smoother mean with less extreme values and a higher variance. The EnKF and ES methodologies turn out to be powerful tools to constrain geological models to dynamic data, and were applied successfully to a real field. The ES gives here similar results in terms of match and predictions while preserving a higher spread within the model ensemble with less extreme petrophysical values.