4D Monitoring of an Underground Gas Storage Case Using an Integrated HistoryMatching Technique

摘要

As time lapse seismic data are frequently available to guidernthe understanding of fluid flow behavior, a methodology hasrnbeen developed to produce a fine scale geological modelrnconstrained by time dependent data, wells and 4D in arnquantitative way. This enables us to keep the coarse modelrnused in fluid flow simulations consistent with the fine scalerngeological model [1] . The objectives of this work are tornvalidate the proposed methodology on a realistic syntheticrncase and to assess the impact of 4D in predictionrnimprovements. Are we doing a better job by using 4D inrnhistory matching?rnA synthetic reference case was derived from a Gaz dernFrance gas storage site where gas injection and production arernalternated over a 20-year period. Seven geological unitsrnshowing various degrees of heterogeneities describe this fullrnfield dynamic model. A fine scale reference geological modelrnwas built (9.7 million cells), then upscaled and used in flowrnsimulator. This model honored the observed static andrndynamic well data (17 wells). Saturation maps from fluid flowrnsimulation and a petro elastic model (PEM) were used tornderive acoustic and elastic parameters. This reference data setrnwill provide a point of comparison to assess and quantify thernprediction performances as the time period is divided intornhistory and prediction.rnA workflow was developed to build a consistent fine scalerngeological model, a fluid flow model and seismic data using arnPEM and accounting for the various scales. The forwardrnmodeling part of the workflow was used to compute simulatedrnwell data, impedances Ip, Is and time shift data. Then using anrnoptimization loop, the models were constrained to observedrnwell data and time lapse seismic data. The case study validatesrnthe methodology and shows an efficient way to includernquantitatively 4D information while keeping consistencyrnthrough the modeling chain. The model constrained by wellrndata and 4D seismic data lead to more accurate predictionsrnthan the model constrained by well data only.