This paper presents a fast, reliable multi-objective history-matching method based on proxy modeling to forecast the production performances of shale gas reservoirs for which all available post-hydraulic-fracturing production data, i.e., the daily gas rate and cumulative-production volume until the given date, are honored. The developed workflow consists of distance-based generalized sensitivity analysis (DGSA) to determine the spatiotemporal-parameter significance, fast marching method (FMM) as a proxy model, and a multi-objective evolutionary algorithm to integrate the dynamic data. The model validation confirms that the FMM is a sound surrogate model working within an error of approximately 2% for the estimated ultimate recovery (EUR), and it is 11 times faster than a full-reservoir simulation. The predictive accuracy on future production after matching 1.5-year production histories is assessed to examine the applicability of the proposed method. The DGSA determines the effective parameters with respect to the gas rate and the cumulative volume, including fracture permeability, fracture half-length, enhanced permeability in the stimulated reservoir volume, and average post-fracturing porosity. A comparison of the prediction accuracy for single-objective optimization shows that the proposed method accurately estimates the recoverable volume as well as the production profiles to within an error of 0.5%, while the single-objective consideration reveals the scale-dependency problem with lesser accuracy. The results of this study are useful to overcome the time-consuming effort of using a multi-objective evolutionary algorithm and full-scale reservoir simulation as well as to conduct a more-realistic prediction of the shale gas reserves and the corresponding production performances.

This paper evaluates the estimated ultimate recovery for 10-year operation at a shale gas reservoir, implementing FMM (Fast Marching Method) as a surrogate model of full-scale numerical simulation and Monte Carlo simulation as a tool for accessing the uncertainty of FMM-based proxy parameters. Sensitivity analysis shows the significant properties affecting the gas recovery that are enhanced permeability, matrix permeability, and porosity in sequence. Using the statistical distributions of these parameters, this study determines P10, P50, and P90 of the 10-year cumulative gas production and compares them with the values from full-physics simulations. The computing time based on the proxy model is much smaller than that of the full-scale simulations while the prediction accuracy is acceptable. FMM can forecast the production profiles reliably without time-consuming simulation and the integration of Monte-Carlo simulation is able to evaluate the uncertainty of gas recovery, quantitatively.