| Summary: | Automatic history matching methods utilize various kinds of inverse modeling techniques. In this dissertation, we examine ensemble Kalman filter as a stochastic approach for assimilating different types of production data and streamline-based inversion methods as a deterministic approach for integrating both production and time-lapse seismic data into high resolution reservoir models. For the ensemble Kalman filter, we develope a physically motivated phase streamline-based covariance localization method to improve data assimilation performance while capturing geologic continuities that affect the flow dynamics and preserving model variability among the ensemble of models. For the streamline-based inversion method, we derived saturation and pressure drop sensitivities with respect to reservoir properties along streamline trajectories and integrated time-lapse seismic derived saturation and pressure changes along with production data using a synthetic model and the Brugge field model. Our results show the importance of accounting for both saturation and pressure changes in the reservoir responses in order to constrain the history matching solutions. Finally we demonstrated the practical feasibility of a proposed structured work- flow for time-lapse seismic and production data integration through the Norne field application. Our proposed method follows a two-step approach: global and local model calibrations. In the global step, we reparameterize the field permeability het- erogeneity with a Grid Connectivity-based Transformation with the basis coefficient as parameters and use a Pareto-based multi-objective evolutionary algorithm to integrate field cumulative production and time-lapse seismic derived acoustic impedance change data. The method generates a suite of trade-off solutions while fitting production and seismic data. In the local step, first the time-lapse seismic data is integrated using the streamline-derived sensitivities of acoustic impedance with respect to reservoir permeability incorporating ...
|
|---|