The principal premise of this thesis is that the ambiguities of reservoir simulation can be and should be reduced by using time-lapse seismic data. Such data can be considered as a sort of reservoir dynamic data, with distinctive features compared to the typical reservoir production data. While well production data are sparse in space and dense in time, 4D time-lapse seismic can be utilized to fill the spatial data gaps between wells. This provides an opportunity to constrain reservoir dynamic behaviour not only at well locations but also between them by honoring time lapse response of the reservoir. This means that seismic assisted history matching should involve a simultaneous minimization of the mismatch between all types of measured and simulated data including seismic data.
This thesis is an effort to discuss critical aspects of integrating 4D time-lapse data in reservoir simulation and history matching. I have illustrated a detailed scheme of seismic assisted history matching with implications on real data, to emphasize the extra value that seismic data can bring into the conventional reservoir history matching. This goal was followed by developing a software application to assess the feasibility of the theory at industrial scales. In addition to the conventional oils, a significant effort has been devoted to extend the scope of the work to viscoelastic heavy oils and their fluid substitution models in thermal cases. I also studied the production/injection induced stresses impacts on anisotropic velocity variations, using coupled geomechanical-flow simulations.
The delineated scheme in seismic assisted history matching, proved to have a crucial contribution that cannot be ignored. The dynamic model for reservoir simulation should always be kept as simple as possible, but not simpler! This thesis is an argument that ignoring time lapse seismic data, builds a model which is simpler than a trustworthy model.