Probabilistic framework-based history matching algorithm utilizing sub-domain delineation and software 'Pro-HMS'

The key idea presented in this thesis is the perturbation of the probability distribution describing the uncertainty in the permeability value within sub-domains of the reservoir using the dynamic information. Probability deformation parameters are defined and optimally established to minimize the deviation of the predicted production response from the observed or measured response. To implement the perturbation method in a parallel computational environment, sub-domains have to be established such that they are the most influential regions and least correlated. In that case the perturbations for establishing the deformation parameters can be done independently in the sub-domains. In contrast to current history matching algorithms that incur significant computational cost in optimizing permeability for each gridblock in the simulation model, our approach reduces the optimizing period by employing innovative method such as sub-domain delineation using principal component analysis (PCA) and a reduced parameter set in the form of deformation parameters that have to be optimized. Once the domains are delineated, a fully probabilistic approach to perturb permeability within the delineated zones is implemented. With this method the information inferred from dynamic data is merged with the prior geologic information. The described process of history matching is effective; however, it can be cumbersome because it combines aspects of stochastic reservoir modeling, flow simulation and iterative optimization. To speed up and enhance the efficiency of history matching process, and to render a streamlined process, an integrated software package Pro-HMS (Probabilistic History Matching Software) was developed. Various synthetic and real field data sets were processed with Pro-HMS, and they show reliable predictions for future reservoir performance.