An effective field scale reservoir model for history matching and reservoir prediction. Part 2. Parametrical upscaling and calibration

The history matching problem in connection to an effective reservoir model concept is defined as a multi-well data inversion problem for a class of process-driven deterministic reservoir models. A two-stage strategy of the production data inversion has been introduced and considered in detail. The strategy includes parametrical upscaling as a pre-processing part and model calibration as a final part. The computer aided approach has been recommended for the parametrical upscaling. It is based on combination of standard step-wise local optimization techniques and active expert-user interface. The fully automatic approach is suggested for the model calibration process. The relevant data inversion algorithm is developed, which combine multi-start strategies (Evolution and Simulated Annealing) with the multi-start mode of a gradient method. The pilot computer programs are created and tested on 2D testbed and on multidimensional synthetic prototypes of a real history matching problem. The results are analyzed based on set of parametrical upscaling and calibration sessions