Applying an optimized low risk model for fast history matching in giant oil reservoir

Abstract EN

History matching is still one of the main challenging parts of reservoir study especially in giant brown fields with lots of wells.

In these cases, history matching with conventional manual technique needs many runs and takes months to get a match.

In this paper the latest approaches for automated history matching (AHM) were applied to a real brown field with 14 active wells with multiple responses (production rate, bottom hole pressure and well block pressure) located in south part of Iran.

Modified support vector machine was employed to create proxy model in which 44 model parameters were incorporated based on design of experimental.

Thereafter, all model parameters were adjusted to reproduce the observed history within the created proxy model.

A robust framework for building the proxy model was programmed with data exchange ability between commercial reservoir simulator software and the proxy model routine.

Accordingly, the proposed proxy model was successfully constructed using 1086 samples based on R2 coefficient of about 0.9 for the trained and test dataset.

Finally, the process was optimized by two main algorithms for reaching best solutions which are genetic and particle swarm optimization.