Reservoir modeling has become an indispensable tool in the oil and gas industry to measure risk associated with alternative production scenarios. It is no secret that nowadays more and more operators use deterministic approach for history matching but even so, the task remains technically challenging. Given a large number of uncertainties associated with subsurface models, it has become critical to produce a set of history-matched models and use them for risk assessment in predicting the future performance of the reservoir.
Quantification of uncertainties is a vital process in complex fields to ensure a proper field development plan (FDP) is in place. In this paper, we present a comprehensive workflow that incorporates all processes from building static and dynamic models to producing ensemble of history matched models. To assess the mismatch with historical production data, a number of geological realizations were created and simulated using a variety of uncertainty parameters. Based on the obtained mismatch value, the posterior probability distribution for uncertain parameters updated. Different optimization methods are used to find candidates with improved match quality, and this loop is repeated until a set of diverse history-matched models are generated. These calibrated reservoir models are then used to estimate prediction uncertainties for further FDP optimization.
This study is focused on automating workflows and generating multiple geological realizations that produce a set of history-matched models to probabilistically estimate prediction uncertainties and optimize FDP. We also show how these models are critical in the process of decision making and risk evaluation.