Abstract
This paper presents a methodology to evaluate the degree of uncertainty of basement fracture reservoir performances when a reservoir has special events like water-breakthrough. Basement fracture reservoir is a particular feature with water-cut performances compared to other common reservoirs due to extremely high fracture permeability and heterogeneity of reservoir properties.
Proxy models are used in this paper to estimate production history and uncertainty. They are useful to reduce the uncertainty contained in a simulation model due to unknown parameters such as reservoir permeability, aquifer size and influx rate, and so on. From a real field application of the proposed methodology, we conclude that the uncertainty of estimated ultimate recovery (EUR) can be significantly reduced when we have water-breakthrough data and proxy models can be utilized to extract probabilistic ranges of EUR with high confidence.
This paper proposes to use only high quality history matching (HQHM) models to configure neural network for a proxy modeling to estimate the uncertainty of EUR. The proxy model to be extracted from HQHM model will reduce the error of estimation and reveal real uncertainty due to uncertain parameters not from the proxy model itself.