Modeling Channel Architecture in a Densely Drilled Oilfield in East China

Abstract

Abstract An "object-type" stochastic model has been used to describe channel architecture within a densely drilled onshore field in East China (Gudong Field). The Gudong dataset comprises 75 wells which are drilled in a regular pattern at a spacing of down to 150 m. The modeled reservoir is 43 m thick with an average net:gross of 0.29. There are 144 observations of channel sandbodies in the 75 wells. The model can simulate channel architectures in the Gudong Field where the average channel width is similar to the well spacing within the most densely drilled part of the field (150 m). The study has demonstrated that "object-type" models can be applied to mature fields with a large number of wells. The key to the conditioning of object models is a correct use of the well data, and enough flexibility in the model to capture complex, realistic geometries. Introduction A number of different object-type stochastic models have been extensively used for modeling channel architecture in fluvial reservoirs since the mid-eighties. The models have been mainly applied to fields in the early phase of development where well data are few. The application of these stochastic models has helped to improve the understanding of flow processes in complex fluvial reservoirs, and realizations of the models have been used to evaluate reservoir connectivity, design well patterns, and predict general production profiles. Many of the larger fluvial reservoirs in the North Sea are now in a more mature phase of development, and the most important reservoir management tasks are related to finding and producing pockets of bypassed oil. The early fluvial channel models could not be applied to such fields where well data are abundant, and complex multiwell conditioning is required. A fluvial model developed at the Norwegian Computing Centre in Oslo has undergone continual modification in response to the increasing maturity of fields on the Norwegian Continental Shelf. Several measures have been adopted for achieving successful multiwell conditioning, including:use of more flexible, realistic channel geometries.use of apriori information on correlation of channels.use of well test data. Completely new stochastic simulation algorithms have also been adopted. This paper describes a test of a recent version of the model. Emphasis has been placed on evaluating the significance of flexible and realistic channel geometries for achieving multiwell conditioning. The test is based on a complex dataset from a densely drilled onshore field in East China (Gudong field). To our knowledge, object-type stochastic models have not yet been applied to datasets with such close, regular well spacing. Gudong Dataset The Gudong Field is located in the Eastern China. The main producing intervals are from fluvial sandbodies of the Guanto Formation. The field has an areal extension of approximately 10 km2. Over 230 wells have been drilled within the field which is being produced using water injection for pressure support. The water cut is very high and an important reservoir management challenge is to locate pockets of bypassed oil. Reservoir heterogeneities within the Gudong Field have previously been modeled using indicator simulation techniques. The abundance of closely spaced wells provides an excellent dataset to test the applicability of object-type models for describing heterogeneities in mature fields. A 43 m thick reservoir interval has been modeled. The interval has an average net-to-gross ratio of 0.29. Due to partial penetration of the reservoir interval, and some data quality problems, only 75 of the 230 wells have been used in the test (Fig. 1). P. 365^