Forecasts of oil recovery from reservoir simulation models of the Huizhou (HZ) fields in the Pearl River Mouth Basin of the South China Sea (Fig.1) have historically underpredicted actual reservoir performance. The major factor identified to explain the trend of underprediction by the simulation models is reliance on deterministic geological modeling techniques that in several cases have resulted in underestimation of original oil in place (OOIP) and reserves.

In the fields’ largest reservoir, a Miocene sandstone known as K22(106), underprediction of oil production by the simulation model during the first five years of production necessitated several revisions to estimates of OOIP and reserves. The reservoir was initially developed in 1993. Subsequently, several upward revisions of OOIP were made by incrementally reinterpreting petrophysical properties. By May 1999, the OOIP estimate was more than twice the post development estimate and yet the reservoir was clearly going to produce more oil than predicted by the deterministic model.

Realizing a new approach was needed the operator, CACT Operators Group, chartered creation of new geological and reservoir simulation models by Chevron, one of the four principals of CACT. The geological reservoir model was built using techniques that had not previously been applied to Huizhou Fields’ models. Porosity, permeability, and water saturation were geostatistically distributed in a finely layered model that was later upscaled to form the basis of a new reservoir simulation model.

The history match of a reservoir model does not yield a unique solution. Multiple history matches may be obtained that do not reflect the actual properties of the field, and hence, will probably not be sucessful in the prediction phase. For this reason, it is important to tie the simulation model closely to the geology and fluid characteristics of the field. The new model accomplishes this by testing geologic scenarios with field production, and then choosing the scenario that matches production history closest. Changes are kept to a minimum in the history matching phase, including only changes justified based on geologic, fluid characteristics, or field development data. Uncertanties in model parameters or production information are taken into account when considering differences between model and field production – matches do not have to be exact.

Working as an integrated multidisciplinary team, multiple geological model realizations were rapidly tested in the reservoir simulation model to obtain the best possible global history match, with only minimal changes to global parameters, such as vertical to horizontal permeability ratio, rock relative permeability curves, and J-Function for water saturation initialization. Without changing the reservoir surface, gross rock volume or oil-water contact, the new techniques led to a 33% increase in the estimated OOIP, and a substantial increase in estimated reserves. By better representing the heterogeneity of the reservoir, areas of unswept oil were identified, which should enable better placed sidetracks and possibly additional sidetracks in the years ahead.

You can access this article if you purchase or spend a download.