Little data yields low-confidence model
Dana Petroleum's challenge in evaluating Nefertiti Field, which is
adjacent to a mature development field, was that very little information had
been collected to characterize reservoir conditions within a static model: only
conventional core analysis and a nuclear magnetic resonance (NMR) log from a
single well.
Low-data fields severely limit confidence in the reservoir model, which
in turn increases the risk posed by costly field decisions. Many assumptions
and correlations must be made to augment the data in constructing a
representative reservoir model, and each of these adds to the uncertainty.
Neural network efficiently determines permeability distribution
Schlumberger proposed using the Petrel E&P software platform to
establish relationships between the core analysis and log interpretations for
delineating spatial trends across the reservoir. In the absence of a facies
model, a neural network was created in the Petrel platform. The routine logs,
such as gamma ray, obtained with the NMR log were used to precondition the
algorithm for predicting the flow zone index (FZI) for the wells across the
permeability distribution. Not only did using a neural network save a
significant amount of time, but it fully represented reservoir heterogeneity
even though the porosity distribution and NMR data were available from only one
well.
History-matched dynamic model thoroughly assesses development
options
Following determination of the permeability distribution for the whole
model and creation of relative permeability and capillary pressure curves, the
dynamic model was ready for initialization. Dynamic modeling honored all the
static model and data preparation uncertainties in estimating the original oil
in place (OOIP) and the P10, P50, and P90 values to support asset valuation and
field development planning.
However, before the development plan could be initiated, it was critical
to achieve a history match. The pressure match was the most challenging because
Nefertiti Field was initially depleted. This was remedied by creating an
aquifer in the Petrel platform model to accurately reflect the pressure
response of the reservoir. With a successful history match, the model could
then be used to evaluate production scenarios.
The different development plans assessed with the matched model were all
focused on optimal well placement of the producers and injectors. Using the
Petrel platform and ECLIPSE simulator made it easy to compare different
trajectories and well controls, and two new producer locations were agreed on
for drilling. To maintain reservoir pressure throughout development, a
waterflooding scenario was assessed within the Petrel platform and ECLIPSE
simulator, and three new injectors were confidently recommended for inclusion
in the next drilling schedule.
