|
The purposes of this study are to modify the traditional
volumetric method in exploration stage and to derive a
stochastic decline curve analysis in development for an oil /
gas field. In addition, the neural network and the grey
prediction theory are used to estimate the probabilistic
reserves, producing rate, and production limit of an oil / gas
field. Finally, the comparisons will be made with different
reserves estimation methods. A "pseudo-normal distribution"
proposed in this study has an advantage of a triangular
distribution (the only need inputs are minimum, maximum and most
likely of parameter), and without the disadvantage of a
triangular distribution (the probability is zero outside of the
minimum and the maximum). In this study, pseudo-normal
distribution was used in volumetric method to estimate the
reserves of gas field in Taiwan. With the same parameters, the
result reveals that the range of the reserves estimated by
pseudo-normal distribution is 50% large than by triangular
distribution. The stochastic decline curve analysis derived in
this study uses the characteristic of production data, decline
curve and Monte Carlo simulation to estimate probabilistic
reserves. For developed field (Shiells Canyon field), the range
of reserves estimated by stochastic decline curve analysis is
almost the same as from a probabilistic decline curve analysis
with *5% variation in parameters (initial production rate,
decline exponential and decline rate). The reserves estimated
by the stochastic decline curve analysis is 20% less then it
from probabilistic decline curve analysis. The production data
of Shiells Canyon field and other related data are used to
establish a neural network for estimating reserves. The neural
network can also be used to study the factor impacting the
production rate. In Shiells Canyon field, oil demand has the
most impact on the production and water production rate has less
effect on oil production rate. With the same economic
conditions, the reserves from neural network prediction is about
30% less than from decline curve analysis. A model to predict
future production has also be built based on the grey prediction
theory, the trend of influence factor itself, Monte Carlo
simulation and neural network. The reserves estimated from the
model is about two times more than from the decline curve
analysis. The main reason for this difference is that the oil
price in the decline curve analysis is fixed, and the predicted
oil price by the grey prediction is increase gradually. The
higher oil price will cause a longer production period and thus
a larger reserves.
|