The
initial state of hydrocarbon mixtures in petroleum reservoirs
is the result of equilibrium among several forces, the most important
of which are the chemical forces arising from chemical potential gradients
of the molecular species in the petroleum accumulation, the gravitational
force arising from the gravitational acceleration, and the thermal
diffusion forces arising from temperature gradients. The equilibrium
among these forces determines the state of pressure and a compositional
gradient and the creation of a gas–oil contact (GOC) in a stationary
reservoir along with the changes in other physical properties. Accurate
modeling of these changes in the development of a proper stationary
model for the reservoir simulation initialization leads to more realistic
predictions of the future behavior of petroleum reservoirs. This is
important especially when phase behavior is important in designing,
modeling, and predicting the performance of the processes used to
maximize the oil recovery, such as in dealing with a gas condensate
reservoir or when miscible displacement is to be done in the enhanced
oil recovery (EOR) stage of the reservoir life. In this study, we
consider the equilibrium among chemical, gravitational and thermal
diffusion forces to predict the changes in reservoir fluid composition
and pressure and also to predict the location of a possible GOC in
a reservoir. Additionally, we develop a simple model to predict the
change of the plus-fraction molecular weight (MW) in a non-isothermal
reservoir using continuous thermodynamics and the theory of irreversible
processes. We propose a method not only to tune the equation of state
(EOS) versus the measured PVT lab data for one fluid
sample but also to accurately model the depths of the GOC and other
fluid samples and their PVT lab data in order to
determine which sample is representative of the reservoir fluid and
also to develop an EOS model that can work for every fluid in the
reservoir, not just a single point. In two case studies, we validate
our calculation procedure for the general compositional gradient,
GOC detection, and the plus-fraction MW change in the reservoir against
two data sets from the literature. The computational results show
that the model developed works satisfactorily to predict the fluid
changes in these two reservoirs. Subsequently, we also report the
results of a series of sensitivity analysis tests to show the factors
affecting the compositional gradient calculations and present examples
of abnormal fluid distributions in a hydrocarbon fluid column where
the fluid becomes denser toward the top of the column or the changes
in fluid properties are highly nonlinear with respect to depth in
the reservoir.