Hydrocarbon reservoirs are large, geologically complex entities with
heterogeneities on all length scales from millimetres to kilometres.
Modern mathematical models of reservoir geology can be very detailed
with up to 20 million gridblocks. This is a much higher level of detail
than can be resolved by numerical flow simulation required to predict
oil recovery rates. In order to span this difference in scales a variety
of methods of upscaling have been used. In this paper we describe how
percolation theory can be used to estimate the large scale flow rates in
hydrocarbon reservoirs. We use a scaling theory to predict the
probability distribution of time to breakthrough of an injected fluid
(typically water). This breakthrough time is extremely important as it
influences the economic recovery rates from oil fields. The scaling law
predicts the entire probability distribution semi-analytically something
that would be extremely computationally time consuming by conventional
Monte Carlo and flow simulation approaches. We also demonstrate the
validity of this approach on a real field example.