Evolving cellular automata to model fluid flow in porous media

Fluid flow in porous media is a dynamic process that is traditionally modeled using PDE (partial differential equations). In this approach, physical properties related to fluid flow are inferred from rock sample data. However, due to the limitations posed in the sample data (sparseness and noise), this method often yields inaccurate results. Consequently, production information is normally used to improve the accuracy of property estimation. This style of modeling is equivalent to solving inverse problems. We propose using a genetic algorithm (GA) as an inverse method to model fluid flow in a pore network cellular automaton (CA). This GA evolves the CA to produce specified flow dynamic responses. We apply this method to a rock sample data set. The results are presented and discussed. Additionally, the prospect of building the pore network CA machine is discussed.