Multi-level delumping strategy for thermal enhanced oil recovery simulations at low pressure

We present a multi-level delumping method suitable for thermal enhanced oil recovery processes, which hydrocarbon components are vaporized under high temperatures, move downstream in the gas phase and condense back to liquid phase. To reduce computational cost, it is standard practice number of (pseudo-)components used reservoir simulation. Depending on type pseudo-components retained simulations, we may not be able capture correct displacement due large errors lumped behavior (flash) computations. address that problem through method: use data obtained from short simulation using most detailed fluid description available, leverage information guide process. temperature as proxy variable composition, select reference temperatures. extract corresponding compositions run them extend an approximate composition. test our six heavy samples, two different processes: hot nitrogen injection in-situ combustion (air exothermic oxidation reactions). The average error mole fraction reduced by 4-12 times (depending samples) compared flash pseudo-components, maximum 6-48 times. illustrate amenable manually adding more about physics some samples. also discuss how efficiently pick For uniformly sampled temperatures (between minimum temperature), conduct sensitivity study led us ran both local (Pattern Search, PS) global (Particle Swarm Optimization, PSO) gradient-free optimization methods.