A robust optimization approach of well placement for doublet in heterogeneous geothermal reservoirs using random forest technique and genetic algorithm

Reinjection is important for sustainable utilization of geothermal energy, and geothermal well placement determines the fate and recovery efficiency of doublet system. The simulation-based optimization method usually requires a large number of intensive forward simulations to evaluate the reservoir performance considering every possible well configuration. In this paper, a surrogate model based on random forest technique was developed to reduce the substantial computational burden of forward simulations, which was combined with genetic algorithm to develop a robust optimization approach of geothermal well placement in heterogeneous geothermal reservoirs. A number of statistical indicators including the maximum and minimum permeabilities in the three different representative areas surrounding the geothermal wells were incorporated into the surrogate model, and its prediction accuracy can be significantly improved. The reasonability and efficiency of the developed optimization method for well placement were demonstrated using three case studies including homogeneous and heterogeneous geothermal reservoirs based on a doublet system in the Dezhou geothermal field, China. The results show that the surrogate model-based optimization method can not only robustly and accurately find the optimal position of injection well given a certain position of production well, but also work well when the simulation-based optimization method fails in complex geothermal reservoirs.