ALBERT

Description: 〈p〉Publication date: 15 November 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 Applied Energy, Volume 254〈/p〉 〈p〉Author(s): Ahinoam Pollack, Tapan Mukerji〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉It has been estimated that Enhanced Geothermal Systems could supply 100 GWe (10%) of total electric capacity in the U.S. An Enhanced Geothermal System (EGS) is created by stimulating an impermeable hot rock, injecting cold water into the hot reservoir, and extracting the heated water to generate electricity. EGS projects are still not commercially feasible, however, due to many challenges, including subsurface uncertainty. There are many uncertain structural and geological features when creating an EGS. With uncertain reservoir properties, it is difficult to optimize decisions that will greatly improve EGS profitability. Currently, a common method of optimizing an EGS is choosing the most representative subsurface reservoir model and optimizing the engineering parameters for this single reservoir model, or Single-Model Optimization (SM-Opt). Due to availability of larger computational power, another feasible option is accounting for subsurface uncertainty by optimizing an EGS given an ensemble of reservoir models, or Multiple Model Optimization (MM-Opt). This option is less common in practice within the geothermal industry since it lags in harnessing computational power. This study compares these two methods for optimizing eight common EGS engineering decisions, including well configuration and fracture spacing. The decisions were optimized to maximize the Net Present Value (NPV) of an EGS. We have found that using SM-Opt, the optimal engineering decisions led to an EGS with a NPV estimate of $32.7 million. This contrasts with the MM-Opt results where the optimal engineering decisions led to a median NPV value of $11 million and a standard deviation of $15 million. This comparison illustrates how ignoring subsurface uncertainty and heterogeneity leads to over-optimistic NPV forecasts. For this study, the SM-Opt optimum decisions were similar to the robust decisions identified using MM-Opt. Yet, in contrast to SM-Opt, the MM-Opt workflow provided an analysis of the influential engineering parameters and a NPV uncertainty range, which was used to ensure decision robustness.〈/p〉〈/div〉 〈/div〉