Toward Integrated Optimization of Subsurface & Surface Dynamics in Geothermal Power Plant

Abstract

In 2017, the Kingdom of Saudi Arabia launched a $500 billion US Dollar project to build a renewable energy-only city in NEOM. Geothermal energy is one of the most abundant renewable resources in NEOM, with great potential for electricity generation via power plants. Nevertheless, existing studies typically separately evaluate the geothermal reservoirs and power plants, which may lead to inaccurate estimations of the electricity capacity from the geothermal reservoirs. To resolve this bottleneck, this work innovatively proposes an optimization framework with an integrated forward model that couples a geothermal reservoir model with a power plant model. In the subsurface, the geothermal surrogate model, which consists of a Weibull-based decline model and a deep neural network model, can take reservoir parameters as input variables to predict the produced fluid temperatures at both the bottom hole and the surface. The decline model is used to flexibly describe the production temperature decline behavior in geothermal reservoirs, and the neural network is used to build the nonlinear mapping between the reservoir model parameters and decline model parameters. In the surface, an in-house binary Organic Rankine Cycle (ORC) geothermal power plant model is developed to integrate with the geothermal model through boundary conditions. Finally, the multi-objective optimization (MOO) based on the Non-dominated Sorting-based Genetic Algorithm II (NSGA-II) is applied to optimize the thermodynamic and economic performances of the integrated system with considering uncertainties in the geothermal reservoir model parameters. In the numerical experiments, the decline model fits the production temperature data well, with a mean relative error of 0.32% for both the production temperatures at the bottom hole and the surface. The deep neural network can accurately predict the 5 parameters in the decline model, with decent R2 scores of 0.973, 0.964, 0.965, 0.998 and 0.996, respectively. The mean relative errors of the predicted bottom-hole and surface temperatures are 0.47% and 0.46%, respectively. Finally, the MOO optimizes the electricity generation and economic performances of the geothermal power plant while minimizing the subsurface risks, which gives optimal subsurface and surface operational parameters with considering uncertainties in reservoir properties. Optimization results recommend the superheated organic Rankine cycle as the best geothermal power plant configuration in NEOM. The optimal electricity output, levelized cost of electricity and net present value are 202.91 GWh, 15.07 US¢/kWh, -9.42 million US$ (without feed-in tariff) and 23.60 million US$ (with feed-in tariff), respectively. Our work, which fully integrates the geothermal field and the power plant, can contribute to more accurate assessments of the power plant outputs and provide guidance to both the subsurface and surface operations under optimal conditions.

Keywords: Geothermal, power plant, optimization, electricity, economics analysis