ANN evaluation of thermodynamic efficiency, economic and environmental analysis of the geothermal-based electricity and hydrogen production system: A case study

This study presents a holistic artificial neural network (ANN)–assisted assessment of the thermodynamic, economic, and environmental performance of a geothermal-based integrated electricity and hydrogen production system. Based on real field data, two distinct operating scenarios are defined: in the S-1 configuration, the entire electricity generated by the Organic Rankine Cycle (ORC) is directed to hydrogen production via a proton exchange membrane (PEM) electrolyzer, whereas in the S-2 configuration the generated electricity is directly supplied to the grid. System performance is evaluated through thermodynamic modelling; the economic analysis is conducted using the levelized cost of energy (LCOE) methodology, and the environmental assessment quantifies avoided CO2 emissions together with natural gas savings. A multi-output ANN model developed using a dataset of 4776 observations is structured to simultaneously predict energy efficiency, exergy efficiency, hydrogen production rate, net power output, and cost indicators. The results indicate that the average exergy efficiencies of the S-1 and S-2 configurations are 6.57% and 7.73%, respectively. While the S-2 configuration appears economically advantageous due to lower LCOE values and shorter payback periods, the S-1 configuration delivers greater CO2 mitigation per unit hydrogen production from an environmental perspective. The ANN model demonstrates moderate predictive performance, with R2 values ranging from 0.47 to 0.67, MAE values as low as 0.000002, and MAPE values between 0.21% and 4.16%, indicating that the complex behaviour of geothermal–hydrogen systems can be reliably captured using data-driven approaches. The low error levels and high coefficients of determination obtained from the ANN model demonstrate that the complex behaviour of geothermal–hydrogen systems can be reliably captured using data-driven approaches. Rather than focusing solely on performance optimisation, this study proposes a multidimensional framework based on real operational data, providing a methodological guideline for the sustainable design and operation of geothermal-based multi-energy production systems.