From lithology prediction to geothermal energy production: Leveraging data-driven and financial insights to unlock geothermal potential of a carbonate reservoir, SE Hungary

Deep Triassic carbonates in Hungary's Pannonian Basin hold substantial geothermal potential but remain difficult to evaluate due to structural complexity, sparse well control, and uncertainty in reservoir continuity. To reduce exploration risk, this study develops a pioneering integrated machine-learning seismic workflow applied to Hungarian geothermal carbonates. A 120 km2 post-stack seismic volume and logs from four wells were used to generate 3D lithology, porosity, and temperature models through seismic inversion, probabilistic neural networks, and supervised Bayesian classification. The workflow delineates the spatial extent of fractured dolomitic limestone platforms, quantifies reservoir quality (5–15 % porosity), and estimates the geothermal resource (HIP ≈ 15,600 PJ; Hrec ≈ 1440 MWth). Probabilistic facies volumes improve the identification of carbonate shale transitions and support uncertainty-aware production scenarios ranging from 48 to 144 MWth. A heat-led techno-economic assessment indicates that a 2P–1I binary CHP design (3.85 MWe, 14.31 MWth) is viable, yielding LCOH values of €102–86/MWh for capacity factors of 0.55–0.65 and annual revenues of €6.9 M from heat sales (plus €0.9 M from seasonal power). Estimated payback ranges from 12.7 to 8.9 years. This integrated geophysical economic framework demonstrates a reproducible pathway for de-risking deep carbonate geothermal systems and supports strategic development of district-heating projects in Hungary.