Predicting Thermal Performance of Aquifer Thermal Energy Storage Systems in Depleted Clastic Hydrocarbon Reservoir via Machine Learning: Case Study from Hungary

• DOI: 10.20944/preprints202504.2474.v1
• Publication Date: 2025-04-30T01:05:33Z
• Authors: Hawkar Abdulhaq; János Geiger; István Vass; Tivadar M. Tóth; Tamás Medgyes; Gabor Bozsó; Balázs Kóbor; Éva Kun; János Szanyi

This study explores a novel strategy to repurpose depleted clastic sediment hydrocarbon reservoirs in Hungary as High-Temperature Aquifer Thermal Energy Storage (HT-ATES) systems, incorporating machine learning to enhance system optimization. Hungary's extensive inventory of depleted fields, predominantly featuring clastic formations, presents significant potential for geothermal energy storage applications. Initially, detailed reservoir models were constructed by analyzing existing well logs and core data. Subsequent advanced numerical simulations of heat transport and groundwater flow were performed within the Bekesi Formation, concentrating on a dual-well configuration—one dedicated to hot fluid injection and extraction and the other to managing cold fluids. State-of-the-art simulation tools, including SGeMS, RockWorks, Python, MODFLOW, and GMS MT3DMS, were utilized to pinpoint optimal brine injection sites by evaluating critical parameters such as thermal conductivity, porosity, and permeability; additional core analyses filled essential gaps in thermal conductivity data. The study's central innovation lies in deploying a Random Forest algorithm to optimize thermal recovery efficiency. Data generated from comprehensive simulations across multiple wells were used to train the model, which then predicted and refined thermal performance for the remaining wells in the field. The outcomes are expected to yield precise identification of optimal injection locations, rigorous heat transport analyses, accurate estimates of storage capacities, and improved predictions of thermal recovery efficiency, thereby establishing a sustainable and data-driven methodology for converting depleted hydrocarbon reservoirs into effective thermal energy storage systems.