Synthetic Geothermal Doublet Simulations for Thermal-Breakthrough Prediction

This dataset contains 600 synthetic two-dimensional geothermal reservoir simulations of a thermal doublet, generated with the MATLAB Reservoir Simulation Toolbox (MRST) and its geothermal module. Each case models conductive-advective heat transport for incompressible single-phase flow on a 50 by 50 Cartesian grid, with a rate-controlled injector and a producer. Heterogeneous permeability and porosity fields are drawn from a von Kármán power spectral density, with permeability constrained to a physical 0.1 to 200 mD range. Well placement, injection rate, and the petrophysical parameters are sampled by Latin hypercube design. The injection and reservoir temperatures are held constant at 308.15 K and 348.15 K. Each simulation spans a 100-year horizon written on a 55-snapshot schedule that resolves the early transient before switching to uniform two-year steps. The HDF5 file stores, per case, the static property and flow fields (porosity, log-permeability, time-of-flight streamline coordinates, velocity), the time-dependent temperature fields, and the well configuration. The dataset was produced for training and benchmarking a physics-aware graph neural network surrogate for thermal-breakthrough prediction. Github Repo (will be public upon publication of preprint): https://github.com/rezanajafisilab/ThermalBreakthrough