Monte carlo dataset for techno-economic assessment of sustainable aviation fuel production via methanol-to-jet
Description
This dataset contains large-scale Monte Carlo simulation results for a techno-economic assessment (TEA) of a methanol-to-jet (MtJ) fuel production pathway. A total of approximately 3 million simulation runs were generated by systematically varying key economic, technical, and operational input parameters within literature-based and industry-relevant bounds. The underlying TEA model is implemented as a steady-state process and cost model, evaluated through an Excel-based framework coupled to Python. For each Monte Carlo sample, the model computes the net production cost (NPC) of sustainable aviation fuel per mass (kg). The dataset is designed to support: -Uncertainty and sensitivity analysis of MtJ techno-economic performance -Development and benchmarking of machine learning surrogate models -Explainable AI (XAI) studies for identifying dominant cost drivers -Reproducible comparison of TEA assumptions across studies To facilitate reuse, the dataset is accompanied by detailed variable descriptions, parameter bounds, and the Python scripts used for data generation. Reduced-size sample files are provided for rapid testing and machine learning prototyping.
Files
Steps to reproduce
Monte Carlo sampling Input parameters were sampled using uniform probability distributions within predefined bounds derived from literature, market data, and engineering assumptions. The sampling procedure was implemented in Python using NumPy. Model evaluation Each Monte Carlo sample was written to an Excel-based techno-economic model using xlwings. The model performs steady-state calculations of capital expenditure, operating expenditure, and economic key performance indicators. Post-processing After each model evaluation, the net production cost (NPC) was read back into Python and stored in a structured CSV format. Invalid or non-converged simulations were filtered out. Data aggregation All valid simulation results were concatenated into a single dataset containing approximately 3 million rows. A reduced subset was extracted for demonstration and testing purposes. Reproducibility The provided Python scripts allow full reproduction of the dataset, given access to the same Excel-based model and identical parameter bounds.
Institutions
- Montanuniversitat Leoben