Machine learning-enabled techno-economic uncertainty analysis of sustainable aviation fuel production pathways

Abstract

Stochastic techno-economic analysis (TEA) is pivotal in assessing the financial viability and risks inherent in biofuel production processes. In this method, the Monte Carlo approach entails the random sampling of input variables and multiple runs of the TEA model to create probability distributions of economic metrics. However, traditional Monte Carlo TEA, reliant on iterative calls to process simulation, is resource-intensive and time-consuming, hindering widespread adoption. To address these challenges, we present an accessible framework that harnesses machine learning methods to estimate techno-economic uncertainty in biofuel production pathways. Our approach streamlines the conventional simulation process by automating dataset generation and machine learning model training. These trained models enable rapid predictions of minimum fuel selling prices at any scale, accommodating randomized input variables based on their defined distributions. We illustrate the efficacy of our framework through examples from sustainable aviation fuel production pathways. Our research entails identifying the primary factors influencing uncertainties in minimum selling prices, exploring the synergistic effects of pathway inputs, and assessing how price variability is impacted by financial, technical, and supply chain factors. These examples underscore the framework's effectiveness in addressing breakeven price uncertainties in biofuel production across diverse input scenarios.