What Really Drives Our Power System Models? Identifying the Assumptions That Shape U.S. Power Futures

The Regional Energy Deployment System (ReEDS) model is widely used to explore long-term U.S. power sector futures, yet its growing complexity makes it difficult to see which assumptions actually shape its headline results. This work develops a practical framework built around ReEDS to expose which techno-economic and policy inputs matter, which do not, and how that knowledge can streamline scenario design. We construct a ReEDS-based experiment that perturbs multiple classes of inputs—technology costs, performance, fuel prices, load growth—using a structured sampling design and the Morris screening method implemented via open-source Python tools. Rather than treating sensitivity analysis as an add-on, we integrate it into the model workflow so that each batch of runs simultaneously supports factor prioritization (which inputs drive system cost, generation mix, transmission build-out, and emissions trajectories), factor fixing (which inputs can be held constant with minimal loss of information), and the design of more informative scenario ensembles. This screening-focused framework is intended as a first-stage filter to reduce dimensionality, guide subsequent in-depth analyses, and support the construction of targeted, information-rich scenario ensembles for ReEDS applications