Multiparametric optimization based decision making approach in energy systems planning under uncertainty

Abstract

Energy infrastructure planning confronts a fundamental temporal asymmetry in which capacity decisions must be committed before uncertainty materializes and yet must remain robust across all future operational scenarios. Traditional approaches either adopt worst-case hedging or ignore operational uncertainty. We address this through a Robust Optimization framework integrating data-driven uncertainty quantification with multiparametric programming, enabling design strategies as explicit functions of realized uncertainty without iterative re-optimization. We quantify forecasting errors via probabilistic deep learning technique to inform uncertainty bounds. Rather than worst-case designs, we model it as bilevel problem solved multiparametrically, deriving "wait-and-see" operations as piecewise affine functions of "here-and-now" decisions and uncertainty realizations.Applied to a multi-source, multi-sink energy system (wind, nuclear, natural gas, battery, electricity and hydrogen demands) under joint supply-demand uncertainty, the framework reveals a strategic shift in technology prioritization as uncertainty quantiles increase, optimal designs preferentially allocate capacity to expensive-but-reliable dispatchables even though economical renewables remain available. The complete parametric solution enables upfront cost-robustness trade-off exploration, providing a tractable and generalizable tool for risk-aware capacity expansion, supporting strategic investment decisions in uncertain and rapidly evolving energy landscape.