June 22, 2023
Place du Levant 2, b.044
The design of emerging energy systems relies on limited market data and long-term forecasts, leading to significant uncertainty in critical model inputs such as efficiencies, resource availabilities, fuel prices and investment costs. We consistently underestimate these uncertainties due to our reliance on the ordinary and the predictable, neglecting exceptional scenarios. This leads to a substantial mismatch between simulations and reality, potentially resulting in a kill-by-randomness of the energy system.
The seminar will be divided into three parts. Part one presents a computationally-efficient method to propagate input uncertainties through expensive models, allowing for a quantification of the resulting uncertainty on the performance (e.g., total system cost). Part two focuses on optimizing the design variables (e.g., number of photovoltaic panels, battery size) for expected performance, robustness, and upside variability. Lastly, we introduce a metric for optimizing antifragility, enhancing positive outcomes when the real-world uncertainties surpass the initial estimations made during simulation.
Even though this workshop focuses on energy systems, the method is non-intrusive and can be easily applied to any other system in any other field.