Scenario space exploration for robust energy planning

Long-term energy planning must contend with deep uncertainty in technology, markets, weather and demand, yet conventional studies compare a handful of narrative scenarios and can miss the thresholds and regime shifts that decide policy outcomes. We propose the scenario-space exploration workflow for large-scale energy-system optimisation models, coupling large-ensemble simulation, global sensitivity analysis, response diagnostics, interaction surfaces and scenario discovery. Applied to the Dutch net-zero energy system in 2050 across 13,200 evaluations of 31 uncertain inputs, it shows first that screening and distribution-based sensitivity disagree on zero-inflated or heavy-tailed outcomes, including the CO₂ price, unserved energy and BECCS use: screening alone misranks the drivers of the outcomes that govern carbon pricing and adequacy. Extreme weather cuts onshore wind generation from 46 to 11 PJ, yet total system cost is unchanged at 49.4 billion euro, because the country switches from importing 31 PJ of electricity to exporting 17 PJ: resilience is traded rather than bought. Other responses are cliff-edged: synthetic-fuel production holds near 20 PJ as the bunker emission-reduction requirement rises to 82%, then falls to zero by 86% as biofuel imports roughly double, so one policy lever switches a carrier off and another on. Scenario discovery turns these responses into decision rules. For example, a biomass import price below 12.2 MEUR per PJ combined with CO₂ storage potential above 30.7 Mt/yr isolates a region in which 95% of runs are both affordable and low in carbon price. Robust energy planning therefore depends not only on knowing which uncertainties matter, but on how they matter: as levers, thresholds or switches between transition pathways.