Adapting long-term generation scheduling to climate variability in hydro-dominant power systems

Hydropower is one of the most significant sources of renewable electricity worldwide, however, it depends on water inflows, which are directly influenced by the increasing climate variability. This variability must be considered in the operation and planning of power systems with significant hydroelectricity production to ensure robust operation. This paper analyzes how climate-projections affect power system operations and investigates different risk measures to manage climate variability: (i) Expected Value, (ii) convex combination of Expected Value and Conditional Value at Risk (Mean-CVaR for short), and (iii) Distributionally Robust Optimization (DRO) with Wasserstein distance. Results considering the Brazilian power system indicate that policies derived from scenario trees purely based on historical observations are poorly prepared for projected hydrological shifts, leading to higher marginal costs and increased load cuts. Increasing risk aversion partially mitigates these effects, but does not fully substitute for explicitly incorporating climate projections into the optimization. When climate-informed scenario trees are used, policies exhibit improved storage management, lower extreme costs, and greater reliability. Moreover, DRO is easier to adjust, requiring calibration of only one parameter, and demonstrates greater stability with respect to parameter calibration. These findings highlight the importance of explicitly modeling climate-driven distributional shifts in long-term energy planning and clarify the distinct roles of tail-risk penalization and distributional robustness.