Uncovering Hidden Biases in Hydropower: Why Detailed Inflow Data is Crucial for Energy System Optimization Models

Abstract

Hydropower plays a critical role in balancing highly renewable power systems, yet its representation in energy system optimization models is often constrained by limited availability of high-resolution inflow data. In practice, aggregated inflow time series are frequently used to reduce data requirements, potentially introducing hidden modeling biases. This paper analyzes the impact of inflow aggregation on generation expansion and operation decisions in energy system optimization models. Using an adapted NREL-118 bus test system and the open-source Low-carbon Expansion Generation Optimization (LEGO) model, we compare planning outcomes based on hourly inflow data against daily, weekly, monthly, and yearly aggregations. By computing expost regret with respect to hourly inflow operation, we quantify the cost and investment distortions caused by aggregation.Our results show that inflow aggregation leads to substantial cost increases and misallocation of generation capacity investments once hydropower becomes a significant share of the energy mix, as coarse aggregations systematically underestimate inflow variability. These findings demonstrate that high-resolution hydropower inflow data is essential for robust energy system planning and highlight the need for openly available, standardized inflow time series to support planning of future energy systems.Hydropower plays a critical role in balancing highly renewable power systems, yet its representation in energy system optimization models is often constrained by limited availability of high-resolution inflow data. In practice, aggregated inflow time series are frequently used to reduce data requirements, potentially introducing hidden modeling biases. This paper analyzes the impact of inflow aggregation on generation expansion and operation decisions in energy system optimization models. Using an adapted NREL-118 bus test system and the open-source Low-carbon Expansion Generation Optimization (LEGO) model, we compare planning outcomes based on hourly inflow data against daily, weekly, monthly, and yearly aggregations. By computing expost regret with respect to hourly inflow operation, we quantify the cost and investment distortions caused by aggregation.Our results show that inflow aggregation leads to substantial cost increases and misallocation of generation capacity investments once hydropower becomes a significant share of the energy mix, as coarse aggregations systematically underestimate inflow variability. These findings demonstrate that high-resolution hydropower inflow data is essential for robust energy system planning and highlight the need for openly available, standardized inflow time series to support planning of future energy systems.