Sensitivity of Forecast Value in Multiobjective Reservoir Operation to Forecast Lead Time and Reservoir Characteristics

Abstract

Streamflow forecasts can be used to improve reservoir operation decision-making, yet the sensitivity of forecast value to forecast lead time and reservoir characteristics in multiobjective reservoir operations is rarely investigated simultaneously. We incorporate streamflow forecasts into reservoir operation by applying radial basis functions (RBFs) and evaluate the forecast value conditioned on different types of forecast lead time and reservoir characteristics. From a case study of the Danjiangkou Reservoir in the Hanjiang River basin, China, forecast value is assessed by comparing Pareto fronts of forecast-informed and no-forecast reservoir operation rules from multiobjective optimization algorithms maximizing power generation and water supply. Subsequently, we vary the installed hydropower plant capacity and capacity-inflow ratio (ratio of active reservoir storage capacity and annual reservoir inflow volume) of the Danjiangkou Reservoir to investigate corresponding forecast values in terms of magnitude and distribution. The results demonstrate that the inclusion of streamflow forecasts predominantly leads to an increase in power generation in wet years, yet the monthly distribution of generation varies with water supply. Additionally, forecast value increases with forecast lead time, resulting in approximately 10, 15, and 18 million yuan annually for forecast lead times of 10, 20, and 30 days, respectively. Finally, forecast value is demonstrated to generally increase as installed capacity increases and decrease as capacity-inflow ratio increases, however, the distribution of forecast value within a year is more sensitive to the capacity-inflow ratio.