The stochastic short-term hydropower generation scheduling considering uncertainty in load output forecasts

Abstract

Load demand forecasts for short-term hydropower generation scheduling (STHGS) has stochastic essence which is impacted by multiple uncertainties from hydrometeorology, streamflow and compensation operation. These dynamically changed uncertainties propagate to startup or shutdown transition, load dispatch and water consumption of hydro-units. However, dynamic uncertainty propagation process is not expounded effectively and its effects on STHGS are ignored to some extent when planners made short-term hydropower plan, leading to potential decision error. Moreover, computation efficiency and precision are also crucial for managers to make reliable decisions and adjustments timely. To this end, framework is developed to provide efficient model solver and identify effect of uncertain load demand forecast on STHGS. The framework contains: (1) stochastic load demand forecast based on martingale model of forecast evolution (MMFE); (2) enhanced shuffled frog leaping algorithm (SFLA) incorporating parallel computing; (3) risk quantification and evolvement revelation based on risk of exceedance (ROE). The STHGS for Three Gorges hydropower station (TGHS) is studied to demonstrate framework. The stochastic STHGS model is efficient solved by algorithms proposed. The propagation of forecast uncertainty and effect of different uncertainty levels and temporal correlations on STHGS is exhibited and fully discussed based on power load scenarios dynamically updated by MMFE.