Stochastic optimization operation for cascade hydropower reservoirs by using precipitation forecasts I. Using aggregation-decomposition methodology to reduce high computational costs

XU Wei,ZHANG Chi,PENG Yong,WANG Ben-de
DOI: https://doi.org/10.13243/j.cnki.slxb.2013.08.013
IF: 2.4
2013-01-01
Journal of Hydraulic Engineering
Abstract:This paper presents an aggregation-decomposition Bayesian stochastic dynamical optimization model(AD-BSDP) to reduce the complexity of the model used in cascade reservoirs and to make the solution computationally tractable.Firstly an aggregation flow stochastic dynamical optimization model(AF-SDP) is constructed in this study by considering aggregate inflow and storages of individual reservoirs as state variables.Secondly an aggregation-decomposition stochastic dynamical optimization model(AD-SDP) is developed to reduce computation burden by aggregating inflows and storages of individual reservoirs.Then the AD-BS-DP model is developed by embedding a Bayesian approach in AD-SDP,and the Bayesian approach is used to address forecast uncertainty and inflow uncertainty.This paper use a 10-day precipitation value of the Quantitative Precipitation Forecasts(QPF) by the Global Forecast System(GFS) as input data to run hydrological rainfall-runoff model forecasting the 10-day average inflow of Hun River.And the AD-BS DP model runs iteratively until if reaches steady stage,using backward recursion,to generate the 10-day steady-state operating policy incorporate inflow forecasts.Finally performance of the AD-BSDP is compared with General Operation Chart,AF-SDP,AD-SDP to evaluate the value of the GFS-QPF in hydropower generation and the efficiency of the models.The simulation results demonstrate that the GFS-QPF is beneficial to the Hun River hydropower system and the AD-BSDP model is efficiency than the others.
What problem does this paper attempt to address?