The aim of the paper is to describe a sequential procedure for reservoir management and to compare it with an existing method, the alternate stochastic optimization (ASO). Following the ASO approach, at each time step the decision on the release is based upon present storage, present and future target releases, and a number of ‘possible’ future inflows given by synthetic generation. In the approach described here the decision is also based on present storage and present and future targets, but the hydrological input is represented by the ‘most likely’ future inflow series, namely, by that inflow series forecast by a real time predictor (in the sense of Box and Jenkins). The management performance of the two methods is substantially the same, but the present procedure allows a very conspicuous computational saving. As a matter of fact, the optimization model (a dynamic program which gives the best decision) is run only once at each time step while it is run 20 or 30 times (depending upon the number of synthetic series used) in the ASO approach.
Real-time predictor versus synthetic hydrology for sequential reservoir management
GUARISO, GIORGIO
1979-01-01
Abstract
The aim of the paper is to describe a sequential procedure for reservoir management and to compare it with an existing method, the alternate stochastic optimization (ASO). Following the ASO approach, at each time step the decision on the release is based upon present storage, present and future target releases, and a number of ‘possible’ future inflows given by synthetic generation. In the approach described here the decision is also based on present storage and present and future targets, but the hydrological input is represented by the ‘most likely’ future inflow series, namely, by that inflow series forecast by a real time predictor (in the sense of Box and Jenkins). The management performance of the two methods is substantially the same, but the present procedure allows a very conspicuous computational saving. As a matter of fact, the optimization model (a dynamic program which gives the best decision) is run only once at each time step while it is run 20 or 30 times (depending upon the number of synthetic series used) in the ASO approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.