Stochastic Dynamic Programming Model for Reservoir Optimization

What is it about?

The monthly time step stochastic dynamic programming (SDP) model has been applied to derive the optimal operating policies of Ukai reservoir, a multipurpose reservoir in Tapi river basin, India. The initial reservoir storages and inflows into the reservoir in a particular month are considered as hydrological state variables. Since flood control and irrigation are the two major purposes of this reservoir, the SDP model is developed with the objective of minimizing annual sum of squared deviations of actual releases and actual storages from their respective target values. The uncertainty in the prediction of inflows into the reservoir system is addressed by incorporating the inflow transition probabilities. In this study, the effect of state variable discretization in deriving optimal operating policies using SDP for the reservoir is investigated. From the model results, it is inferred that, by adopting unequal interval storage discretization approach over equal interval storage discretization approach, there is an improvement of about 8–58% in the values of the objective function. The optimal operating policies derived using unequal interval storage discretization have been expressed in terms of final storages (levels) for each month for various combinations of inflows and initial storages.

Featured Image

Why is it important?

Key Highlights: (i) Introducing and adopting unequal interval (equi-probable) discretization over the conventional equal interval discretization of reservoir storage in Stochastic Dynamic Programming (SDP) optimization model. (ii) Proposing operation rule curves for each month based upon the combinations of different inflow and storage states for reservoir optimization considering flood control also.