A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1991

In the present study, two dynamic programming models, Deterministic Dynamic programming (DDP) and Stochastic Dynamic Programming (SDP) are developed to find the optimum operation policies of a flood control reservoir system. Ubol Ratana Reservoir was selected as a case study. The purpose of flood control was given priority for Ubol Ratana Reservoir. Both models were derived by using the different alternative objective function with the same hypothetical loss function defining the flood loss due to nonideal reservoir operation for flood control. In this study, Deterministic Dynamic Programming model used a min-max type of objective function. The stochastic nature of inflows to the reservoir was explicitly taken into account in SDP model. The synthetic inflow data generated by AR(2) model for 100 years were applied in the second order Markov Chain model (lag two), which would be used to describe the stochastic nature of inflows to the reservoir. Transition probability matrices, estimated from the generated synthetic inflows with the second order Markov chain, are used to incorporate serial correlation among inflows to the reservoir. To test the usefulness of the both DDP and SDP model in reservoir operation policies reservoir operation simulation models were constructed. The optimal reservoir operation policies derived by DDP and SDP models were then applied in the operation of simulation model and their performances were evaluated and compared. Finally, the operational results obtained from the reservoir operation simulation models by using DDP and SDP based optimal operation policies were compared with the operational results obtained from the simulation of Ubol Ratana Reservoir operation based on the existing rule curve and two principal targets of flood control strategies previously suggested.