A new reservoir operation policy generation method (SOSM) using scenarios optimization (SO) and surrogate model (SM)

Abstract

In this paper, a model combining Scenarios Optimization (SO) and Surrogate Model (SM) using Radial basis Function (RBF) SOSM is developed for reservoir operation optimization. In order to incorporate the inflow uncertainties in optimal operation policies, different inflow scenarios are generated with conditional probability on historical data firstly and then deterministic DP is applied to select the optimal release policy for each inflow scenario. The solutions from DP, as well as the corresponding inflow and storage, are used to plot a curve surface. This surface is approximated by SM using RBF, from which, the optimal release can be read according to current storage of the reservoir and the predictive inflow. This SOSM approach is applied to the Three Gorges Project (TGP), and weekly operating rules are determined for one year. The result shows this approach is valid and valuable in selecting reservoir operating rules. © 2012 Springer-Verlag GmbH Berlin Heidelberg.