H41G-1423
Robust multi-objective optimization for short-term reservoir operation under inflow uncertainty
Thursday, 17 December 2015
Poster Hall (Moscone South)
Duan Chen, Oregon State University, Corvallis, OR, United States and Oregon State University
Abstract:
Robust optimization offers optimal solutions that are insensitive to uncertainty. This study proposes a robust multi-objective optimization model for short-term reservoir operation under inflow uncertainty. Variations in prediction of stream inflows are considered as uncertainty and are modelled with a mean inflow surrounded by an uncertainty envelope. The generalized polynomial chaos expansion method is used to connect the inflow uncertainty with the uncertainty of the solutions e.g., objectives. To quantify the uncertainty, the variance of each objective is defined as an additional objective. A recently developed optimization algorithm called BorgMOEA is used as the optimizer to solve this many-objective optimization (more than three objectives) problem. The model finds Pareto-optimal solutions for all objectives, which are used determining robust solutions. The model is applied to a multi-reservoir system for short-term operation under inflow uncertainty. Trade-offs between the mean and variance of the objectives are explicitly explored. Different levels of robust solutions (i.e., robustness) are determined based on the trade-offs and decision making preferences. The proposed robust multi-objective optimization model shows a promising result for robust decision making in reservoir operation under uncertainty.