HR: 1340h
AN: H43A-0958 [Abstracts]
TI: Probabilistic Long-Term Reservoir Storage Forecasting Using Multi-Objective Genetic Algorithms
AU: * Kim, T
EM: chaucer@yonsei.ac.kr
AF: Yonsei University, Civil Engineering
Hydro Laboratory, Seoul, 120749, Korea, Republic of
AU: Shin, J
EM: ausran@yonsei.ac.kr
AF: Yonsei University, Civil Engineering
Hydro Laboratory, Seoul, 120749, Korea, Republic of
AU: Heo, J
EM: jhheo@yonsei.ac.kr
AF: Yonsei University, Civil Engineering
Hydro Laboratory, Seoul, 120749, Korea, Republic of
AB:
In this study, the implicit stochastic optimization approach is used, and instead of using regression analysis for
the optimization results, the developed reservoir operating rule is found directly from the optimization model. The
piecewise-linear operating rule for the Soyanggang reservoir was developed using a multiobjective genetic
algorithm (NSGA-II) and the synthetic inflow that was generated by time series modeling.
In order to formulate the operating rule effectively, two aspects of the piecewise-linear operating rule are
examined in detail: search space determination and effects of inflow and constraints. First, the upper and lower
limits of the first and last end points are determined by frequency analysis. If the upper and lower limits are simply
set to the storage values corresponding to normal pool and low water levels, respectively, the search space of
NSGA-II would become very large, particularly for the non-flood season. Therefore, each quantile having 1%
exceedance and nonexceedance probabilities is computed by the frequency analysis of the historical storage
record on the first day of a month; these quantities are used as the upper and lower limits of the optimization
model. Furthermore, these limits could be reasonable estimations with slight variations from the historical
maximum and minimum values.
In the case study, the simulation results are obtained by using the developed piecewise-linear operating rule.
Four- and five-segmented operating rules are adopted along with six years of historical inflow data of the
Soyanggang reservoir. The reservoir operation results show that the developed piecewise-linear operating rule
can handle various inflow series that have different characteristics and can generally satisfy the constraints
defined in the optimization model including the constraint of terminal storage. In addition, a probabilistic long-
term reservoir storage forecast is provided. This storage forecast would be useful information since a system
operator is able to evaluate the current status of the reservoir quantitatively, but not qualitatively.
DE: 1880 Water management (6334)
SC: Hydrology [H]
MN: 2007 Fall Meeting