| Summary: | 碩士 === 國立臺灣大學 === 農業工程學系 === 81 === The main purpose of this study is to investigate the suita-
bility of using a modified self-organization algorithm ,
SGMDH (Stepwise regression Group Method of Data Handling) to
simulate the long-term hydrological process. The ten-day
streamflow data obtained from Te-Chi reservoir (Ta-Chia chi
basin) & Kankou hydrometric station (Tanshui river basin) are
used as the case study. The results both on calibration
and verification stage obtained by SGMDH model and AR model
which is a traditional approach show that the SGMDH model has
less parameters, efficient structure, and small volume error
than the AR model. In the case of using the SGMDH model
for simulating the generating time series from autoregression
(AR) models, it shows that the SGMDH model can have better
performance as the weight of white noise component is
increased in AR models. For improving the model performance in
fitting the streamflow in dry and wet seasons, the SGMDH2
model, which combines the rainfall and streamflow data for
its structure , aggregated with a simple linear function
by means of fuzzy inference modeling is developed and
used to simulate the streamflow of above watersheds. The
results show that the aggre gated model has pertinent
performances both on calibration & verification events
especially in reducing the volume error which is one of the
most important index in the long-term period water resource
planning. Consequentiy , it can concludes that SGMDH model
is simplicity, utility, and great compatibility.
|
|---|
