Daily Stream-flow Discharge Forecasting Model for Qian-Feng Bridge Station on Wu River

碩士 === 逢甲大學 === 土木及水利工程所 === 90 === Distinct wet and dry seasons, along with uneven spatial and temporal distribution of rainfall, lead a great change of stream runoff in Taiwan. The reliable water demand is more hopeful owing to the industry and commercial development. So daily stream-flow discha...

Full description

Bibliographic Details
Main Authors: Hsu-Chung Liao, 廖旭崇
Other Authors: Chang-Shian Chen
Format: Others
Language:zh-TW
Published: 2002
Online Access:http://ndltd.ncl.edu.tw/handle/qny9v8
Description
Summary:碩士 === 逢甲大學 === 土木及水利工程所 === 90 === Distinct wet and dry seasons, along with uneven spatial and temporal distribution of rainfall, lead a great change of stream runoff in Taiwan. The reliable water demand is more hopeful owing to the industry and commercial development. So daily stream-flow discharge data are used in operation in order to dear with the variable water demand. In this study, the daily stream-flow discharge data of Qian-Feng bridge station on Wu River are used to establish forecasting model. The research tries to utilize time series analysis to simulate and control system variation characteristics. In the analysis of linear time series models, ARMA model, Outlier Detection, and adding rainfall characteristic’s TFN model are set up. In non-linear models, to select the parameters of models is deduced through Genetic Algorithm. TAR model, BL model, and TARBL model with the characteristics of both TAR and BL models will be established. In general, the TFN model with rainfall characteristics can be better than any other linear models was identified and verified. As to the capability of forecasting in wet and dry seasons, TAR model is better than any other non-linear models except the peak value of flood. The BL model cannot get a good result enough because bilinear term may result in the downward vibration. The TARBL model, with piecewise and simulated high flood discharge properties, cannot get a good result as well as BL model. In conclusion, the ability to forecast daily stream-flow discharge, the non-linear model can be better than the linear model and is consistent with the complex data of physical characteristics.