考慮集水區物理機制之類神經網路降雨逕流模式
| Autor: | WEI-CHIAO HUNG, 洪暐樵 |
|---|---|
| Rok vydání: | 2004 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 92 Since the watershed rainfall-runoff process is a nonlinear system and exhibit spatial and temporal variability, the artificial neural network (ANN) model, which is suitable for simulating nonlinear processes, has been widely applied to hydrology fields. The training of the ANN network is based completely on the reliability of the available hydrologic records. If the training data of the ANN model can provide more physical phenomena of the watershed, then the model is promising to have a better simulation for the rainfall-runoff processes. The objective of this study is to develop a more efficient ANN model for watershed rainfall-runoff simulating through physical insight. Hydrologic records from four watersheds, namely, Heng-Chi, Wu-Tu, Sung-Mao, Nan-Hu, were collected for analysis. A validated physically based runoff model, the KW-GIUH model, was adopted to identify the hydrologic characteristics of the selected storm events. Hourly rainfall and discharge records were aggregated into different input vectors for ANN model network training for rainfall-runoff simulating. In order to considering extreme flood conditions, a design hyetograph method associated with the KW-GIUH model were proposed to simulate the runoff conditions resulting from extreme storms. The generated severe floods data were then included into the training data sets for further ANN model network training. The analytical results show that the developed ANN model can provide good simulation results for rainfall-runoff processes in the study watersheds. Key words: ANN model, physical insight, rainfall-runoff simulation |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|