| Author | Nguyen Tan Danh |
| Call Number | AIT Thesis no. WM-95-13 |
| Subject(s) | Hydraulics
|
| Note | A thesis submitted in partial fulfilment of the requirements for the degree of Master of Engineering, School of Civil Engineering |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. WM-95-13 |
| Abstract | The study considers two existing models: Tank Model and Back Propagation Neural Network Model (BPNN Model). Another model called Tank- AR(l) Model, which is a stochasticdeterministic model, was developed in this study to improve the results calculated by the Tank Model. These three models are applied to forecast daily discharges from daily rainfall and daily evaporation data for two basins: Da Nhim and La Nga, in Vietnam. The concept behind the Tank- AR(l) Model is that a forecast value is considered to consist of two components: one is the output of a deterministic model (Tank Model), and the other is an error. The error is defined as the difference between observed discharge and the correspondent discharge forecast by the deterministic model. The error is simulated and forecast by a stochastic model, which is an Autoregressive Model of order one [AR(l)]. The Tank Model although simple but yields good outputs. In forecasting discharge of Da Nhim Basin, due to the variety of the hydrological years the results were not very good in general case, but better in specific cases which characterise the type of hydrological years. It is also noted that forecast discharge for Da Nhim in one day ahead is difficult because its concentration time is only 12 hours at the most. For La Nga Basin, good agreement between forecast and observed discharges was obtained and forecasting discharges with one day ahead is possible and successful. The Tank-AR(l) Model does improve the forecast values. It is realised that the goodness of the result from the Tank-AR(l) depends so much on the performance of the Tank Model. The Tank-AR(l) can improve to some extent the discrepancies in the discharge forecast by the Tank Model. The BPNN Model is particularly promising. In forecasting discharges for Da Nhim Basin and for La Nga Basin, the results were much better compared to the results from the Tank and Tank-AR(l) Models. Another good point found out in this study is that the discharges can be forecast from rainfall and evaporation of all the stations within a basin instead of using mean areal values as in deterministic lumped models. The BPNN Model learns very fast and produces very good outputs. The learning rate of 0.5 and the learning cycle of 500 were found appropriate in this study. The forecast results of the BPNN Model were enhanced by means of a BPNN-AR(l) Model, whose concept is the same as that of the Tank-AR(l) Model. The model provided very good improvement in the accuracy of the forecast values. It is suggested that the three models could be used to forecast daily discharge of the two basins. Among them the BPNN is highly recommended. It is also recommended to apply these models to other basins. |
| Year | 1995 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. WM-95-13 |
| Type | Thesis |
| School | School of Civil Engineering |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Water Engineering and Management (WM) |
| Chairperson(s) | Gupta, Ashim Das; |
| Examination Committee(s) | Huynh Ngoc Phien;Loof, Rainer;Ammentorp, H.C.; |
| Scholarship Donor(s) | DANIDA; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 1995 |