| Summary: | 碩士 === 逢甲大學 === 土木及水利工程所 === 92 === With dramatic variations both in temporal and spatial scale, the difference between plentiful and withered discharge is greatly in Taiwan. On temporal about 80% concentration at May to October is plentiful water period, and November to April is withered water period. The rainfall is few in south Taiwan, only 10% precipitation declines in the withered water period; and on spatial, the total amount in north Taiwan is more than south Taiwan. The estimate of the 10-day discharge is important to the reservoir operation; the accurate estimate 10-day discharge will be advantageous to allocate the water resource.
This text use 10-day discharge data with Gou-Pin river Li-Lin bridge station and LAO-NUNG station, it establish line time sequence analysis list change to measure an ARMA model, double to change to measure a conversion function noise TFN model, dry and wet season to separate the ARMA and the TFN mode, and the gray GM (1,1) model and the Artificial Neural-Fuzzy Inference system, and analysis of these models.
In general, the TFN model with rainfall characteristics was identified and verified the results better than the ARMA model. In our three tested models, the predicted results of ANFIS model are the most stable and steady. Therefore, as water resources allocation, using ANFIS model is better choice when predicting inflow on 10-day discharge of the real-time reservoir operation.
|
|---|
