Studies on Order Series Method with Application to Non-Gaussian Hydrological Time Series
| Autor: | Ming-De Chuang, 莊明德 |
|---|---|
| Rok vydání: | 1999 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 87 Most of tradition methods for generating and forecasting non-Gaussian hydrological time series is established by raw data directly. So, the process of data transformation or building models are always affected by the variation of sample datum,which will cause inappropriate data transformation, bias parameters or misjudge the distribution of noise, etc. In order to overcome these defections of tradition methods, this research proposed “Time Series Separation Theory” and ”Order Series Transformation Theory” by the characters of order series, and application in synthesizing, generating and forecasting hydrological time series. “Time Series Separation Theory” takes time series are composed by data marginal distribution and data order series. So, we can use lots of theoretical distributions and stochastic models to analyze these datum separately ; therefore, it is no need to use non-Gaussian or nonlinear stochastic models to generate and forecast hydrological time series. As a result of these studies, it is found that the order series generation method breakthrough the limitation of tradition linear autogressive model, which can generate both high correlative and high skew time series. Besides, the accuracy of order series forecast method is close to nonlinear instantaneous transformation forecast method (Yu and Wen (1996)), but the calculation speed of computer is almost thirty times to tradition method. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|