頁籤選單縮合
題 名 | 非常態週期性水文序列之預處理及其在預測上之應用=Prehanding of Non-Normal Periodic Hydrologic Series and its Application to Forecasting |
---|---|
作 者 | 李宗仰; 林淑真; | 書刊名 | 臺灣水利 |
卷 期 | 46:2=182 1998.06[民87.06] |
頁 次 | 頁71-84 |
分類號 | 443.046 |
關鍵詞 | Box-Cox轉換; 無母數週期標淮化; 常態性; 定常性; Box-cox transformation; Non-parametric periodic standardization; Normality; Stationarity; |
語 文 | 中文(Chinese) |
中文摘要 | 一序率水文模式的建立,通常基於常態性與定常性的兩個基本假設,然而水文序 列經常是非常態與非定常之過程,為能使其套用於模式,本文針對非常態週期性時間序列提 出一預處理程序, 方法是結合 Box-Cox 轉換與標準化之觀念,使其同時地轉換至最大常態 性並完成定常性。含雨量及流量二水文量共 20 組的水文時間序列將被用於驗証所述之方法 ,而其中的 6 組將用於建立模式並給於預測。研究結果顯示:結合 Box-Cox 轉換與標準化 之資料預處理方式,可最佳逼近或滿足常態性且完成了定常性;而經預處理後的水文序列, 可有效降低模式的複雜度且在預測的精度上亦能提昇。此外,在標準化程序中可再度改善僅 以轉換序列所逼近得之常態性,特別是在端點值的修正上更為顯著。 |
英文摘要 | Many stochastic models are developed based on assumptions that data series are normally distributed and/or stationary, but hydrologic data are generally non-normal and/or non-stationary. This work studied a prehanding technique for a non-normal periodic time series. The Box-Cox procedure and non-parametric periodic standardization were utilized to transform data series to maximum normality and stationarity. Twenty hydrologic series, including rainfall and flow, were used to identify the decribed method. Six series of those were used for modelling and prediction. The results show that the optimal transformation with combination of standardization can bring the data near normality and achieve stationarity. The prehanded data are employed to identify the model and to diminish the complexity in modeling and to promopt the accuracy of forecasting. Besides, the standardized process further improves normality, especially, on the two sides of the probability plot. |
本系統中英文摘要資訊取自各篇刊載內容。