目標函數對連續型降雨－逕流模式率定之影響

本研究旨在探討目標函數對率定連續型降雨－逕流模式之影響。文中採用歐洲廣泛使用之連續型降雨－逕流模式（HBV模式）為模式架構，並做適當修正以適合台灣本土。一般於降雨－逕流模式中較常使用之目標函數爲均方根誤差與平均絕對百分比誤差，但由於上述目標函數無法兼顧洪水與枯水期間流量之模擬，因此本研究提出多階段平均絕對百分比誤差爲目標函數進行比較。模式率定之方法則採用目前頗具效率之全域最佳參數搜尋法（SCE法）。本研究發現：多階段平均絕對百分比誤差能同時滿足洪水與枯流量之模擬精度，爲適合模式率定之目標函數。

關鍵字

連續型降雨－逕流模式；目標函數；全域最佳參數搜尋法

並列摘要

The purpose of this study is to investigate the effects of objective functions on the calibration results of a continuous rainfall-runoff model. The structure of HBV model used extensively in Europe is modified for application in Taiwan. Both root mean square error and mean absolute percent error are commonly used as the objective functions for model calibration. As they cannot well calibrate the high and low flow simultaneously, this study proposes a new objective function, referred as multistage mean absolute percent error. The SCE algorithm is adopted for model calibration to search the global optimal parameters. The study concludes that the multistage mean absolute percent error is a suitable objective function for model calibration, as it can well calibrate both high and low flow of observed hydrograph.

並列關鍵字

Continuous rainfall-runoff model ； Objective functions ； Global optimization algorithm

主題瀏覽