由於大類的大肆開發,造成近幾年來地球暖化非常嚴重,全球各地也因此受到大自然嚴重的反撲,無論是暴風雪、大地震、熱浪、颶風等等,而台灣又地處西太平洋颱風路徑之要衝,經常在每年七至九月間面臨颱風之侵襲,由於颱風所帶來之強勁風勢及豐沛雨量,隨之而來的使台灣許多的淨水場無法負荷而正常供水,因爲豪雨挾帶著大暈泥沙,使原水濁度比一般時候提高了許多倍。鑑於此,加上近年來類神經網路已成爲學術界所廣泛的應用,不只限於應有在資訊科學,在水利工程治水防洪研究上亦有相當的份量。因此,本研究應用類神經網路來探討台灣主要淨水場之原水濁度與與上游水文特性之相關性,以淨水場上游集水區水文環境因子爲輸入因子,以原水濁度爲輸出因子,本研究發現往水文因子中以降雨對濁度有頗高的敏感性,相關係數般高有達至0.91;藉此結果,後續研究即能暴雨設計而建立淨水場供水預警系統,以解決民生缺水之苦。
Water resource management is more and more important in 21(superscript st) Century. Global climate change increases the occurring frequency of extreme weather events. Typhoon frequently attacks Taiwan in recent years. Heavy rainstorm can cause landslides and debris flows, which threaten the local environment and citizens' lives. High turbidity water would impact water treatment plants and increase the uncertainty of water supply. This study applies Back-Propagation Network (BPN) model to assess the relationship between hydrologic properties in upstream watershed and the turbidity of raw water in three major water treatment plants in Taiwan. The result shows that rainfall property is highly related to turbidity of raw water in Ban-Xin water treatment plant. Early prediction systems of turbidity can be established according to the results of relationship analysis.