- 期刊
不同模式因子對優養預測模式最佳化之研究
To Optimize the Prediction Ability of Eutrophication Model by Different Model Factors
摘要
本研究中利用時間序列模式來預測德基水庫未來的優養化變化趨勢，以協助水庫優養化問題的管理與規劃。研究中以單因子實驗設計原理為基礎，尋求預測模式之最佳化。模式因子包括預測模式、預測時距和模式時距3部分：第1 部分是選擇研究中所需之基礎預測模式與模式檢定方法，第2 部分是選取模式之最佳預測時距，第3 部分則是尋求模式之最佳模式時距。結果顯示，相加模式中之虛擬變數法與考慮誤差之加減法，以及相乘模式中不考慮誤差之相乘法有較佳的預測表現；而模式之最佳預測時距則均為「月」至於在模式時距的研究中，發現以相加模式中之虛擬變數法進行預測時，其最佳模式時距為「14 年」，而加減法與相乘法則同樣為「8 年」。整體而言，經預測模式、預測時距與模式時距等3 個因子校正後，模式預測能力平均提升達50%以上，其中又以相加模式中的加減法表現最佳，其預測值最能呈現水庫未來最真實的優養化變化趨勢。
並列摘要
This research was to predict the trophic status of Te-Chi reservoir by the time series model. We hope the results could help the management of eutrophication in reservoir. Single factor design was used to find out the optimum conditions of prediction model. The study involved three parts, which were (1) to select the fundamental prediction models and examination methods, (2) to find out the optimum time period of prediction, and (3) to find out the optimum time period of model. The results showed that the dummy variables method, the addition/subtraction method with error, and the multiplicative method without error had better prediction ability. For all models, the best time period of prediction was ”month”. The investigation from the time period of model exhibited that the optimum time period was ”14 years” in the dummy variables method, and both of the addition/subtraction and multiplicative method were ”8 years”. As a whole, the mean improving rates about the prediction abilities ofmodelsreachedabove50% by the calibration of prediction model, time period of prediction, and time period of model. Additionally, the addition/subtraction model was the best, its prediction values could present the realest trend of trophic status in reservoir.