摘要 本研究採用ARMA及GM (1, 1) 模型,分別預測舊濁水溪溶氧 (Dissolved Oxygen, DO) 、生化需氧量 (biochemical oxygen demand, BOD) 、化學需氧量(chemical oxygen, COD) 、及氨氮 (ammonia, NH3-N) 之濃度及比較其結果。以ARMA預測舊濁水溪DO時平均絕對百分誤差 (mean abs perc Error, MAPE) 介於37.04至98.50之間,相關係數 (correlation coefficient, R) 介於-0.59至0.62之間;預測BOD時MAPE介於142.29至511.79之間,預測時的R值介於-0.57至0.19之間;預測COD時MAPE介於106.72至300.09之間,預測時的R值介於-0.51至0.13之間;預測NH3-N時MAPE介於63.81至140.68之間,預測時的R值介於-0.50至-0.05之間。以GM(1, 1)預測舊濁水溪DO時MAPE介於25.31至36.38之間,預測時的R值介於0.39至0.73之間;預測BOD時MAPE介於49.28至82.00之間,預測時的R值介於0.28至0.51之間;預測COD時MAPE介於48.74至117.38之間,預測時的R值介於-0.11至0.46之間;預測NH3-N時MAPE介於45.64至60.05之間,預測時的R值介於0.29至0.37之間。無論預測DO、BOD、COD或NH3-N,皆以GM (1, 1, x(0)) 、GM (1, 1, a) 、GM (1, 1, b) 三種GM (1, 1) 模型較佳。
Abstract This study used ARMA and GM (1, 1) models to predict the concentrations of DO (Dissolved Oxygen), BOD (biochemical oxygen demand), COD (chemical oxygen), and ammonia (NH3-N) in the Old Jhuoshuei River and compare the results. the MAPE of using ARMA to predict the DO in the Old Jhuoshuei is between 37.04 and 98.50. The CR (correlation coefficient) is between -0.59 and 0.62; the MAPE of predicting the BOD is between 142.29 and 511.79; the R value of the prediction is between -0.57 and 0.19. The MAPE of predicting COD is between 106.72 and 300.09; the R value of prediction is between -0.51 and 0.13. The MAPE of predicting NH3-N is between 63.81 and 140.68; the R value of prediction is between -0.50 and -0.05. The MAPE of using GM(1, 1) model to predict the DO of the Old Jhuoshuei River is between 25.31 and 36.38; the R value of prediction is between 0.39 and 0.73. The MAPE of predicting BOD is between 49.28 and 82.00; the R value of prediction is between 0.28 and 0.51. The MAPE of predicting COD is between 48.74 and 117.38; the R value of prediction is between -0.11 and 0.46. The MAPE of predicting NH3-N is between 45.64 and 60.05; the R value of prediction is between 0.29 and 0.37. Regardless of predicting DO, BOD, COD or NH3-N, the three GM (1, 1) models of GM (1, 1, x (0)), GM (1, 1, a), GM (1, 1, b) are relatively better in performance.