STUDY ON RESERVOIR EUTROPHICATION FORECASTING MODEL BY SELF-ORGANIZATION ALGORITHM

• Main Authors: Wen-Yang Li, 李文揚
• Other Authors: Pei-Hwa Yen
• Format: Others
• Language: zh-TW
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/76069167842961913965

碩士 === 國立成功大學 === 水利及海洋工程學系 === 88 === The capacity of reservoir is one of the most important water resources in Taiwan, and reservoir’s water quality is related much closely with human’s activities. So, the eutrophication of reservoir is paid much attention in water quality control. A framework based on GMDH (Group Method of Dada Handling) is proposed in this paper to forecast the water quality of specific reservoir. Then, step regression GMDH(SGMDH) has been introduced in this framework to improve the GMDH algorithm which can easily induced the high order nonlinear terms so as to reduce the benefit of predicting procedure. The eutrophication situation of Fei-tsui reservoir Taipei, TAIWAN, during the year of 1988~1998 is the database for this water quality forecasting simulation. Monthly Carlson’s TSI will be the output in this forecasting model to judge the water quality of reservoir capacity. Mean monthly temperature and discharge inflow data of reservoir of 2 months prior as well as Carlson’s TSI values of 4 months prior are the most suitable combination of input variables in this water quality-forecasting model. This optimun water quality-forecasting model is developed by the progressive monthly data input of 3 year (i.e., 36 groups of input data). Hence, monthly information of Fei-tsui reservoir during the year of 1988~1990 were used to established the forecasting model and predict the eutrophication situation of Fei-tsui reservoir through the year of 1991 to 1998 respectively. Results show that the SGMDH is the prefer algorithm to make the Carlson’s TSI forecasting and most of the measurement TSI value are in the confidence interval of 95% in this specific reservoir. Time variant properties of eutrophication of Fei-tsui reservoir were tested as well in this paper. An update water quality-forecasting model was carried out to renew the forecasting value by using the input data of 3 year prior each year of 1992 to 1998. Good prediction results could be obtained except several months in 1994 and 1996 respectively with few variances from the predicted value estimated by the original time-invariant model. The simulation results through the year of 1991 to 1998 shows that the SGMDH model has a good validity in Carlson’s TSI forecasting of water quality in specific reservoir capacities.