A Study of Adaptive Network Fuzzy Inference System for Storm-Stage Forecasting in River Basin

• Main Authors: JIUN-FENG FANG, 方俊峰
• Other Authors: YUNG DUAN
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/25193297540607502117

碩士 === 中國科技大學 === 土木與防災應用科技研究所 === 97 === Rivers in Taiwan are generally short and steep. Water flow is extremely scant during dry periods. However, during rainy seasons, floods will rush into the riverbeds in no time, causing abrupt rise of water level and seriously endangering people’s lives and assets. This study will investigate the applicability by using the Adaptive Network Fuzzy Inference System (ANFIS) for floods forecast in Bazhang River. The study examines the ANFIS forecasting of flood stage in the case of five floods and nine storms and concludes from the comparison of forecast results and the observed water stages. From the result of the forecast stage of five typhoons, we get the range of absolute error is from 0.02 to 0.44. The range of relative error is from 0.85 to 9.07. The range of determination coefficient is from 0.7364 to 0.9877. The range of root mean square error is from 0.0217 to 0.764. From the result of the forecast stage of five storms, we get the range of absolute error is from 0.03m to 0.29m. The range of relative error is from 1.32％ to 8.17％. The range of determination coefficient is from 0.5935 to 0.9737. The range of root mean square error is from 0.0335 to 0.5279. These show the credible range. The ANFIS model can accurate forecast the flood stage during floods and storms, and which based on inputs of forecasted precipitation and advanced flood water stage records. The accurate advanced forecasting hour is about three hours. The ANFIS models can make accurate forecast for typhoons and storms.