考慮單點或多點風速輸入的適應性類神經波浪推算模式

本文建立臺北港風速與示性波高關係的兩種適應性類神經模式，並與迴歸的經驗模式比較波浪模擬的能力。兩模式為單純輸入風速率的適應性類神經ANFIS-U 單點模式及輸入風速分量的ANFIS-(u, v) 單點模式。本文為獲得適應性類神經模式架構內的最佳參數，經測試適合停止的疊代次數為190 次。由模式檢驗指標R^2 及RMSE 發現ANFIS-U 單點模式僅稍佳於迴歸模式，但是ANFIS-(u, v) 單點模式在三種模式中是最佳的。輸入參數增加觀測樁外圍25 km 半徑上W 及NW 的計算風速所建立的多點模式均有助於模式的波浪推算能力，而增加觀測樁外圍25 km 在NW 處的計算風速輸入的多點模式比增加W 處風速的模式更適合來推算波浪。最佳的ANFIS(u, v) 多點模式的R^2 提高至0.8044，RMSE 降至0.3238 m，此結果比ANFIS(u, v) 單點模式有11.0% 的RMSE 改善幅度。

關鍵字

適應性類神經模式；波浪推算；模式適配度；多點風速

並列摘要

The paper developed two adaptive neuro-fuzzy wave models to establish the relation between observed data of wind velocity and wave heights at the Taipei Harbour and compared with the empirical equation in a quadratic form. The ANFIS-U single-point model includes an input variable of observed wind speeds and the ANFIS-(u, v) single-point model uses two input variables, two components of wind velocity. The minimum number of epoch was suggested to be 190 times for obtaining optimal parameters in both ANFIS models. Both indexes of R^2 and RMSE for model fitting show that the ANFIS-U model has similar simulation capacitty with the empirical model and the ANFIS-(u, v) model is the best one among three single-point models. Multi-point ANFIS models with additional inputs of wind data at the position away from the offshore observation pole by 25 km in the W or NW directions have higher simutation capacity than single-point models. The muli-point model with NW wind inputs is a little better than that with W wind inputs. The best ANFIS-(u, v) muli-point model with NW wind inputs has the highest R^2 of 0.8044 and the lowest RMSE of 0.3238 m among all models and decreses RMSE by 11.0% compared with the ANFIS-(u, v) simple-point model.

並列關鍵字

adaptive neuro-fuzzy models ； wave forecasting ； model fitting ； multi-point winds

被引用紀錄

張家羲（2016）。多點風速輸入的FCM-ANFIS波浪推算模式發展〔碩士論文，國立交通大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0030-0803201714434894

國際替代計量

考慮單點或多點風速輸入的適應性類神經波浪推算模式

全文下載

主題瀏覽