臺灣能源期刊發行
- 創刊日期:
102年11月30日
- 發行所:
經濟部能源署
- 發行人:
游振偉
- 地址:
台北市復興北路2號13樓
- 電話:
02-2772-1370
- 執行單位:
財團法人工業技術研究院
- 地址:
新竹縣竹東鎮中興路四段195號26館
- 服務專線:
03-5916006
- 服務信箱:
- 總編輯:
王漢英胡均立
- 顧問:
王運銘童遷祥王人謙
- 執行主編:
劉子衙陳志臣
- 編輯委員:
方良吉王錫福朱家齊李堅明李叢禎林師模馬鴻文陳希立廖芳玲廖肇寧劉文獻蕭志同顧洋(依筆畫順序排列)
:::
臺灣能源期刊論文全文
臺灣能源期刊第8卷第3期內容
出刊日期:September, 2021
- 題目
- 使用迴歸與類神經網路法進行風力發電量預測分析
- Title
- Prediction and Analysis of Wind Power by Regression and Neural Network Method
- 作者
- 陳銘宏
- Authors
- Ming-Hong Chen1
- 摘要
- 本研究之目的在於使用迴歸與類神經網路法進行風力發電預測。在NAR (Nonlinear Autoregressive) 模式下,將原始資料拆解並去除部分之後,可有較佳之預測結果,但須同時考慮去除之資料與原始 數據之比例,尤其是小型風機;在NARX (Nonlinear Autoregressive with External Input)模式下,則無 需刪除資料亦可改善預測之結果。本研究接著應用所提出之預報系統至MW等級之風場進行案例分 析。計算之結果顯示,使用NARX模式結合數據拆解前處理程序時,可獲得較佳之結果,而採用小 波轉換法處理數據後,對於預測準確度沒有明顯的影響。
- 關鍵字
- 風能,預報,迴歸,類神經網路
- Abatract
- The purpose of this study is to conduct the wind power prediction by the regression and neural network methods. Results show that smaller errors are achieved with more deleted data set in the Nonlinear Autoregressive (NAR) model. However, the ratio of the deleted data and raw data should be considered simultaneously, especially for the small wind turbine. In the case using Nonlinear Autoregressive with External Input (NARX) model, additional decomposed data set is employed, and the improvement is still achieved without the alteration of raw data. The performance of the proposed prediction model on the MW scale wind farm is also investigated. Results show that better performance is obtained using the NARX model combined with the decomposed data sets, and the resulted mean absolute percentage error (MAPE) is less than 5%. There is no obvious improvement (< 1%) in the prediction by using the method of wavelet transformation.
- Keywords
- wind power, prediction, regression, neural network.