【正文】 In this paper, we adopt the method of neural works to GRNN to research wind power forecasting. First, we study the history wind power data of one wind farm and analysis them, then truncate the wind power data and build the GRNN neural works model with time series。 generalization ability。 expansion coefficient 目 錄 摘要 .................................................................................................................................................. I Abstract ...........................................................................................................................................II 1 緒論 .............................................................................................................................................. 1 引言 ........................................................................................................................................... 1 課題研究背景 ........................................................................................................................ 1 課題研究意義 ........................................................................................................................ 2 風(fēng)電場功率預(yù)測的國內(nèi)外研究現(xiàn)狀 ....................................................................................... 3 國外研究現(xiàn)狀 ........................................................................................................................ 3 國內(nèi)研究現(xiàn)狀 ........................................................................................................................ 3 風(fēng)電場功率的預(yù)測方法 ........................................................................................................... 4 物理方法 ................................................................................................................................ 4 統(tǒng)計(jì)方法 ................................................................................................................................ 4 學(xué)習(xí)方法 ................................................................................................................................ 5 三種方法的比較 .................................................................................................................... 5 本文的主要工作 ....................................................................................................................... 5 本章小結(jié) ................................................................................................................................... 6 2 徑向基神經(jīng)網(wǎng)絡(luò)的基本理論 ...................................................................................................... 7 徑向基函數(shù)（ RBF）神經(jīng)網(wǎng)絡(luò) ................................................................................................ 7 RBF 神經(jīng)網(wǎng)絡(luò)概述 ................................................................................................................ 7 RBF 神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)模型 ........................................................................................................ 7 廣義回歸（ GRNN）神經(jīng)網(wǎng)絡(luò) ................................................................................................ 8 GRNN 神經(jīng)網(wǎng)絡(luò)概述 ............................................................................................................ 8 GRNN 神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu) ............................................................................................................ 9 GRNN 神經(jīng)網(wǎng)絡(luò)的理論基礎(chǔ) .............................................................................................. 10 3 預(yù)測數(shù)據(jù)及結(jié)構(gòu)參數(shù)的預(yù)處理 ................................................................................................ 12 歷史數(shù)據(jù)的預(yù)處理 ................................................................................................................. 12 預(yù)測誤差分析 ......................................................................................................................... 12 神經(jīng)網(wǎng)絡(luò)的泛化能力 ............................................................................................................. 13 偏差 方差分解 ..................................................................................................................... 13 “ 欠擬合 ” 與 “ 過擬合 ” .................................................................................................. 14 神經(jīng)網(wǎng)絡(luò)模型的評估 .......................................................................................................... 15 GRNN 網(wǎng)絡(luò)設(shè)計(jì)要點(diǎn) ............................................................................................................. 16 SPREAD 參數(shù)的物理本質(zhì) .................................................................................................. 16 選擇 SPREAD 的方法 ......................................................................................................... 16 4 基于 GRNN 神經(jīng)網(wǎng)絡(luò)的風(fēng)功率預(yù)測建模方法 ....................................................................... 20 問題描述 ................................................................................................................................. 20 數(shù)據(jù)預(yù)處理 ............................................................................................................................. 20 網(wǎng)絡(luò)設(shè)計(jì)與訓(xùn)練 ..................................................................................................................... 21 網(wǎng)絡(luò)模型的評估方法 .......................................................................................................... 21 網(wǎng)絡(luò)訓(xùn)練 .............................................................................................................................. 21 訓(xùn)練參數(shù)的選擇 .................................................................................................................. 22 5 基于 GRNN 神經(jīng)網(wǎng)絡(luò)的風(fēng)功率預(yù)測仿真應(yīng)用 ....................................................................... 24 仿真背景 ................................................................................................................................. 24 仿真實(shí)驗(yàn) ....