基于CEEMDAN?TCN的短期風(fēng)電功率預(yù)測(cè)研究

摘" 要： 風(fēng)力發(fā)電作為可再生能源的重要組成部分，在電力系統(tǒng)規(guī)劃和日常運(yùn)行中扮演著重要的角色，準(zhǔn)確的短期風(fēng)電功率預(yù)測(cè)對(duì)于電網(wǎng)的穩(wěn)定運(yùn)行和優(yōu)化調(diào)度具有重要意義。為提高短期風(fēng)電功率預(yù)測(cè)的準(zhǔn)確性，提出一種基于自適應(yīng)噪聲完備集合經(jīng)驗(yàn)?zāi)B(tài)分解和時(shí)間卷積網(wǎng)絡(luò)的短期風(fēng)電功率預(yù)測(cè)方法。首先利用自適應(yīng)噪聲完備集合經(jīng)驗(yàn)?zāi)B(tài)分解對(duì)初始風(fēng)電功率數(shù)據(jù)進(jìn)行分解，得到多個(gè)相對(duì)穩(wěn)定的子數(shù)據(jù)序列；然后將其分別作為時(shí)間卷積網(wǎng)絡(luò)的輸入，利用時(shí)間卷積網(wǎng)絡(luò)模型進(jìn)行特征提取和功率預(yù)測(cè)；最后將所有預(yù)測(cè)值進(jìn)行匯總，得到最終的功率預(yù)測(cè)值。使用寧夏某地區(qū)真實(shí)風(fēng)電功率數(shù)據(jù)進(jìn)行驗(yàn)證，并與傳統(tǒng)預(yù)測(cè)模型比較，結(jié)果表明所提方法具有較高的預(yù)測(cè)精度，可為風(fēng)電功率短期預(yù)測(cè)等相關(guān)工作提供相關(guān)參考。

關(guān)鍵詞： 短期風(fēng)電功率預(yù)測(cè)； 自適應(yīng)噪聲的完備集合經(jīng)驗(yàn)?zāi)B(tài)分解（CEEMDAN）； 時(shí)間卷積網(wǎng)絡(luò)（TCN）； 特征提取； 預(yù)測(cè)精度； 時(shí)間序列分析

中圖分類號(hào)： TN911.23?34" " " " " " " " " " " " "文獻(xiàn)標(biāo)識(shí)碼： A" " " " " " " " " " " "文章編號(hào)： 1004?373X（2025）02?0097?06

Research on short?term wind power forecasting based on CEEMDAN?TCN

LI Ao， RAN Huajun， LI Linwei， WANG Xinquan， GAO Yue

（College of Electrical Engineering and New Energy， China Three University， Yichang 443002， China）

Abstract： Wind power generation， as an important component of renewable energy， plays a crucial role in power system planning and daily operation. Therefore， accurate short?term wind power forecasting is crucial for the stable operation and optimized scheduling of electrical grids. In order to enhance the precision of short?term wind power forecasting， a method of short?term wind power forecasting based on complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN） and temporal convolutional networks （TCN） is proposed. The CEEMDAN is used to decompose the initial wind power data， so as to obtain multiple several relatively stable sub?data sequences. The sub?data sequences are used as inputs for TCN， and the TCN model is used to conduct the feature extraction and power forecasting. All predicted values are aggregated to obtain the final power prediction value. The proposed method is verified by the real wind power data from a certain region in Ningxia， and compared with traditional prediction models. The results indicate that the proposed method has high prediction accuracy and can provide relevant references for short?term wind power forecasting and other related work.

Keywords： short?term wind power forecasting; complete ensemble empirical mode decomposition with adaptive noise; temporal convolutional network; feature extraction; prediction accuracy; time series analysis

0" 引" 言

風(fēng)能作為一種重要的低碳、可再生能源，在能源行業(yè)中獲得了廣泛的應(yīng)用[1]。然而，風(fēng)力發(fā)電的高隨機(jī)性和強(qiáng)波動(dòng)性給風(fēng)電并網(wǎng)造成了不利的影響[2]，因此提高風(fēng)電功率預(yù)測(cè)的準(zhǔn)確性對(duì)于電力系統(tǒng)穩(wěn)定運(yùn)行具有重要意義。

目前國(guó)內(nèi)外常用的風(fēng)電功率預(yù)測(cè)方法主要有物理法、統(tǒng)計(jì)法、人工智能法和組合模型法[3]。物理法通過建立風(fēng)電場(chǎng)及外界環(huán)境與風(fēng)電功率之間的物理模型進(jìn)行預(yù)測(cè)[4]，但其受風(fēng)電機(jī)組模型參數(shù)和外界環(huán)境影響較大，預(yù)測(cè)精度較低。……