《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 其他 > 設(shè)計(jì)應(yīng)用 > 基于XGBoost與LightGBM集成的 電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型
基于XGBoost與LightGBM集成的 電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型
2022年電子技術(shù)應(yīng)用第9期
吳 丹1,雷 珽1,李芝娟2,王 寧3,段 艷3
1.國(guó)網(wǎng)上海市電力公司,上海200122;2.浦東供電公司,上海200122;3.同濟(jì)大學(xué) 汽車學(xué)院,上海201804
摘要: 隨著電動(dòng)汽車規(guī)模化發(fā)展,充電站負(fù)荷對(duì)電網(wǎng)造成一定影響,為保障電網(wǎng)平穩(wěn)運(yùn)行,提出一種基于極端梯度提升(eXtreme Gradient Boosting,XGBoost)與輕量級(jí)梯度提升機(jī)(Light Gradient Boosting Machine,LightGBM)融合的電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型。該方法運(yùn)用Stacking集成學(xué)習(xí)的策略:首先根據(jù)時(shí)間特征與歷史負(fù)荷數(shù)據(jù)采用XGBoost與LightGBM算法構(gòu)建負(fù)荷預(yù)測(cè)的基學(xué)習(xí)器,然后采用嶺回歸(Ridge Regression,RR)算法將基學(xué)習(xí)器的輸出結(jié)果進(jìn)行融合之后輸出負(fù)荷預(yù)測(cè)值。為了對(duì)比多種不同的負(fù)荷預(yù)測(cè)模型,采用上海市嘉定區(qū)的充電站訂單數(shù)據(jù)進(jìn)行試驗(yàn),結(jié)果表明,該方法所構(gòu)建的負(fù)荷預(yù)測(cè)模型相比單一算法模型具有更高的預(yù)測(cè)準(zhǔn)確度,對(duì)電網(wǎng)平穩(wěn)運(yùn)行有一定理論及實(shí)用價(jià)值。
中圖分類號(hào): TM910.6;U469.72
文獻(xiàn)標(biāo)識(shí)碼: A
DOI:10.16157/j.issn.0258-7998.212316
中文引用格式: 吳丹,雷珽,李芝娟,等. 基于XGBoost與LightGBM集成的電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型[J].電子技術(shù)應(yīng)用,2022,48(9):44-49.
英文引用格式: Wu Dan,Lei Yu,Li Zhijuan,et al. Electric vehicle charging load forecasting based on XGBoost and LightGBM integration model[J]. Application of Electronic Technique,2022,48(9):44-49.
Electric vehicle charging load forecasting based on XGBoost and LightGBM integration model
Wu Dan1,Lei Yu1,Li Zhijuan2,Wang Ning3,Duan Yan3
1.State Grid Shanghai Municipal Electric Power Company,Shanghai 200122,China; 2.State Grid Shanghai Pudong Electric Power Supply Company,Shanghai 200122,China; 3.Institute of Automobile,Tongji University,Shanghai 201804,China
Abstract: With the scale development of electric vehicles, the load of charging stations has a certain impact on the power grid. In order to ensure the power grid run steadily, an electric vehicle charging load forecasting model based on the integration of eXtreme Gradient Boosting(XGBoost) and Light Gradient Boosting Machine(LightGBM) is proposed. This method uses the strategy of stacking integrated learning. Firstly, the base models of load forecasting are constructed based on XGBoost and LightGBM respectively. And then Ridge Regression(RR) algorithm is used to fuse the output results of the base models, the fusion result is the load forecasting value. Based on a variety of different load forecasting models, comparative experiments are carried out with the order data of charging station located in Jiading District, Shanghai. The results show that the load forecasting model constructed by this method has higher forecasting accuracy than the model based on single algorithm, and has certain theoretical and practical value for the smooth operation of power grid.
Key words : electric vehicle;load forecasting;Stacking integrated learning

0 引言

    近年來(lái)電動(dòng)汽車的保有量快速上升,電動(dòng)汽車規(guī)?;瘜?duì)電網(wǎng)的輸電網(wǎng)絡(luò)、配電網(wǎng)絡(luò)、充電設(shè)施等多方面帶來(lái)影響[1-8],因此準(zhǔn)確的電動(dòng)汽車充電負(fù)荷預(yù)測(cè)對(duì)于電網(wǎng)平穩(wěn)運(yùn)行具有重要意義。

    電動(dòng)汽車充電負(fù)荷預(yù)測(cè)是根據(jù)過(guò)去一段時(shí)間的用電負(fù)荷及日期類型等相關(guān)數(shù)據(jù)預(yù)測(cè)未來(lái)一段時(shí)間的用電負(fù)荷[9],構(gòu)建準(zhǔn)確的電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型不僅有利于電網(wǎng)對(duì)充電站的充電負(fù)荷進(jìn)行調(diào)度與管理,也有利于充電站制定科學(xué)的運(yùn)營(yíng)計(jì)劃。不少國(guó)內(nèi)外學(xué)者從用戶端及車端出發(fā)對(duì)電動(dòng)汽車的充電負(fù)荷預(yù)測(cè)展開(kāi)了研究[10-18],通過(guò)融合電動(dòng)汽車出行特征、用戶行為特點(diǎn)和道路交通狀況等因素,建立電動(dòng)汽車充電負(fù)荷預(yù)測(cè)模型。真實(shí)的充電過(guò)程與從車端仿真結(jié)果存在差異,所以從充電站端得到的負(fù)荷預(yù)測(cè)結(jié)果比車端更能真實(shí)反映電動(dòng)汽車充電對(duì)電網(wǎng)造成的影響。目前從充電站端對(duì)充電負(fù)荷進(jìn)行預(yù)測(cè)的相關(guān)研究較少,并多數(shù)是以深度學(xué)習(xí)算法構(gòu)建負(fù)荷預(yù)測(cè)模型,具有一定局限性,例如文獻(xiàn)[19]采用模糊聚類分析與BP神經(jīng)網(wǎng)絡(luò)相結(jié)合的方法建立電動(dòng)汽車充電負(fù)荷的短期預(yù)測(cè)模型,文獻(xiàn)[20]采用隨機(jī)森林與神經(jīng)網(wǎng)絡(luò)相結(jié)合的方法建立電動(dòng)汽車充電站短期負(fù)荷的預(yù)測(cè)模型,深度學(xué)習(xí)算法在輸入序列較長(zhǎng)時(shí)存在梯度消失問(wèn)題,模型無(wú)法克服對(duì)異常值敏感的缺點(diǎn),導(dǎo)致模型預(yù)測(cè)準(zhǔn)確度變差。

    針對(duì)上述問(wèn)題,本文從充電站端的數(shù)據(jù)出發(fā),通過(guò)挖掘電動(dòng)汽車充電負(fù)荷隨時(shí)間的變化規(guī)律,提取負(fù)荷影響因素作為模型的輸入特征。為了實(shí)現(xiàn)較高的負(fù)荷預(yù)測(cè)準(zhǔn)確度,本文采用數(shù)據(jù)挖掘比賽中表現(xiàn)優(yōu)異的XGBoost與LightGBM算法分別構(gòu)建負(fù)荷預(yù)測(cè)模型,再結(jié)合Stacking集成學(xué)習(xí)的策略,利用嶺回歸模型將XGBoost與LightGBM模型的輸出結(jié)果進(jìn)行融合之后再輸出。實(shí)驗(yàn)結(jié)果表明,XGBoost與LightGBM模型實(shí)現(xiàn)了較高的預(yù)測(cè)準(zhǔn)確度高,再采用Stacking集成學(xué)習(xí)方法將XGBoost與LightGBM模型的預(yù)測(cè)結(jié)果進(jìn)行融合后,模型的預(yù)測(cè)準(zhǔn)確度得到了進(jìn)一步提升。




本文詳細(xì)內(nèi)容請(qǐng)下載:http://theprogrammingfactory.com/resource/share/2000004919。




作者信息:

吳  丹1,雷  珽1,李芝娟2,王  寧3,段  艷3

(1.國(guó)網(wǎng)上海市電力公司,上海200122;2.浦東供電公司,上海200122;3.同濟(jì)大學(xué) 汽車學(xué)院,上海201804)




wd.jpg

此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載。