甘露情 劉媛華



摘 要:建立有效的空氣質(zhì)量指數(shù)預(yù)測模型,可以為個人出行及相關(guān)部門治理大氣污染提供指導(dǎo)。選取北京市的歷史空氣數(shù)據(jù)以及氣象數(shù)據(jù)作為研究對象,建立基于BP(Back Propagation)神經(jīng)網(wǎng)絡(luò)和SVR(Support Vector Regression)支持向量機回歸的BP-SVR組合預(yù)測模型。首先利用灰狼優(yōu)化算法分別對BP模型和SVR模型參數(shù)進(jìn)行尋優(yōu);然后運用該組合模型對空氣質(zhì)量指數(shù)進(jìn)行預(yù)測。實驗結(jié)果表明,BP-SVR模型的平均絕對百分誤差、均方根誤差、平均絕對誤差均小于單一預(yù)測模型,分別為0.217 5、37.032 0、25.157 5。BP-SVR組合模型具有更高的預(yù)測精度,泛化能力更強,可以對空氣質(zhì)量指數(shù)進(jìn)行有效預(yù)測。
關(guān)鍵詞:空氣質(zhì)量指數(shù)預(yù)測;灰狼算法;BP模型;SVR模型;BP-SVR模型
DOI:10. 11907/rjdk. 201217
中圖分類號:TP301文獻(xiàn)標(biāo)識碼:A 文章編號:1672-7800(2020)010-0080-04
Abstract: Establishing an effective air quality index prediction model can provide guidance for individual travel and related departments to control air pollution. The historical air data and meteorological data of Beijing were selected as research objects, and a BP-SVR combined prediction model based on BP (Back Propagation) neural network and SVR (Support Vector Regression) support vector machine regression was established. First, the gray wolf optimization algorithm was used to optimize the parameters of the BP model and the SVR model, and then the combined model was used to predict the air quality index. Experimental results show that the average absolute percentage error, root mean square error, and average absolute error of the BP-SVR model are smaller than that of a single prediction model, which are 0.2175, 37.032 0, and 25.157 5, respectively. The BP-SVR combination model has higher prediction accuracy and stronger generalization ability, and can effectively predict the air quality index.
Key Words: air quality index prediction; gray wolf algorithm; BP model; SVR model; BP-SVR model
0 引言
我國工業(yè)化進(jìn)程不斷加快,能源消耗量持續(xù)增加,產(chǎn)生大量的污染顆粒,導(dǎo)致我國空氣污染日趨嚴(yán)峻[1-2]。近些年,我國眾多城市出現(xiàn)大規(guī)模霧霾天氣的頻率越來越高,特別是2015年北京發(fā)生史上最嚴(yán)重霧霾,PM2.5濃度峰值接近1 000μg/m3??諝馕廴静坏珪θ梭w健康造成傷害,還會造成交通停滯等問題,對社會經(jīng)濟造成重大損失[3]。目前亟需建立模型對空氣質(zhì)量指數(shù)進(jìn)行有效預(yù)測,減少大氣污染對人體的危害,還可為相關(guān)部門治理大氣污染提供數(shù)據(jù)支持。
國內(nèi)外早期主要從數(shù)值模型和統(tǒng)計模型角度預(yù)測空氣質(zhì)量指數(shù),數(shù)值模型需要對污染物有充分認(rèn)識,數(shù)據(jù)采集難度較大。統(tǒng)計模型計算簡單,然而空氣質(zhì)量數(shù)據(jù)具有非線性特點,用線性模型進(jìn)行預(yù)測難以取得較好結(jié)果?!?br>