《電子技術(shù)應(yīng)用》
您所在的位置:首頁 > 其他 > 設(shè)計應(yīng)用 > 基于互信息變量選擇的燃煤機組SCR脫硝系統(tǒng)PSO-ELM建模
基于互信息變量選擇的燃煤機組SCR脫硝系統(tǒng)PSO-ELM建模
網(wǎng)絡(luò)安全與數(shù)據(jù)治理 9期
張瑾,姜浩,,金秀章
(華北電力大學(xué)控制與計算機工程學(xué)院,,河北保定071003)
摘要: 針對燃煤機組SCR脫硝系統(tǒng)出口NOx濃度存在測量滯后以及吹掃時數(shù)據(jù)失真等問題,,提出了一種基于特征提取和粒子群算法(PSO)優(yōu)化極限學(xué)習(xí)機(ELM)超參數(shù)的燃煤機組SCR脫硝系統(tǒng)模型。利用互信息(MI)進行時間遲延補償,,采用最大相關(guān)最小冗余(mRMR)方法篩選輔助變量,通過PSO優(yōu)化算法確定ELM最優(yōu)超參數(shù)并建立預(yù)測模型,,最后進行對比驗證,。仿真結(jié)果表明:采用本文方法所建立的PSO-ELM預(yù)測模型的均方誤差和相關(guān)系數(shù)分別為0.931 4 mg/m3和0.978 6,預(yù)測精度高,,能夠為脫硝系統(tǒng)出口NOx的現(xiàn)場優(yōu)化控制提供技術(shù)支持,。
中圖分類號:X773
文獻標識碼:A
DOI:10.19358/j.issn.2097-1788.2023.09.013
引用格式:張瑾,姜浩,,金秀章.基于互信息變量選擇的燃煤機組SCR脫硝系統(tǒng)PSO-ELM建模[J].網(wǎng)絡(luò)安全與數(shù)據(jù)治理,,2023,42(9):88-95.
PSO-ELM modeling of SCR denitrification system of coal-fired units based on mutual information variable selection
Zhang Jin,,Jiang Hao ,,Jin Xiuzhang
( School of Control and Computer Engineering,North China Electric Power University,,Baoding 071003,,China)
Abstract: Aiming at the problems of NOx concentration at the outlet of selective catalytic reduction (SCR) denitration system of coal-fired units, such as measurement lag and data distortion during purging, a SCR denitration system model of coal-fired units based on feature extraction and particle swarm optimization (PSO) to optimize extreme learning machine (ELM) hyperparameters is proposed in this paper. Mutual information (MI) was used to compensate the time delay, maximum correlation minimum redundancy (mRMR) was used to screen the auxiliary variables, and the optimal ELM hyperparameters were determined by PSO optimization algorithm and the prediction model was established. Finally, the comparison and verification were carried out. The simulation results show that the mean square error and correlation coefficient of the PSO-ELM prediction model established by the method in this paper are 0.931 4 mg/m3 and 0.978 6 respectively, with high prediction accuracy, which can provide technical support for the on-site optimization control of NOx at the exit of the denitrification system.
Key words : mutual information;PSO algorithm,;SCR-DeNOx system,;extreme learning

0     引言

燃煤機組產(chǎn)生的氮氧化物(NOx)是大氣污染的首要排放物之一,在空氣質(zhì)量方面影響較為嚴重[1],。煙氣排放連續(xù)檢測系統(tǒng)(Continuous Emission Monitoring Systems,,CEMS)對煙氣取樣管路要按時反向吹掃,以避免積灰堵塞,,從而會導(dǎo)致NOx測量結(jié)果存在間斷性失真,,同時,由于煙氣取樣管路長度一般為40~60 m,,造成測量結(jié)果出現(xiàn)時滯現(xiàn)象,,控制系統(tǒng)的控制難度也因此得到提升,。因此,建立脫硝系統(tǒng)預(yù)測模型,,對于燃煤機組的優(yōu)化運行,,噴氨量的控制以及污染物的監(jiān)測管理都具有重要意義[2]。

隨著神經(jīng)網(wǎng)絡(luò)的發(fā)展,,許多建模方法被應(yīng)用到脫硝系統(tǒng)當(dāng)中,。楊文玉等人[3]利用RBF神經(jīng)網(wǎng)絡(luò)建立了脫硝系統(tǒng)出口NOx的預(yù)測模型,該模型在處理時序預(yù)測問題時并沒有明顯優(yōu)勢,。張淑清等人[4]利用ELM神經(jīng)網(wǎng)絡(luò)建立了電網(wǎng)負荷的預(yù)測模型,,并利用飛蛾優(yōu)化算法對模型參數(shù)進行優(yōu)化,該文所用訓(xùn)練數(shù)據(jù)過少,,容易導(dǎo)致模型過擬合,。劉延泉等人[5]將互信息與LSSVM方法結(jié)合,對脫硝系統(tǒng)入口NOx濃度進行了預(yù)測,,但模型未考慮輸入變量的對模型的影響,。

除了建模方法,特征選擇也會影響模型的預(yù)測能力,。特征選擇常見的方法有過濾式(Filter),、封裝式(Wrapper)和嵌入式(Embedded)三種。輸入變量的直接選擇決定了模型的結(jié)構(gòu)與輸出,,輸入變量的選擇通常對工業(yè)機理進行分析,,從待選變量進行篩選獲取[6-7],。金秀章等人[8]利用mRMR算法篩選出符合模型的輸入變量,,建立了出口SO2質(zhì)量濃度預(yù)測模型,但正則化仍不能計算出隱層節(jié)點的具體數(shù)量,。趙文杰等人[9]利用互信息與優(yōu)化算法結(jié)合確定系統(tǒng)最優(yōu)的輸入變量集合,,將互信息特征提取方法應(yīng)用到高維系統(tǒng)中,建立了脫硝系統(tǒng)的預(yù)測模型,,但該方法計算量大,,耗時較長,實施起來較為困難,。錢虹等人[10]采用隨機森林算法進行變量選擇,,并對SCR脫硝系統(tǒng)出口NOx質(zhì)量濃度進行了預(yù)測,但模型未解決煙氣采樣管道長度較長而導(dǎo)致的時滯問題,。


本文詳細內(nèi)容請下載:http://forexkbc.com/resource/share/2000005666




作者信息:

張瑾,,姜浩,金秀章

(華北電力大學(xué)控制與計算機工程學(xué)院,河北保定071003)

微信圖片_20210517164139.jpg

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