《電子技術(shù)應(yīng)用》
您所在的位置:首頁 > 模擬設(shè)計(jì) > 設(shè)計(jì)應(yīng)用 > 基于LSTM的濕法煙氣脫硫漿液pH值建模
基于LSTM的濕法煙氣脫硫漿液pH值建模
《信息技術(shù)與網(wǎng)絡(luò)安全》2020年第8期
金秀章,,景 昊
華北電力大學(xué) 控制與計(jì)算機(jī)工程學(xué)院,河北 保定071003
摘要: 針對(duì)燃煤電廠濕式石灰石-石膏濕法煙氣脫硫(WFGD)過程中漿液pH值測(cè)量時(shí)間長(zhǎng),,不利于WFGD作業(yè)的問題,,建立高精度的漿液pH值模型,。基于深度學(xué)習(xí)的框架,,利用長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(LSTM)算法對(duì)時(shí)間序列處理上的優(yōu)越性進(jìn)行建模,,該模型具有良好的精確度和泛化能力,。將燃煤機(jī)組實(shí)際運(yùn)行數(shù)據(jù)中與漿液pH值變化相關(guān)的變量作為模型的輔助變量,建立基于LSTM神經(jīng)網(wǎng)絡(luò)的漿液pH值預(yù)測(cè)模型,。對(duì)模型進(jìn)行仿真驗(yàn)證,,并分別與BP神經(jīng)網(wǎng)絡(luò)模型和最小二乘支持向量機(jī)(LSSVM)模型比較,結(jié)果表明LSTM神經(jīng)網(wǎng)絡(luò)模型的預(yù)測(cè)精度最高,,驗(yàn)證了LSTM神經(jīng)網(wǎng)絡(luò)在工業(yè)建模中的優(yōu)良性能,。
中圖分類號(hào): TK39;TP183
文獻(xiàn)標(biāo)識(shí)碼: A
DOI: 10.19358/j.issn.2096-5133.2020.08.012
引用格式: 金秀章,,景昊. 基于LSTM的濕法煙氣脫硫漿液pH值建模[J].信息技術(shù)與網(wǎng)絡(luò)安全,,2020,39(8):62-66.
Modeling of pH value of wet flue gas desulfurization slurry based on LSTM
Jin Xiuzhang,,Jing Hao
School of Control and Computer Engineering,,North China Electric Power University,Baoding 071003,,China
Abstract: Aiming at the problem that the measurement time of slurry pH in wet limestone-gypsum wet flue gas desulfurization(WFGD) process in coal-fired power plants is long, which is not conducive to WFGD operation, a high-precision slurry pH model was established.Therefore, based on the framework of deep learning, the long-term and short-term memory neural network(LSTM) algorithm was used to model the superiority of time series processing.The model has good accuracy and generalization ability.The slurry pH prediction model based on LSTM neural network was established by using variables related to slurry pH changes in actual operating data of coal-fired units as the auxiliary variables of the model.The model was simulated and verified,and compared with the BP neural network model and the least square support vector machine(LSSVM) model.The results show that the LSTM neural network model has the highest prediction accuracy, which verifies the excellent performance of the LSTM neural network in industrial modeling.
Key words : slurry pH prediction,;long and short-term memory network(LSTM);wet limestone-gypsum wet flue gas desulfurization(WFGD),;time series



          目前燃煤電廠的SO2排放量已經(jīng)超過了SO2排放總量的一半,,并且呈現(xiàn)逐年遞增的趨勢(shì)。我國(guó)先后頒布的《火電廠大氣污染物排放標(biāo)準(zhǔn)》和《煤電節(jié)能減排升級(jí)與改造行動(dòng)計(jì)劃(2014—2020年)》等一系列政策法規(guī),,明確指出火電廠的SO2排放濃度必須控制在35 mg/m3以下,。石灰石-石膏濕法煙氣脫硫技術(shù)(WFGD)是目前最有效的燃煤機(jī)組SO2控制技術(shù)之一。WFGD工藝中漿液pH值是決定煙氣脫硫效率的關(guān)鍵參數(shù),,因此pH值的測(cè)量需要迅速,、準(zhǔn)確。

         在WFGD現(xiàn)場(chǎng)測(cè)量時(shí)由于環(huán)境惡劣,,且pH值變化具有較大的慣性,,導(dǎo)致測(cè)量時(shí)長(zhǎng)較大,無法及時(shí)得到漿液pH值的準(zhǔn)確值,,對(duì)于脫硫作業(yè)十分不利。因此需要對(duì)漿液pH值進(jìn)行預(yù)測(cè),。

         pH值測(cè)量作為非線性系統(tǒng)一直是研究熱點(diǎn),。利用燃煤機(jī)組的運(yùn)行數(shù)據(jù),再結(jié)合機(jī)理分析,,采用實(shí)驗(yàn)建模的方法可以辨識(shí)出精確合理的系統(tǒng)模型,。文獻(xiàn)[5]和文獻(xiàn)[6]把神經(jīng)網(wǎng)絡(luò)等自適應(yīng)模糊系統(tǒng)用于pH中和過程。BP神經(jīng)網(wǎng)絡(luò),、RBF神經(jīng)網(wǎng)絡(luò),、Elman神經(jīng)網(wǎng)絡(luò)等方法是pH值建模的典型方法,但上述算法本身在時(shí)間序列的處理上并沒有突出的優(yōu)勢(shì),。

         隨著技術(shù)的進(jìn)步,深度學(xué)習(xí),、遞歸神經(jīng)網(wǎng)絡(luò),、卷積神經(jīng)網(wǎng)絡(luò)等也在pH值建模得到應(yīng)用。LSTM神經(jīng)網(wǎng)絡(luò),,注重?cái)?shù)據(jù)間的時(shí)間特性,,在大遲延時(shí)間序列預(yù)測(cè)中具有突出優(yōu)勢(shì)。LSTM神經(jīng)網(wǎng)絡(luò)的特點(diǎn)在于發(fā)現(xiàn)當(dāng)前時(shí)刻數(shù)據(jù)與之前數(shù)據(jù)間的聯(lián)系,,利用本身具有的記憶能力,,將之前數(shù)據(jù)的狀態(tài)進(jìn)行保存,同時(shí)根據(jù)保存的信息影響后續(xù)的預(yù)測(cè)值及變化趨勢(shì),。

         因此,,本文提出一種基于LSTM神經(jīng)網(wǎng)絡(luò)的pH值預(yù)測(cè)模型。以某600 MW機(jī)組為研究對(duì)象,,使用機(jī)組實(shí)際運(yùn)行數(shù)據(jù),,經(jīng)過機(jī)理和相關(guān)性分析,確定pH值模型的輔助變量,,建立高精度的pH值預(yù)測(cè)模型,。


本文詳細(xì)內(nèi)容請(qǐng)下載http://forexkbc.com/resource/share/2000003249

作者信息:

金秀章,景  昊

(華北電力大學(xué) 控制與計(jì)算機(jī)工程學(xué)院,,河北 保定071003)


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