矸石電廠煤泥輸送管道堵塞預測研究
發(fā)布時間：2018-09-03 13:53
【摘要】：煤泥作為煤炭廢料,矸石電廠將其作為燃料在循環(huán)流化床鍋爐內摻燒,通過鋪設管道進行煤泥輸送。煤泥輸送管道的堵塞問題一直是影響煤泥輸送系統(tǒng)生產效率和安全的重要因素之一。對堵塞進行故障預測和定位將能顯著提高煤泥輸送系統(tǒng)的運行性能,也是保障循環(huán)流化床鍋爐正常與安全運行的重要課題。本文研究了基于小波變換和極限學習機的煤泥輸送管道堵塞預測及定位方法,主要開展了以下研究工作:針對煤泥輸送系統(tǒng)的監(jiān)測數(shù)據(jù)易受到干擾影響的問題,分析現(xiàn)場監(jiān)測數(shù)據(jù)的特點,對監(jiān)測數(shù)據(jù)進行缺失數(shù)據(jù)補齊和數(shù)據(jù)去噪預處理。采用三次指數(shù)平滑法對數(shù)據(jù)進行補齊處理,采用小波變換法進行數(shù)據(jù)去噪,比較四種不同閾值方法的去噪效果,并將小波空域相關法用于信號去噪,取得了更好的去噪效果,實現(xiàn)了煤泥輸送管道監(jiān)測數(shù)據(jù)的預處理,并為后期的壓力分布建模、堵塞預測及定位奠定基礎。針對復雜管道煤泥輸送時阻力損失的問題,分析煤泥輸送管道的阻力損失機理及影響因素,確定阻力損失是剪切應力和摩擦阻力共同作用的結果;建立基于機理分析法的復雜管道(水平直管、傾斜管道、垂直管道)的壓力分布模型。對管道內的一段煤泥進行受力分析,建立力的平衡方程,并在非線性參數(shù)約束條件下對其進行求解,得出復雜管道壓力分布的數(shù)學模型,其服從復雜的指數(shù)關系,并在確定模型摩擦阻力系數(shù)中改進了常用的的計算方法;分析壓力分布的多變量影響因素,建立基于量子遺傳(QGA)BP神經網絡的QGA-BP的壓力分布模型。針對煤泥輸送管道堵塞預測問題,提出基于粒子群優(yōu)化核函數(shù)極限學習機(PSOKELM)的煤泥輸送管道堵塞預測方法,該方法在分析煤泥輸送管道堵塞預測機理的基礎上,確定堵塞預測的特征量,將支持向量機核函數(shù)引入極限學習機,并通過粒子群算法進行參數(shù)優(yōu)化。利用黃陵煤矸石熱電廠實際測試數(shù)據(jù)進行仿真實驗,并與粒子群算法優(yōu)化支持向量機(PSOSVM)預測模型和核函數(shù)極限學習機(KELM)預測模型進行比較,結果證明基于PSOKELM的預測模型在預測速度和準確性方面均優(yōu)于PSOSVM預測模型,在預測精度上優(yōu)于KELM預測模型。針對煤泥輸送管道堵塞定位問題,分析瞬態(tài)正負壓波法的定位原理,利用小波變換空域相關法對正負壓波信號去噪,提出基于小波包預處理經驗模態(tài)分解和小波變換模極大值法相結合(WPEMD-WTM)的正負壓波突變點檢測方法,確定正負壓波時間差,結合壓力波波速進行堵塞點定位,并通過仿真驗證了該堵塞定位方法的有效性和準確性。針對煤泥輸送管道堵塞故障的報警問題,研究堵塞故障異常分析方法。通過計算壓力預測值及預測區(qū)間,結合壓力樣本數(shù)據(jù)的統(tǒng)計特征,判斷壓力異常狀況,確定警示閾值,劃分警示等級,根據(jù)不同等級,采取不同的安全控制措施,并驗證了堵塞故障安全控制方法的合理性。本文提出的煤泥輸送管道壓力分布模型,可對管道設計和膏體泵選擇提供依據(jù);基于核函數(shù)極限學習機的堵塞預測模型和基于小波分析的堵塞定位方法,能對于煤泥輸送系統(tǒng)的堵塞故障問題提供新的安全控制決策和手段。
[Abstract]:Slime is a kind of coal waste, which is burned in a circulating fluidized bed boiler as fuel in a gangue power plant. Slime transportation is carried out by laying pipelines. The blockage of the pipelines has always been one of the important factors affecting the production efficiency and safety of the sludge transportation system. The operation performance of the feeding system is also an important issue to ensure the normal and safe operation of CFB boilers. This paper studies the prediction and location method of slurry pipeline blockage based on wavelet transform and extreme learning machine. The characteristics of field monitoring data are analyzed, and the missing data are complemented and the data are denoised by cubic exponential smoothing method and wavelet transform method. Good denoising effect has realized the pretreatment of monitoring data of slime transportation pipeline, and laid the foundation for modeling the pressure distribution in the later period, predicting and locating the blockage. The pressure distribution model of the complex pipeline (horizontal straight pipe, inclined pipe, vertical pipe) based on the mechanism analysis method is established. The force analysis of a section of slime in the pipeline is carried out, the force balance equation is established, and the mathematical model of the pressure distribution of the complex pipeline is obtained under the condition of nonlinear parameter constraints. The model obeys complex exponential relation and improves the commonly used calculation method in determining the friction coefficient of the model; analyzes the multi-variable influencing factors of pressure distribution, establishes the pressure distribution model of QGA-BP based on the quantum genetic algorithm (QGA) BP neural network. Kernel function extreme learning machine (PSOKELM) is used to predict the blockage of coal slurry pipeline. Based on the analysis of the blockage prediction mechanism of coal slurry pipeline, the characteristic quantity of blockage prediction is determined. The kernel function of support vector machine is introduced into the extreme learning machine, and the parameters are optimized by particle swarm optimization. The test data are simulated and compared with PSOSVM and KELM prediction models. The results show that PSOKELM prediction model is superior to PSOSVM prediction model in prediction speed and accuracy, and is superior to KELM prediction model in prediction accuracy. Based on the analysis of the location principle of the transient positive and negative pressure wave method and the denoising of the positive and negative pressure wave signal by the spatial correlation method of wavelet transform, a detection method of the catastrophe point of the positive and negative pressure wave based on wavelet packet pre-processing empirical mode decomposition and wavelet transform modulus maximum method (WPEMD-WTM) is proposed to determine the time of the positive and negative pressure wave. The validity and accuracy of this method are verified by simulation. Aiming at the alarming problem of the blocking fault of the coal slurry pipeline, the analysis method of abnormal blocking fault is studied. The prediction value and prediction interval are calculated, and the statistical characteristics of the pressure sample data are combined to judge the blocking fault. According to different levels, different safety control measures are adopted, and the rationality of the safety control method for blocking fault is verified. The pressure distribution model of the slurry pipeline proposed in this paper can provide a basis for pipeline design and paste pump selection; based on kernel function limit learning machine The blockage prediction model and the blockage location method based on wavelet analysis can provide a new safety control decision and means for the blockage fault of coal slurry transportation system.
【學位授予單位】：西安科技大學
【學位級別】：博士
【學位授予年份】：2014
【分類號】：TM621
，
本文編號：2220161
[Abstract]:Slime is a kind of coal waste, which is burned in a circulating fluidized bed boiler as fuel in a gangue power plant. Slime transportation is carried out by laying pipelines. The blockage of the pipelines has always been one of the important factors affecting the production efficiency and safety of the sludge transportation system. The operation performance of the feeding system is also an important issue to ensure the normal and safe operation of CFB boilers. This paper studies the prediction and location method of slurry pipeline blockage based on wavelet transform and extreme learning machine. The characteristics of field monitoring data are analyzed, and the missing data are complemented and the data are denoised by cubic exponential smoothing method and wavelet transform method. Good denoising effect has realized the pretreatment of monitoring data of slime transportation pipeline, and laid the foundation for modeling the pressure distribution in the later period, predicting and locating the blockage. The pressure distribution model of the complex pipeline (horizontal straight pipe, inclined pipe, vertical pipe) based on the mechanism analysis method is established. The force analysis of a section of slime in the pipeline is carried out, the force balance equation is established, and the mathematical model of the pressure distribution of the complex pipeline is obtained under the condition of nonlinear parameter constraints. The model obeys complex exponential relation and improves the commonly used calculation method in determining the friction coefficient of the model; analyzes the multi-variable influencing factors of pressure distribution, establishes the pressure distribution model of QGA-BP based on the quantum genetic algorithm (QGA) BP neural network. Kernel function extreme learning machine (PSOKELM) is used to predict the blockage of coal slurry pipeline. Based on the analysis of the blockage prediction mechanism of coal slurry pipeline, the characteristic quantity of blockage prediction is determined. The kernel function of support vector machine is introduced into the extreme learning machine, and the parameters are optimized by particle swarm optimization. The test data are simulated and compared with PSOSVM and KELM prediction models. The results show that PSOKELM prediction model is superior to PSOSVM prediction model in prediction speed and accuracy, and is superior to KELM prediction model in prediction accuracy. Based on the analysis of the location principle of the transient positive and negative pressure wave method and the denoising of the positive and negative pressure wave signal by the spatial correlation method of wavelet transform, a detection method of the catastrophe point of the positive and negative pressure wave based on wavelet packet pre-processing empirical mode decomposition and wavelet transform modulus maximum method (WPEMD-WTM) is proposed to determine the time of the positive and negative pressure wave. The validity and accuracy of this method are verified by simulation. Aiming at the alarming problem of the blocking fault of the coal slurry pipeline, the analysis method of abnormal blocking fault is studied. The prediction value and prediction interval are calculated, and the statistical characteristics of the pressure sample data are combined to judge the blocking fault. According to different levels, different safety control measures are adopted, and the rationality of the safety control method for blocking fault is verified. The pressure distribution model of the slurry pipeline proposed in this paper can provide a basis for pipeline design and paste pump selection; based on kernel function limit learning machine The blockage prediction model and the blockage location method based on wavelet analysis can provide a new safety control decision and means for the blockage fault of coal slurry transportation system.
【學位授予單位】：西安科技大學
【學位級別】：博士
【學位授予年份】：2014
【分類號】：TM621
，
本文編號：2220161
本文鏈接：http://sikaile.net/kejilunwen/dianlilw/2220161.html
