【摘要】：隨著當前以酸雨、霧霾為首的大氣污染問題日益嚴重,環(huán)境污染的治理逐漸成為能源生產(chǎn)的重要環(huán)節(jié)。SCR脫硝技術(shù)以其成熟、應用廣泛的特點成為我國燃煤電廠控制NOX污染的主要手段之一。深入研究SCR脫硝技術(shù),針對SCR脫硝系統(tǒng)變工況運行的特性,選用試驗建模的方法,選取能夠反應系統(tǒng)動態(tài)特性的輸入輸出數(shù)據(jù),建立有效的SCR脫硝系統(tǒng)模型,以此作為優(yōu)化控制過程的基礎(chǔ),是提高脫硝效率的重要途徑,對SCR脫硝裝置的設(shè)計和運行具有重要的參考價值。本文以火電廠變負荷運行常態(tài)為背景,綜述了SCR脫硝技術(shù)及應用,基于對模型泛化能力強,推廣性可靠,計算速度快,適于在線工作的考慮,選用了最小二乘支持向量機的方法建立非線性模型。針對變量之間的多重共線性,首先采用了相似度函數(shù)優(yōu)化處理方法,去除變量冗余數(shù)據(jù);其次為了更好的體現(xiàn)系統(tǒng)對象的動態(tài)特性,利用核主元分析和多變量過程監(jiān)測影響SCR反應器入口NOX濃度的主導因素,采用網(wǎng)格搜索結(jié)合粒子群(PSO)算法確定模型中參數(shù),建立了LSSVM預測模型。針對機組變負荷要求,本文結(jié)合滑動窗口法,預定義可根據(jù)鍋爐負荷的變化而自適應改變的更新閾值,分別是基于剔除最舊數(shù)據(jù)準則和基于預報誤差來自適應更新模型參數(shù),建立了兩種動態(tài)模型。分別將上述建模方法應用于SCR脫硝系統(tǒng)建模,進行了測試,對比結(jié)果顯示出了動態(tài)模型的預測精度較高、泛化能力強。最后將預測模型應用到預測控制算法中,轉(zhuǎn)化為非線性優(yōu)化問題,利用粒子群算法優(yōu)化噴氨量,使SCR反應器出口NOX濃度能夠很好地跟蹤設(shè)定值,從而提高脫硝效率。實驗結(jié)果表明相比傳統(tǒng)的PID控制方式,該方法實現(xiàn)了較好的控制效果。
[Abstract]:With the acid rain, the air pollution problem led by haze is becoming more and more serious, and the control of environmental pollution has gradually become an important link in the energy production. SCR denitrification technology is mature with its maturity. The widely used features have become one of the main means to control NOX pollution in coal-fired power plants in China. The technology of SCR denitrification is studied in depth. According to the characteristics of SCR denitrification system under different operating conditions, the experimental modeling method is selected, and the input and output data which can reflect the dynamic characteristics of the system are selected to establish an effective SCR denitrification system model. As the basis of optimization control process, it is an important way to improve denitrification efficiency and has important reference value for the design and operation of SCR denitrification plant. Based on the background of variable load operation normality of thermal power plant, this paper summarizes the technology and application of SCR denitrification, which is based on the consideration of strong generalization ability, reliable generalization, fast calculation speed and suitable for on-line operation. The least square support vector machine (LS-SVM) is used to establish the nonlinear model. Aiming at the multiple collinearity between variables, the similarity function optimization method is used to remove the redundant data. Secondly, in order to better reflect the dynamic characteristics of the system object, the kernel principal component analysis and multivariable process are used to monitor the main factors affecting the NOX concentration at the inlet of SCR reactor, and the parameters in the model are determined by grid search combined with particle swarm optimization (PSO) algorithm. LSSVM prediction model is established. According to the variable load requirement of the unit, this paper combines the sliding window method to predefine the renewal threshold which can be changed adaptively according to the change of boiler load, which is based on the elimination of the oldest data criterion and the prediction error based on the adaptive updating model parameters, respectively. Two dynamic models are established. The above modeling methods are applied to the modeling of SCR denitrification system respectively. The comparison results show that the dynamic model has high prediction accuracy and strong generalization ability. Finally, the predictive model is applied to the predictive control algorithm, which is transformed into a nonlinear optimization problem. The particle swarm optimization algorithm is used to optimize the amount of ammonia injection, so that the NOX concentration at the outlet of the SCR reactor can track the set value well, thus improving the denitrification efficiency. The experimental results show that this method has better control effect than the traditional PID control method.
【學位授予單位】：華北電力大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：X773

【參考文獻】

相關(guān)期刊論文 前10條

1 王f，

本文編號：2338428