基于ELMAN的化工污水生化系統(tǒng)建模及水質(zhì)預(yù)測(cè)

論文編號(hào):DQ268   論文字?jǐn)?shù):24690,頁(yè)數(shù):49 

摘　　要
 化工污水處理是污水處理的主要內(nèi)容，是影響與制約水環(huán)境改善的關(guān)鍵因素。其水質(zhì)控制技術(shù)也一直是水處理領(lǐng)域研究的重要課題。本文通過對(duì)國(guó)內(nèi)外污水的處理工藝、水質(zhì)監(jiān)測(cè)技術(shù)及處理過程自動(dòng)控制技術(shù)現(xiàn)狀的了解與研究，針對(duì)污水處理過程中關(guān)鍵水質(zhì)參數(shù)無法在線監(jiān)測(cè)等問題，提出人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)技術(shù)是未來污水處理測(cè)控領(lǐng)域的一個(gè)重要發(fā)展方向。
 鑒于污水處理過程的強(qiáng)非線性、大時(shí)變、嚴(yán)重滯后的特點(diǎn)，難以通過機(jī)理分析建立精確的數(shù)學(xué)模型，本文采用基于動(dòng)態(tài)神經(jīng)網(wǎng)絡(luò)技術(shù)的辨識(shí)建模方法。主要內(nèi)容有以下三方面：
 1、基于化工污水處理工程實(shí)際的數(shù)據(jù)預(yù)處理方法研究
 以來自于實(shí)際的化工污水處理廠的現(xiàn)場(chǎng)歷史數(shù)據(jù)，結(jié)合工藝實(shí)踐與數(shù)據(jù)處理技術(shù)，研究數(shù)據(jù)庫(kù)建立、誤差處理、數(shù)據(jù)插值與修復(fù)等數(shù)據(jù)預(yù)處理方法，為神經(jīng)網(wǎng)絡(luò)建模提供了良好的基礎(chǔ)數(shù)據(jù)。
 2、研究與建立污水處理出水水質(zhì)人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)模型
 建立基于Elman動(dòng)態(tài)神經(jīng)網(wǎng)絡(luò)的污水處理核心系統(tǒng)——生化系統(tǒng)的神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)模型，以此來預(yù)測(cè)污水出水水質(zhì)的氨氮和COD濃度。對(duì)比基本Elman算法，從算法和網(wǎng)絡(luò)結(jié)構(gòu)兩方面進(jìn)行研究、改進(jìn)。提出了改進(jìn)的Levenberg-Marquardt反向傳播算法和增加預(yù)測(cè)項(xiàng)的歷史輸出值的Elman結(jié)構(gòu)算法。
 3、污水水質(zhì)參數(shù)預(yù)測(cè)模型的仿真研究
 利用MATLAB神經(jīng)網(wǎng)絡(luò)工具箱進(jìn)行水質(zhì)參數(shù)預(yù)測(cè)模型的仿真研究，仿真研究結(jié)果表明：通過算法與結(jié)構(gòu)的改進(jìn)，建立的生化系統(tǒng)神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)模型在精度、訓(xùn)練時(shí)效上都有明顯提高，可以很好地預(yù)測(cè)出水的氨氮和COD濃度。
 總之，本文的研究工作將人工神經(jīng)網(wǎng)絡(luò)技術(shù)與化工污水處理工程實(shí)際相結(jié)合，從理論與仿真兩方面驗(yàn)證了神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)技術(shù)應(yīng)用于污水處理過程出水水質(zhì)參數(shù)預(yù)測(cè)的可行性，為進(jìn)一步實(shí)現(xiàn)出水水質(zhì)控制奠定了堅(jiān)實(shí)基礎(chǔ)。
 
 
 
Abstract
Chemical sewage treatment are the main contents of the sewage treatment, It is a key factor in the influence and constraints of water environmental improvement. Its water quality control has been an important subject of research in the field of water treatment technology. In this paper, through understanding and research for the sewage treatment process both at home and abroad, water quality monitoring technology and treating processes automatic control technical aspect, artificial neural network forecasting the future of wastewater treatment technology is an important field of measurement and control the direction of development, for key water quality parameters can not be the issue of online monitoring in sewage treatment process.
In view of the characteristics of strongly nonlinear, large time-varying, a serious lag in the sewage treatment process,it is difficult to establish establish accurate mathematical models through the mechanism analysis. This paper uses modeling methods to identify based on dynamic neural network technology. The main contents contain the following three aspects:In this paper, the main contents contains the following two aspects:
1、Chemical sewage treatment works based on the actual method of data preprocessing.
From the on-site historical data in the the actual chemical sewage treatment plant, Combination of technology and data processing technology practice,research databases, error handling, data interpolation and restoration, such as data pre-processing, It provides a good basis for data for neural network modeling.
2、Research and the establishment of sewage treatment effluent quality artificial neural network prediction model.
The core of the sewage treatment system based on Elman Dynamic Neural Network - biochemical system's neural network forecast model, forecasts the sewage water leakage water quality by the ammonia nitrogen and COD density. Contrasting the basic Elman algorithm, and conducting the research and improvement from the algorithm and the network architecture two aspects, It proposes the improving Levenberg-Marquardt antipropagation algorithm and the Elman structure algorithm of the increasing forecast item′s history value of exports.
3、Simulation research for sewage water quality parameter forecast model.
Working on the water quality parameter forecast model using the MATLAB neural network toolbox the simulation research，The simulation results indicates that: Through the algorithm and the structure improvement, the established biochemical system neural network forecast model has the distinct enhancement in the precision and the training effectiveness, which also can forecast the water leakage ammonia nitrogen and the COD density well.
In a word,this research work actually unifies the artificial neural networks technology and the chemical industry sewage treatment project,and it has confirmed the neural network forecasting technology to apply in the sewage treatment process water leakage water quality parameter forecast feasibility from the theory and the simulation, for further realizing the water leakage water quality control to lay the solid foundation.

目　　錄

摘　　要 1
Abstract 2
引　　言 1
1 綜述 1
1.1 國(guó)內(nèi)外污水處理工藝發(fā)展?fàn)顩r 1
1.2 污水水質(zhì)監(jiān)測(cè)技術(shù)的現(xiàn)狀及存在的問題 2
1.2.1 國(guó)內(nèi)外污水水質(zhì)監(jiān)測(cè)技術(shù)的現(xiàn)狀 2
1.2.2 污水水質(zhì)監(jiān)測(cè)中現(xiàn)存的問題 4
1.3 國(guó)內(nèi)外污水處理過程自動(dòng)控制的研究現(xiàn)狀[18-26] 5
2 人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)技術(shù)及方案論證 7
2.1 人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)技術(shù)及其實(shí)施 7
2.2 人工神經(jīng)網(wǎng)絡(luò)的原理 7
2.3 人工神經(jīng)網(wǎng)絡(luò)模型的方案論證 17
3 污水處理過程及Elman動(dòng)態(tài)人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)模型建立的基礎(chǔ) 19
3.1 某化工污水廠污水處理工藝流程簡(jiǎn)介 19
3.2 輸入輸出變量確定原則 21
3.3 現(xiàn)場(chǎng)數(shù)據(jù)采集與數(shù)據(jù)預(yù)處理 22
3.3.1 現(xiàn)場(chǎng)數(shù)據(jù)采集 22
3.3.2 數(shù)據(jù)預(yù)處理 22
4 研究與建立污水處理出水水質(zhì)人工神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)模型 25
4.1 污水水質(zhì)數(shù)據(jù)來源 25
4.2 確定輸入輸出變量 25
4.3 Elman神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)的設(shè)計(jì) 26
5污水水質(zhì)參數(shù)預(yù)測(cè)模型的仿真研究 31
5.1 改進(jìn)算法的Elman神經(jīng)網(wǎng)絡(luò)模型仿真 33
5.2 改進(jìn)結(jié)構(gòu)的Elman神經(jīng)網(wǎng)絡(luò)模型仿真 36
結(jié)　　論 40
參 考 文 獻(xiàn) 41
致　　謝 44