在污水處理過(guò)程中,關(guān)鍵出水參數(shù)的實(shí)時(shí)測(cè)量對(duì)于出水水質(zhì)起著至關(guān)重要的作用。然而,受現(xiàn)存技術(shù)的限制,實(shí)際污水處理系統(tǒng)中存在很多難以在線測(cè)量的變量,生化需氧量（BOD₅）就是其中之一。軟測(cè)量技術(shù)的出現(xiàn)有效地解決了這類問(wèn)題。在軟測(cè)量建模中,傳統(tǒng)的機(jī)器學(xué)習(xí)方法得到了廣泛的應(yīng)用,但是,這類方法通常被認(rèn)為是具有一個(gè)隱藏層模型結(jié)構(gòu)的淺層學(xué)習(xí)方法。淺層學(xué)習(xí)對(duì)于簡(jiǎn)單的非線性過(guò)程逼近發(fā)揮了很好的作用,但當(dāng)面對(duì)高度復(fù)雜的過(guò)程時(shí),此類方法則顯得力不從心。鑒于此,本文針對(duì)污水處理廠可獲取的有限樣本、過(guò)程的高度非線性和動(dòng)態(tài)特性等問(wèn)題,基于深度學(xué)習(xí)研究了BOD₅的軟測(cè)量預(yù)測(cè)建模方法。主要工作包括:1)為了解決有限標(biāo)記樣本和變量間的嚴(yán)重非線性的問(wèn)題,考慮深度神經(jīng)網(wǎng)絡(luò)堆疊自動(dòng)編碼器（Stacked Autoencoders,SAE）在復(fù)雜非線性方面的強(qiáng)處理能力,以及遺傳算法（Genetic Algorithm,GA）在尋優(yōu)方面的良好性能,將二者結(jié)合提出了一種污水處理BOD₅在線監(jiān)測(cè)的SAE+GA軟測(cè)量預(yù)測(cè)建模方法。該方法首先根據(jù)實(shí)驗(yàn)數(shù)據(jù),選擇與BOD

【文章來(lái)源】：蘭州理工大學(xué)甘肅省

【文章頁(yè)數(shù)】：66 頁(yè)

【學(xué)位級(jí)別】：碩士

【文章目錄】：
摘要
Abstract
Chapter1 Introduction
    1.1 Research Background and Significance
    1.2 Soft Sensor Technologies and Applications in Wastewater System
        1.2.1 Research Status in World
        1.2.2 Research Status in China
    1.3 Overview of Soft Sensor Modeling Methods
        1.3.1 First Principle Models
        1.3.2 Regression Analysis Modeling Methods
        1.3.3 Artificial Intelligence Modeling Methods
    1.4 Thesis Outline
    1.5 Summary
Chapter2 Model Analysis of Wastewater Treatment Process
    2.1 Introduction
    2.2 Overview of the Wastewater Treatment Process
        2.2.2 Key Effluent Parameters Analysis of the WWTPs
        2.2.3 Limited Characteristics of Label Samples
        2.2.4 Process Nonlinearity
        2.2.5 Process Dynamic Characteristics
    2.3 Deep Learning Application of Soft Sensor in the WWTPs
        2.3.1 Stacked Autoencoders and its Parameter Optimization Algorithm
        2.3.2 Recurrent Neural Network and Long-Short Term Memory
        2.3.3 Model Structure Identification Using a Genetic Algorithm
    2.4 Summary
Chapter3 Soft Sensor Modeling of Key Effluent Parameter in Wastewater Treatment Process Based on SAE and GA
    3.1 Introduction
    3.2 Soft Sensor Modeling of Key Effluent Parameter BOD5 Based on SAE and GA
        3.2.1 Soft Sensor Modeling Based on SAE and GA
        3.2.2 Modeling Steps for Soft Sensor Based on SAE and GA
    3.3 Simulation Study
        3.3.1 Case Description
        3.3.2 Augmentation Processing and Data Preprocessing
        3.3.3 Setting Parameters of the Deep Neural Network
        3.3.4 Simulation Experiment and Result Analysis
    3.4 Summary
Chapter4 Dynamic Soft Sensor Modeling of Key Effluent Parameter in Wastewater Treatment Process Based on Recurrent Neural Network LSTM
    4.1 Introduction
    4.2 Dynamic Soft Sensor Modeling of Key Effluent Parameter BOD5 Based on LSTM
        4.2.1 Soft Sensor Modeling Based on LSTM and GA
        4.2.2 Modeling Steps for Soft Sensor Based on LSTM
    4.3 Simulation Study
        4.3.1 Data Preprocessing
        4.3.2 Hyperparameters Optimization
        4.3.3 Simulation Experiment and Result Analysis
    4.4 Summary
Conclusions and Future Work
    Conclusions
    Future Work
Reference
Acknowledgment
List of Publications



本文編號(hào)：3029518