碳中和趨勢(shì)下數(shù)學(xué)模擬在污水處理系統(tǒng)中的發(fā)展與綜合應(yīng)用

陳治池,何 強(qiáng),蔡 然,羅華瑞,羅 南,宋忱馨,程 鴻*

碳中和趨勢(shì)下數(shù)學(xué)模擬在污水處理系統(tǒng)中的發(fā)展與綜合應(yīng)用

陳治池1,何 強(qiáng)1,蔡 然2,4,羅華瑞3,羅 南2,宋忱馨4,程 鴻1*

(1.重慶大學(xué)三峽庫(kù)區(qū)生態(tài)環(huán)境教育部重點(diǎn)實(shí)驗(yàn)室,重慶 400045；2.北京首創(chuàng)生態(tài)環(huán)保集團(tuán)股份有限公司,北京 100000；3.深圳市環(huán)水投資集團(tuán)有限公司,廣東 深圳 518031；4.四川水匯生態(tài)環(huán)境治理有限公司,四川 內(nèi)江 641000)

數(shù)學(xué)模擬技術(shù)在污水處理方面被廣泛應(yīng)用,為了系統(tǒng)總結(jié)相關(guān)技術(shù),本文回顧了污水處理系統(tǒng)中數(shù)學(xué)模擬技術(shù)的發(fā)展歷程;綜述了活性污泥模型(ASM)與機(jī)器學(xué)習(xí)(ML)在水質(zhì)預(yù)測(cè)及參數(shù)工況優(yōu)化領(lǐng)域中的應(yīng)用;重點(diǎn)探究了污水處理系統(tǒng)中溫室氣體排放模型,以及多目標(biāo)優(yōu)化模型在污水處理系統(tǒng)中溫室氣體排放(GHG)、出水質(zhì)量(EQI)和運(yùn)行成本(OCI)的權(quán)衡問(wèn)題;歸納了數(shù)學(xué)模擬技術(shù)在實(shí)現(xiàn)污水廠能量自給與資源回收的應(yīng)用發(fā)展.研究結(jié)果表明數(shù)學(xué)模擬技術(shù)能準(zhǔn)確預(yù)測(cè)出水水質(zhì)、快速優(yōu)化工藝參數(shù)、權(quán)衡溫室氣體排放、出水水質(zhì)與運(yùn)行成本之間的關(guān)系、以及提高資源回收效率等.因此，數(shù)值模擬技術(shù)可有效指導(dǎo)污水處理工藝的運(yùn)行優(yōu)化以及管理,為污水處理行業(yè)減污降碳協(xié)同增效提供技術(shù)支撐.

碳中和；活性污泥模型；機(jī)器學(xué)習(xí)；溫室氣體；多目標(biāo)優(yōu)化；資源回收

目前,傳統(tǒng)的污水處理廠主要依靠技術(shù)人員自身經(jīng)驗(yàn)來(lái)管理,過(guò)量曝氣和投藥的現(xiàn)象普遍存在,這不僅使污水處理廠的管理運(yùn)行效率變低,更造成能耗藥耗的浪費(fèi)[1],最終增大碳排放量.當(dāng)今碳減排形勢(shì)下,污水處理廠在提標(biāo)改造的過(guò)程中必須同時(shí)兼顧當(dāng)前技術(shù)水平下的水質(zhì)提升、運(yùn)行成本和環(huán)境影響.而污水處理中的數(shù)學(xué)模擬技術(shù)是基于生化反應(yīng)機(jī)理和大量運(yùn)行數(shù)據(jù),對(duì)污水處理系統(tǒng)的各項(xiàng)指標(biāo)進(jìn)行模擬預(yù)測(cè)及分析的一項(xiàng)技術(shù),已在污水處理廠的設(shè)計(jì)、提標(biāo)改造以及運(yùn)行優(yōu)化等方面得到廣泛應(yīng)用[2-4].

數(shù)學(xué)模擬技術(shù)正在逐步成為污水處理廠同步實(shí)現(xiàn)減污降碳協(xié)同增效的重要工具.國(guó)際水協(xié)自1987年以來(lái)陸續(xù)推出了4套ASM系列模型,而隨著計(jì)算機(jī)技術(shù)的不斷發(fā)展,一些基于ASM系列模型的污水處理仿真軟件也相繼被開(kāi)發(fā)出來(lái)[5-6].近年來(lái),機(jī)器學(xué)習(xí)在流域水環(huán)境以及城市給排水系統(tǒng)的水質(zhì)預(yù)測(cè)方面的應(yīng)用日趨多樣化[7].此外,多目標(biāo)優(yōu)化算法模型的出現(xiàn),為污水處理系統(tǒng)的技術(shù)、經(jīng)濟(jì)及環(huán)境指標(biāo)的量化權(quán)衡提供了便利[8].然而,目前尚缺乏數(shù)學(xué)模擬技術(shù)在污水處理方面應(yīng)用的系統(tǒng)總結(jié).

鑒于此,本文從活性污泥模型的發(fā)展到實(shí)際應(yīng)用的角度,梳理了數(shù)學(xué)模擬在實(shí)現(xiàn)污水處理工藝節(jié)能降耗及碳減排中發(fā)揮的作用和存在的問(wèn)題.系統(tǒng)總結(jié)了以出水質(zhì)量指標(biāo)與能耗經(jīng)濟(jì)指標(biāo)預(yù)測(cè)為核心的數(shù)學(xué)模擬技術(shù)在關(guān)鍵參數(shù)工況優(yōu)化中的應(yīng)用,對(duì)模擬軟件的應(yīng)用、機(jī)器學(xué)習(xí)等諸多方面進(jìn)行梳理,重點(diǎn)綜述了數(shù)學(xué)模擬技術(shù)在溫室氣體模擬的應(yīng)用,以及多目標(biāo)優(yōu)化思想在研究技術(shù)、環(huán)境和經(jīng)濟(jì)權(quán)衡問(wèn)題中的展現(xiàn)的創(chuàng)新潛力,并分析了應(yīng)用數(shù)學(xué)模擬技術(shù)對(duì)污水處理系統(tǒng)中資源回收進(jìn)行管理的意義.本研究旨在為碳中和背景下污水處理行業(yè)未來(lái)的發(fā)展與水質(zhì)、經(jīng)濟(jì)和碳減排目標(biāo)的實(shí)現(xiàn)提供理論依據(jù)和實(shí)踐基礎(chǔ).

1 污水處理系統(tǒng)中數(shù)學(xué)模擬技術(shù)

如圖1所示,數(shù)學(xué)模擬技術(shù)在污水廠的應(yīng)用方向主要包括水質(zhì)預(yù)測(cè)與管理以及技術(shù)優(yōu)化與運(yùn)行兩個(gè)方面.其中Biowin、GPS-X及WEST等主流模擬軟件是以逐步發(fā)展起來(lái)的ASM系列模型為基礎(chǔ),在模型過(guò)程及參數(shù)組分等方面進(jìn)行優(yōu)化和改進(jìn),利用構(gòu)建的仿真模型科學(xué)合理地對(duì)工藝進(jìn)行診斷和優(yōu)化.而利用人工神經(jīng)網(wǎng)絡(luò)、支持向量機(jī)等機(jī)器學(xué)習(xí)算法可以充分挖掘污水廠運(yùn)行數(shù)據(jù)價(jià)值[9],以滿足污水處理廠的智慧化管理需求.此外,多目標(biāo)優(yōu)化模型為權(quán)衡討論出水質(zhì)量(EQI)、運(yùn)行成本(OCI)及溫室氣體排放量(GHG)等指標(biāo)提供了理論基礎(chǔ)[10],而NSGA-Ⅱ等多目標(biāo)遺傳算法的發(fā)展對(duì)于污水處理系統(tǒng)中的多優(yōu)化問(wèn)題求解的穩(wěn)定性和效率具有重要意義.

圖1 數(shù)學(xué)模擬技術(shù)在污水處理中的發(fā)展及應(yīng)用路線

1.1 活性污泥模型的發(fā)展

活性污泥模型的發(fā)展可以追溯到米門(mén)方程和Monod模型的提出,自此之后,一系列描述污水處理過(guò)程中微生物生化反應(yīng)過(guò)程的模型被不斷地開(kāi)發(fā)和完善.此外,一些以ASM系列模型為內(nèi)核的模擬軟件也被相繼推出,為研究以大時(shí)滯、多變量及復(fù)雜非線性為特點(diǎn)的污水處理過(guò)程提供了便利,具體見(jiàn)表1.

表1 活性污泥相關(guān)模型與典型模擬軟件的發(fā)展

續(xù)表1

1.2 活性污泥模型的優(yōu)化與改進(jìn)

(1)進(jìn)水組分的測(cè)定與優(yōu)化

作為搭建活性污泥數(shù)學(xué)模型的第一步,進(jìn)水組分的分析尤為關(guān)鍵.研究表明,進(jìn)水組分與城市污水的性質(zhì)有關(guān),不同的城市排水管道系統(tǒng)、生物預(yù)處理方式、經(jīng)濟(jì)發(fā)展水平、生活習(xí)慣都會(huì)對(duì)其產(chǎn)生影響[27].由于有機(jī)物的模擬過(guò)程就是以COD的組成成分為基礎(chǔ),本文以ASM2D為例,其COD組分拆解見(jiàn)圖2.

COD組分的測(cè)定分析方法很多,包括OUR呼吸計(jì)量法,缺氧批量NUR法、混凝-過(guò)濾法、化學(xué)分析法等.不同的COD組分測(cè)定難度不一,可采用不同測(cè)定方法或者通過(guò)取經(jīng)驗(yàn)值確定.通過(guò)總結(jié)近年來(lái)國(guó)內(nèi)外學(xué)者開(kāi)展COD組分測(cè)定方法相關(guān)的研究,得到各成分測(cè)定方法如表2所示.

值得注意的是,進(jìn)水組分的測(cè)定(如I,A等)還未形成統(tǒng)一的測(cè)試方法;此外,某些組分(如X等)的測(cè)定過(guò)程也相對(duì)復(fù)雜;因此,盡快建立操作簡(jiǎn)便、標(biāo)準(zhǔn)化的組分測(cè)定方法體系顯得尤為重要.

(2)模型過(guò)程的優(yōu)化

雖然ASM系列的四套核心模型能較好地描述污水處理的過(guò)程,但在實(shí)際應(yīng)用方面,研究者們?yōu)榱颂岣吣Ｐ偷哪M精度,往往在其基礎(chǔ)上進(jìn)行改良以確保模型更貼合污水廠實(shí)際情況.目前,ASM系列模型的改進(jìn)大致可以分為模型的簡(jiǎn)化與模型過(guò)程的增加[35].

圖2 COD組分樹(shù)狀圖

對(duì)于模型的簡(jiǎn)化,是指通過(guò)分析和計(jì)算變量參數(shù)對(duì)系統(tǒng)的敏感性,優(yōu)化或分類參數(shù),減少模型的復(fù)雜程度來(lái)提高模型的實(shí)用性.在分類參數(shù)方面,孫培德等[36]在充分分析活性污泥系統(tǒng)中生物反應(yīng)機(jī)理后將系統(tǒng)中微生物劃分為8類菌群,充分考慮微生物間的相互作用建立了全耦合活性污泥模型.在工藝參數(shù)方面,Kim等[37]在ASM2的基礎(chǔ)上刪除了總懸浮固體等不直接參與生化反應(yīng)的模型組分,精準(zhǔn)模擬了SBR工藝中的脫氮除磷過(guò)程.此外,Bahar等[38]在ASM1基礎(chǔ)上僅保留4種組分,在ASM3的基礎(chǔ)上僅保留5種組分,分別簡(jiǎn)化其反應(yīng)過(guò)程用于模擬污泥消化過(guò)程.雖然模型簡(jiǎn)化可以在不損失模型模擬準(zhǔn)確度的前提有效降低工作量,但該簡(jiǎn)化過(guò)程是基于參數(shù)敏感性分析,因此需確保敏感性分析和參數(shù)篩選的準(zhǔn)確性.

表2 COD組分測(cè)定方法匯總

對(duì)于模型過(guò)程的增加,雖然ASM系列模型中的反應(yīng)過(guò)程已比較全面,但在實(shí)際運(yùn)用中還是會(huì)出現(xiàn)某些偏差[35],這時(shí)往往通過(guò)增加模型過(guò)程,通過(guò)拓展的方式來(lái)提高模型的精確度和適應(yīng)度.Ni等[39]基于ASM3模型應(yīng)用同步存儲(chǔ)和生長(zhǎng)的概念建立了兩步反硝化模型,同時(shí)利用實(shí)例對(duì)改進(jìn)模型進(jìn)行了評(píng)價(jià)和比較.此外,Man等[40]嘗試在ASM3模型中加入了亞硝酸鹽等組分,并增加了氨氧化菌和亞硝酸鹽氧化菌的生長(zhǎng)等過(guò)程,拓展后的新模型較準(zhǔn)確地模擬了實(shí)驗(yàn)室運(yùn)行的SBR反應(yīng)器.模型過(guò)程的增加無(wú)疑會(huì)增加模型預(yù)測(cè)的準(zhǔn)確性,但同時(shí)應(yīng)該兼顧項(xiàng)目工作量,應(yīng)在研究目的范圍內(nèi)權(quán)衡參數(shù)的數(shù)量,高效地提高模型的準(zhǔn)確度.

1.3 機(jī)器學(xué)習(xí)

20世紀(jì)50年代,Hebb提出了解釋學(xué)習(xí)過(guò)程中大腦神經(jīng)元所發(fā)生變化的赫布理論[41],象征著機(jī)器學(xué)習(xí)思想的萌芽,之后出現(xiàn)了大量的機(jī)器學(xué)習(xí)算法.機(jī)器學(xué)習(xí)的本質(zhì)[42]是找到一個(gè)目標(biāo)函數(shù),使其成為輸入變量到輸出變量之間的最佳映射:().通俗來(lái)講就是使計(jì)算機(jī)能夠去模擬人類的學(xué)習(xí)行為,在學(xué)習(xí)的過(guò)程中獲取經(jīng)驗(yàn)技能,不斷完善自我性能.

目前國(guó)內(nèi)外研究中水工程領(lǐng)域常用機(jī)器算法的流程及原理見(jiàn)圖3.線性回歸與多項(xiàng)式回歸都是基于最小二乘法的擬合模型,屬于監(jiān)督學(xué)習(xí)的一種,K-最近鄰算法適合類域交叉重疊較多的大數(shù)據(jù)[43];支持向量機(jī)在解決時(shí)間序列分析、分類問(wèn)題等問(wèn)題上表現(xiàn)優(yōu)異[44];隨機(jī)森林計(jì)算時(shí)由每棵樹(shù)投票或取均值的方式來(lái)決定最終結(jié)果[45];對(duì)于人工神經(jīng)網(wǎng)絡(luò)而言,感知機(jī)是它的結(jié)構(gòu)單元.為了能夠處理高維度的數(shù)據(jù)而提出了多層感知機(jī)[46],而徑向基函數(shù)神經(jīng)網(wǎng)絡(luò)由于其簡(jiǎn)單的拓?fù)浣Y(jié)構(gòu)和全局逼近能力,在非線性系統(tǒng)的控制建模上應(yīng)用廣泛[47].循環(huán)神經(jīng)網(wǎng)絡(luò)主要用于處理序列數(shù)據(jù)[48],長(zhǎng)短期記憶網(wǎng)絡(luò)是一種特殊的循環(huán)神經(jīng)網(wǎng)絡(luò),可以很好的解決循環(huán)神經(jīng)網(wǎng)絡(luò)存在的長(zhǎng)時(shí)依賴問(wèn)題[49].深度神經(jīng)網(wǎng)絡(luò)可以理解為有很多隱藏層的神經(jīng)網(wǎng)絡(luò),每一層構(gòu)成一個(gè)非線性信息處理單元[50],與傳統(tǒng)的機(jī)器學(xué)習(xí)方法相比具有更大的特征學(xué)習(xí)和特征表達(dá)能力[51].

機(jī)器學(xué)習(xí)可被應(yīng)用于解決生化工程、材料科學(xué)、環(huán)境及水科學(xué)問(wèn)題各個(gè)方面.例如,Mowbray等[52]回顧了過(guò)去20年來(lái)機(jī)器學(xué)習(xí)在生物化學(xué)工程中的應(yīng)用;Zhong等[53]總結(jié)了機(jī)器學(xué)習(xí)在環(huán)境科學(xué)與工程領(lǐng)域的四種主要應(yīng)用類型:預(yù)測(cè)、提取特征重要性、檢測(cè)異常值和對(duì)化學(xué)材料的探索;Huang等[7]全面總結(jié)了機(jī)器學(xué)習(xí)在自然和工程水系統(tǒng)中的應(yīng)用及發(fā)展.隨著機(jī)器學(xué)習(xí)的迅猛發(fā)展,它在評(píng)價(jià)和預(yù)測(cè)污水處理廠性能方面的應(yīng)用也愈加深入,為污水處理廠及時(shí)獲取水質(zhì)信息和高效運(yùn)行管理作出了貢獻(xiàn).

圖3 水工程領(lǐng)域常用機(jī)器學(xué)習(xí)算法的流程及原理

1.4 多目標(biāo)優(yōu)化

污水處理廠在提標(biāo)改造的過(guò)程中應(yīng)保證污水處理廠出水質(zhì)量達(dá)標(biāo)的同時(shí)控制運(yùn)行成本達(dá)到提質(zhì)降耗的目的,此時(shí)構(gòu)成了一個(gè)多目標(biāo)優(yōu)化問(wèn)題.多目標(biāo)優(yōu)化的定義是使多個(gè)目標(biāo)在給定區(qū)域確定一組代表最優(yōu)解并形成帕累托(pareto)邊界的點(diǎn)[54],即考慮個(gè)自變量和個(gè)目標(biāo)函數(shù)的多目標(biāo)優(yōu)化問(wèn)題(MOP):

①非支配解:多目標(biāo)優(yōu)化問(wèn)題并不存在一個(gè)最優(yōu)解,所有可能的解都稱為非支配解,也稱為pareto解.

②pareto最優(yōu)解:在無(wú)法在改進(jìn)任何目標(biāo)函數(shù)的同時(shí)不削弱至少一個(gè)其他目標(biāo)函數(shù).這種解稱作非支配解或pareto最優(yōu)解.

即pareto前沿是pareto最優(yōu)解的集合在目標(biāo)函數(shù)空間上的像.

基于傳統(tǒng)數(shù)學(xué)規(guī)化原理的多目標(biāo)優(yōu)化模型在實(shí)際工程優(yōu)化問(wèn)題應(yīng)用中經(jīng)常表現(xiàn)出一定的脆弱性.“演化算法”在解決這類優(yōu)化問(wèn)題中相比與傳統(tǒng)的優(yōu)化算法顯示出了其優(yōu)越的性能.目前,演化算法可以分為四個(gè)主要的研究領(lǐng)域:遺傳算法、遺傳規(guī)劃、演化規(guī)劃和演化策略[56].其中遺傳算法是迄今為止演化算法中應(yīng)用最多、比較成熟、廣為人知的算法[57],其本質(zhì)是一種求解問(wèn)題的高度并行性全局搜索算法.

圖4 決策空間和目標(biāo)空間之間的關(guān)系以及兩目標(biāo)優(yōu)化問(wèn)題的解定義

1985年Schaffer提出了向量評(píng)估算法[58](VEGA),但該算法在搜索空間非凸時(shí)無(wú)法求得帕累托最優(yōu)解.1999年Zitzler等提出了SPEA算法[59],該算法性能優(yōu)越,但算法過(guò)程比較復(fù)雜.1995年Srinivas等提出了非支配排序遺傳算法NSGA[60], 2002年Deb等提出了帶精英策略的非支配排序遺傳算法(NSGA-II)[61].NSGA-II的優(yōu)良性能包括求解的穩(wěn)定性、解的收斂性和均勻分布性等,因此被國(guó)內(nèi)外學(xué)者在求解多目標(biāo)優(yōu)化問(wèn)題時(shí)廣泛引用.總之針對(duì)多目標(biāo)優(yōu)化的遺傳算法面臨的問(wèn)題主要有兩個(gè):一是如何與實(shí)際問(wèn)題更有效的結(jié)合,二是算法的效率問(wèn)題,這也是研究者們未來(lái)主要的研究方向.

在“碳中和”的大背景下,越來(lái)越多的學(xué)者將溫室氣體(GHG)的排放納入了污水處理的性能評(píng)價(jià)范疇,并與出水質(zhì)量(EQI)、運(yùn)行成本(OCI)等傳統(tǒng)性能評(píng)價(jià)指標(biāo)共同構(gòu)成了污水處理過(guò)程的多目標(biāo)優(yōu)化問(wèn)題.即目標(biāo)函數(shù)為EQI、OCI和GHG,此時(shí)這類離散且非線性的多目標(biāo)優(yōu)化問(wèn)題[62]可被描述為:

式中:(決策變量)為DO,外回流比,內(nèi)回流比,排泥量,藥劑投加量等污水處理工藝運(yùn)行參數(shù).

針對(duì)不同的實(shí)際問(wèn)題,EQI的定義有著明顯差異,例如李宸等[63]將其定義為污染物的出水濃度與排放標(biāo)準(zhǔn)中各污染物限值的比值之和,而在另外一項(xiàng)研究[64]中對(duì)每種出水指標(biāo)賦予了權(quán)重,并定義了違規(guī)成本作為約束條件;對(duì)于OCI而言,一般基于Nopens等[65]提出的方法,即運(yùn)行成本定義為電能、化學(xué)藥劑成本與污泥處置成本之和;GHG則一般定義為污水處理系統(tǒng)直接產(chǎn)生的溫室氣體與電能藥耗間接產(chǎn)生的溫室氣體之和.當(dāng)然,嚴(yán)格意義上OCI與GHG的核算都應(yīng)該減去掉資源回收所產(chǎn)生的收益[66].

2 基于數(shù)學(xué)模擬的水質(zhì)指標(biāo)預(yù)測(cè)

2.1 活性污泥模型在水質(zhì)指標(biāo)模擬預(yù)測(cè)中的應(yīng)用

準(zhǔn)確的預(yù)測(cè)出水水質(zhì)有助于污水處理廠及時(shí)調(diào)整運(yùn)行策略[67],可利用基于活性污泥模型的仿真軟件對(duì)污水處理系統(tǒng)進(jìn)行模擬從而預(yù)測(cè)出水水質(zhì),目前相對(duì)誤差(REL)、均方誤差(RMSE)和西爾不等式系數(shù)(TIC)等指標(biāo)[68]常被用來(lái)評(píng)價(jià)模型擬合的準(zhǔn)確度.Vitanza等[69]利用Biowin軟件對(duì)三個(gè)污水廠進(jìn)行了工藝全流程建模,用相對(duì)誤差指標(biāo)分別評(píng)價(jià)了模型擬合的準(zhǔn)確度.

為提高活性污泥模型的適用性,新的水質(zhì)參數(shù)指標(biāo)不斷被開(kāi)發(fā)介入模型中.Hu等[70]將溶解性有機(jī)氮(mDON)的產(chǎn)生和消耗納入了傳統(tǒng)的ASM系列模型,并在全流程的實(shí)際污水處理廠驗(yàn)證了模型的適用性,為實(shí)現(xiàn)污水廠氮控制與成本的有效管理打下了基礎(chǔ).此外,數(shù)學(xué)模擬技術(shù)也逐步應(yīng)用于非傳統(tǒng)的污水處理工藝或反應(yīng)器.Eldyasti等[71]分別用BioWin和AQUIFAS對(duì)中試規(guī)模的循環(huán)流化床生物反應(yīng)器(CFBBR)進(jìn)行了數(shù)學(xué)模擬,發(fā)現(xiàn)BioWin和AQUIFAS都能夠預(yù)測(cè)大部分水質(zhì)參數(shù),平均百分比誤差在20%以下;Dorofeev A等[72]使用BioWin軟件對(duì)基于厭氧氨氧化工藝進(jìn)行了數(shù)學(xué)模擬,探究了溫度及溶解氧的最佳運(yùn)行條件.

盡管活性污泥模型針對(duì)不同水質(zhì)指標(biāo)或者工藝展現(xiàn)出較為準(zhǔn)確的預(yù)測(cè),但隨著污水處理工藝的發(fā)展,工藝類型變得愈加多樣與復(fù)雜,活性污泥模型在未來(lái)應(yīng)隨著污染物去除路徑與機(jī)理研究的深入,結(jié)合目標(biāo)工藝特征參數(shù)的挖掘,提高模型預(yù)測(cè)出水指標(biāo)的精度與泛化能力.

2.2 機(jī)器學(xué)習(xí)在水質(zhì)指標(biāo)預(yù)測(cè)中的應(yīng)用

由于傳統(tǒng)活性污泥模型在建模過(guò)程中,需要依托各水質(zhì)指標(biāo)參數(shù) (如進(jìn)水組分等),因而,水質(zhì)參數(shù)表征和的準(zhǔn)確性制約著模型模擬的效果.同時(shí),由于參數(shù)表征需耗費(fèi)大量時(shí)間和人力,活性污泥模型也展現(xiàn)出一定的低效性[73].隨著機(jī)器學(xué)習(xí)方法的發(fā)展,以數(shù)據(jù)為驅(qū)動(dòng)的預(yù)測(cè)模型通過(guò)挖掘污水處理廠“數(shù)據(jù)湖”的內(nèi)在規(guī)律,為模擬復(fù)雜的生化過(guò)程處理后的出水水質(zhì)指標(biāo)提供了另一個(gè)途徑[74].

各類機(jī)器學(xué)習(xí)算法已被應(yīng)用于污水處理廠水質(zhì)指標(biāo)的預(yù)測(cè),其中,人工神經(jīng)網(wǎng)絡(luò)(ANN) 已被證明是理解和模擬污水處理廠非線性行為的有效分析工具[75].例如,Hamed等[75]開(kāi)發(fā)了兩個(gè)基于人工神經(jīng)網(wǎng)絡(luò)的模型來(lái)預(yù)測(cè)污水處理廠的出水BOD和SS濃度,為污水廠的運(yùn)行提供了幫助.此外,支持向量機(jī)(SVM)[76]、模糊神經(jīng)網(wǎng)絡(luò)(ANFIS)[77]、極限學(xué)習(xí)機(jī)(ELM)[3]等算法也被用來(lái)預(yù)測(cè)污水處理系統(tǒng)的性能.

對(duì)算法進(jìn)行比較和創(chuàng)新可以節(jié)省大量時(shí)間成本并提高預(yù)測(cè)精度.Yang等[67]提出了一種自適應(yīng)動(dòng)態(tài)非線性偏最小二乘(PLS)模型,實(shí)驗(yàn)結(jié)果表明所提出的模型在預(yù)測(cè)精度、穩(wěn)定性和執(zhí)行效率方面體現(xiàn)了優(yōu)越性.對(duì)于出水氮指標(biāo)的預(yù)測(cè),Bagherzadeh等[78]提出了一種高精度的污水處理廠總氮預(yù)測(cè)模型,在此項(xiàng)研究中發(fā)現(xiàn)選擇合適的特征可提高20%預(yù)測(cè)精度,并且發(fā)現(xiàn)使用決策樹(shù)算法相比人工神經(jīng)網(wǎng)絡(luò)算法高出了10%的準(zhǔn)確度.而Alejo[79]等認(rèn)為支持向量機(jī)(SVM)在對(duì)于出水指標(biāo)中的氨氮預(yù)測(cè)方面是優(yōu)于人工神經(jīng)網(wǎng)絡(luò)(ANN) 的.同樣地,在預(yù)測(cè)凱氏氮的濃度方面,Manu等[80]認(rèn)為可以支持向量機(jī)(SVM)算法能夠通過(guò)有效地設(shè)置SVM參數(shù)提高預(yù)測(cè)效率,達(dá)到比模糊神經(jīng)網(wǎng)絡(luò)(ANFIS)更好的預(yù)測(cè)效果.此外,Picos等[81]還將人工神經(jīng)網(wǎng)絡(luò)(ANN)和遺傳算法(GA)相結(jié)合以尋找UASB反應(yīng)器的最佳工況,經(jīng)驗(yàn)證ANN-GA模型可以幫助UASB反應(yīng)器達(dá)到最佳工況,同時(shí)使含鹽廢水處理的能耗降到最低.

然而,上述研究?jī)H限于水質(zhì)指標(biāo)的直觀預(yù)測(cè),而忽略了微生物指標(biāo)在污水處理過(guò)程中的重要作用.活性污泥系統(tǒng)中的微生物狀態(tài)與出水質(zhì)量聯(lián)系非常緊密,因此在未來(lái)的研究中,微生物組學(xué)信息作為重要的特征信息與機(jī)器學(xué)習(xí)相結(jié)合也許會(huì)成為一個(gè)新的方向.

3 基于數(shù)學(xué)模擬的關(guān)鍵工藝參數(shù)工況控制及優(yōu)化

為了滿足日益嚴(yán)格的污水排放標(biāo)準(zhǔn),往往涉及到對(duì)生物處理單元的工況進(jìn)行優(yōu)化控制以充分發(fā)揮污水處理系統(tǒng)的效能.然而,活性污泥系統(tǒng)相當(dāng)復(fù)雜,依靠傳統(tǒng)方法對(duì)設(shè)計(jì)工況參數(shù)進(jìn)行優(yōu)化具有不確定性且低效性[82],因此數(shù)學(xué)模擬成為了污水處理系統(tǒng)設(shè)計(jì)升級(jí)和管理過(guò)程中不可或缺的工具.

3.1 出水質(zhì)量指標(biāo)優(yōu)化

國(guó)內(nèi)外不乏利用成熟的商用仿真軟件對(duì)參數(shù)進(jìn)行優(yōu)化的研究.魏忠慶等[83]利用GPS-X對(duì)福建某污水處理廠構(gòu)建了提標(biāo)工藝模型,在部分指標(biāo)達(dá)到準(zhǔn)Ⅳ類地表水水質(zhì)標(biāo)準(zhǔn)的同時(shí),優(yōu)化了工藝混合液回流比和污泥回流比,使工藝總體積降低了906m3.污水處理過(guò)程中的氮指標(biāo)優(yōu)化也相當(dāng)關(guān)鍵.針對(duì)改良AAO工藝運(yùn)行過(guò)程中存在的TN處理效果不穩(wěn)定問(wèn)題,周振等[84]利用WEST軟件模擬對(duì)比了A2/O,倒置A2/O以及AO三種工藝模式下的運(yùn)行效果,并通過(guò)優(yōu)化回流比的方式顯著改善了出水水質(zhì).污泥齡(SRT)是控制氮指標(biāo)的關(guān)鍵因素,Elawwad 等[85]使用BioWin為目標(biāo)污水處理廠建立了全流程模型,通過(guò)不同場(chǎng)景進(jìn)行模擬,發(fā)現(xiàn)在不改變污水處理廠現(xiàn)有規(guī)模和工藝的情況下,通過(guò)將SRT從2.7d增加到7d,同時(shí)延長(zhǎng)曝氣池的缺氧區(qū),可以輕松實(shí)現(xiàn)分別為94%和62.4%的硝化率和反硝化率.

為了比較改良型A2/O和Orbal氧化溝兩種工藝,邵袁等[17]利用WEST軟件對(duì)目標(biāo)污水處理廠分別構(gòu)建工藝模型,開(kāi)展工藝的優(yōu)化運(yùn)行研究,提出并實(shí)施了工藝的優(yōu)化運(yùn)行方案,有效提高了各類污染物的去除效率.鑒于溫度是影響污水廠運(yùn)行的重要因素,柳蒙蒙等[86]針對(duì)寒冷地區(qū)的城鎮(zhèn)污水處理廠,利用GPS-X分別對(duì)污泥回流比、反應(yīng)區(qū)體積比、充水比、運(yùn)行周期和不同水溫的CASS運(yùn)行方案等進(jìn)行了數(shù)值模擬優(yōu)化,冬季的運(yùn)行結(jié)果表明改造后的CASS工藝出水指標(biāo)能夠穩(wěn)定達(dá)到一級(jí)A排放標(biāo)準(zhǔn).

除了單獨(dú)應(yīng)用仿真軟件對(duì)污水處理系統(tǒng)進(jìn)行模擬優(yōu)化,也可以應(yīng)用試驗(yàn)設(shè)計(jì)思想耦合模型來(lái)指導(dǎo)污水處理廠運(yùn)營(yíng).例如,Cao等[68]篩選并分析了61個(gè)參數(shù)的靈敏度,對(duì)6個(gè)關(guān)鍵工藝參數(shù)建立二次多項(xiàng)式響應(yīng)面模型進(jìn)行優(yōu)化.結(jié)果表明,通過(guò)優(yōu)化生物單元Do和SRT顯著提高了工藝的脫氮效能.

3.2 基于數(shù)學(xué)模擬的能耗與運(yùn)行成本優(yōu)化

污水處理廠在污水的收集、處理和排放的過(guò)程中需要消耗大量能源[87],且能源的產(chǎn)生過(guò)程也是溫室氣體排放的間接來(lái)源之一.能源不僅用于污水處理系統(tǒng)(如反應(yīng)器、曝氣和泵裝置的電力消耗),還用于生產(chǎn)和運(yùn)輸處理過(guò)程中使用的化學(xué)藥劑[88].研究表明,污水處理過(guò)程中能耗最多可占污水處理廠運(yùn)行成本的48%[89],因此降低污水處理系統(tǒng)的能耗已經(jīng)成為污水廠實(shí)現(xiàn)可持續(xù)發(fā)展的最重要的主題之一,也是響應(yīng)國(guó)家碳減排號(hào)召的有效途徑.

為降低污水處理廠的平均能耗,首先需要分析影響污水處理系統(tǒng)能耗的關(guān)鍵因素.Bagherzadeh等[90]基于2014-2019年收集的數(shù)據(jù),研究了水質(zhì)、水力和氣候參數(shù)對(duì)目標(biāo)污水處理廠日平均能耗的影響,研究表明污水處理廠的能耗與溫度和濕度等天氣情況呈正相關(guān),并且總氮等指標(biāo)對(duì)目標(biāo)污水處理廠的能耗具有顯著影響.此外利用控制器對(duì)溶解氧和加藥量進(jìn)行優(yōu)化可為污水處理系統(tǒng)減少大量運(yùn)行成本,郝二成等[91-92]為探尋保證出水穩(wěn)定達(dá)標(biāo)的邊界運(yùn)行條件,建立模型對(duì)比了不同工藝控制方案,結(jié)果顯示按氨氮濃度變化設(shè)定Do值進(jìn)行曝氣控制、優(yōu)化除磷藥劑的投加濃度的方式實(shí)現(xiàn)了節(jié)能降耗.

數(shù)學(xué)模擬技術(shù)開(kāi)始被逐漸應(yīng)用在不同類型的污水進(jìn)水類型中.邵袁等[93]基于WEST建立了鄉(xiāng)鎮(zhèn)生活污水處理廠A2/O模型,通過(guò)軟件模擬優(yōu)化了工藝的回流比等參數(shù),使除磷藥劑月均使用量較優(yōu)化前下降56%,全廠用電量由7.1(GW·h)/月降至5.5 (GW·h)/月.數(shù)學(xué)模擬技術(shù)還可對(duì)工業(yè)廢水進(jìn)行優(yōu)化,Wu等[94]利用WEST軟件對(duì)進(jìn)水成分復(fù)雜的工業(yè)廢水污水廠進(jìn)行模擬優(yōu)化,使運(yùn)行成本從6.2歐元/m3降至5.5歐元/m3.

曝氣過(guò)程能耗超過(guò)污水處理廠總能耗的50%[95-97],因此利用數(shù)學(xué)模型對(duì)曝氣過(guò)程進(jìn)行模擬優(yōu)化可節(jié)省大量運(yùn)行成本.Lizarralde等[98]建立了污水處理廠曝氣過(guò)程的數(shù)學(xué)模型,用于研究曝氣過(guò)程的效率和詳細(xì)描述不同水相和氣相之間的傳質(zhì)過(guò)程;Asteriadis等[99]通過(guò)使用溶解氧級(jí)聯(lián)控制器對(duì)IAFB反應(yīng)器中的曝氣進(jìn)行模擬,找到了最大氧轉(zhuǎn)移速率水平對(duì)應(yīng)的溶解氧水平條件,避免了系統(tǒng)過(guò)度曝氣.數(shù)據(jù)與模型的集成系統(tǒng)是實(shí)現(xiàn)污水處理廠自動(dòng)化控制的基礎(chǔ),Sean等[100]使用GPS-X 進(jìn)行穩(wěn)態(tài)建模,開(kāi)發(fā)了帶監(jiān)督控制的數(shù)據(jù)采集(SCADA)系統(tǒng),發(fā)現(xiàn)使用GPS-X模擬優(yōu)化可節(jié)省20%的能耗.此外,Holenda等[101]結(jié)合遺傳算法(GA)討論了間歇曝氣污水處理廠的曝氣優(yōu)化問(wèn)題,結(jié)果發(fā)現(xiàn)與傳統(tǒng)的控制策略相比,遺傳算法的模型優(yōu)化策略可使能耗削減10%.不同的運(yùn)行方式也能造成不同的成本效益.Abbasi等[102]對(duì)常規(guī)活性污泥法、接觸穩(wěn)定化和分步曝氣三種方案分別進(jìn)行了模擬,使用GPS-X軟件計(jì)算了維修成本和能耗成本,發(fā)現(xiàn)進(jìn)行接觸穩(wěn)定化式運(yùn)行具有更高的成本效益.然而,出水水質(zhì)與能耗削減目標(biāo)往往互相矛盾,因此多目標(biāo)優(yōu)化模型的應(yīng)用可以為我們選擇最適運(yùn)行工況提供建議,以平衡出水水質(zhì)與能耗成本.

要想建設(shè)智慧化的污水廠,僅靠數(shù)學(xué)模擬工具對(duì)運(yùn)行成本進(jìn)行優(yōu)化是遠(yuǎn)遠(yuǎn)不夠的,面對(duì)水質(zhì)水量的沖擊變化,為了保證時(shí)效性,只有對(duì)數(shù)據(jù)進(jìn)行全面的在線監(jiān)測(cè)并建立數(shù)據(jù)庫(kù),才能對(duì)污水廠的運(yùn)行管理進(jìn)行及時(shí)、高效、準(zhǔn)確的指導(dǎo).

4 基于數(shù)學(xué)模擬的溫室氣體模擬及多目標(biāo)優(yōu)化研究

與發(fā)達(dá)國(guó)家相比,中國(guó)實(shí)現(xiàn)“雙碳”目標(biāo)時(shí)間更緊、幅度更大.水處理行業(yè)排放的溫室氣體約占人類活動(dòng)產(chǎn)生的溫室氣體的3%[103].研究發(fā)現(xiàn),N2O和CH4兩種強(qiáng)效溫室氣體的溫室效應(yīng)分別是CO2的298倍和25倍[104],而污水處理廠作為水處理行業(yè)溫室氣體排放的主要來(lái)源,對(duì)其產(chǎn)生的CO2、N2O及CH4等溫室氣體建立模型進(jìn)行模擬估算具有重大意義.

4.1 污水處理系統(tǒng)中溫室氣體排放模型

一般來(lái)說(shuō),污水處理廠主要通過(guò)三種機(jī)制(圖5)排放溫室氣體,即直接排放、間接內(nèi)部排放和間接外部排放[105].污水處理廠的直接排放主要與生物過(guò)程(如微生物呼吸產(chǎn)生的CO2排放、硝化和反硝化產(chǎn)生的N2O排放以及厭氧消化產(chǎn)生的CH4排放)有關(guān);間接內(nèi)部排放與污水廠內(nèi)部的電能或熱能的消耗有關(guān);間接外部排放與污水處理廠內(nèi)與外界的聯(lián)系(如污水廠所需化學(xué)藥劑的生產(chǎn)及各類運(yùn)輸產(chǎn)生的溫室氣體)有關(guān),它們的溫室氣體排放因子有著明顯差異(表3).

對(duì)于溫室氣體排放模型的研究,Ni等[106]基于已知的異養(yǎng)反硝化菌和氨氧化細(xì)菌(AOB)產(chǎn)生N2O代謝途徑開(kāi)發(fā)了N2O排放的數(shù)學(xué)模型并首次針對(duì)大型污水廠處理廠進(jìn)行模擬.溶解氧水平是影響N2O的產(chǎn)生的重要因素,Boiocchi等[107]基于BSM2仿真平臺(tái)對(duì)污水處理系統(tǒng)的溶解氧水平進(jìn)行了在線控制,最大限度地減少了N2O的排放.此外,Huang等[108]建立了針對(duì)污水廠內(nèi)不同排放源的溫室氣體估算模型,并且評(píng)估了溶解氧控制來(lái)減少溫室氣體排放的潛力.

圖5 典型污水廠工藝流程溫室氣體排放主要來(lái)源分布

表3 部分污水廠溫室氣體直接/間接排放因子

盡管近年來(lái)已有不少關(guān)于溫室氣體模型的研究,但由于不同類型污水處理工藝中N2O形成排放機(jī)制的尚未厘清,且難以收集到足夠的溫室氣體數(shù)據(jù)支撐模型,導(dǎo)致如今的溫室氣體模型過(guò)于參數(shù)化.

4.2 全流程溫室氣體的評(píng)估

在評(píng)估污水廠全流程溫室氣體排放過(guò)程方面,Mannina等[117]對(duì)研究污水處理過(guò)程中溫室氣體的建模工具進(jìn)行了總結(jié),提出了污水廠全流程建?？梢詭椭鬯畯S減少溫室氣體排放的理念.Kyung等[118]則開(kāi)發(fā)了一個(gè)綜合模型來(lái)估算污水處理廠的直接和間接溫室氣體排放量,并使用敏感性分析研究了溫室氣體排放的潛在變化.此外,Koutsou等[119]首次在歐洲國(guó)家范圍內(nèi)估算了希臘污水處理廠的溫室氣體排放量,提出了確保小型污水處理廠的高效運(yùn)行,促進(jìn)資源化利用等建議.這些研究無(wú)一例外的強(qiáng)調(diào)了對(duì)污水廠進(jìn)行溫室氣體排放控制的重要性,然而仍缺少環(huán)境經(jīng)濟(jì)目標(biāo)下溫室氣體模型結(jié)合具體控制方法的系統(tǒng)總結(jié),這對(duì)于可持續(xù)化推進(jìn)污水處理廠碳減排具有重要的意義.

4.3 多目標(biāo)優(yōu)化算法的應(yīng)用與發(fā)展

基于遺傳算法的多目標(biāo)優(yōu)化模型應(yīng)用于污水處理系統(tǒng)中已取得不少研究成果.陳文亮等[120]構(gòu)建了多目標(biāo)優(yōu)化模型,探究了污水處理廠出水水質(zhì)與運(yùn)行成本、反應(yīng)池體積之間的權(quán)衡問(wèn)題.而為了解決出水水質(zhì)與能耗之間的權(quán)衡問(wèn)題,周紅標(biāo)等[121]以能耗與出水質(zhì)量為目標(biāo),提出了混合多目標(biāo)骨干粒子群優(yōu)化算法,實(shí)現(xiàn)了溶解氧和硝態(tài)氮設(shè)定值的動(dòng)態(tài)尋優(yōu)、智能決策和底層跟蹤控制.經(jīng)過(guò)不斷的優(yōu)化創(chuàng)新,如今基于遺傳算法的多目標(biāo)優(yōu)化在算法層面更加高效,與實(shí)際問(wèn)題的結(jié)合也更加緊密.

4.4 碳中和趨勢(shì)下的多目標(biāo)優(yōu)化權(quán)衡

在多目標(biāo)優(yōu)化算法幫助下,可對(duì)溫室氣體(GHG)排放與出水質(zhì)量(EQI)、運(yùn)行成本(OCI)之間進(jìn)行權(quán)衡討論.首先,研究工藝參數(shù)變量的敏感性非常重要,Flores等[122]在評(píng)估污水處理廠的控制/運(yùn)營(yíng)策略時(shí),把溫室氣體排放的納入并制定了評(píng)價(jià)體系,研究了以下4個(gè)變量對(duì)EQI、OCI和GHG的影響:(1)曝氣池溶解氧水平;(2)初沉池對(duì)總懸浮物的去除率;(3)厭氧消化池的溫度;(4)厭氧消化上清液的流量.然后,Sweetapple等[123]對(duì)EQI、OCI及GHG進(jìn)行了權(quán)衡,研究表明有多種方案可在減少溫室氣體排放的同時(shí)不產(chǎn)生額外的運(yùn)營(yíng)成本并保持一定的出水質(zhì)量.為了對(duì)污水處理系統(tǒng)整體性能進(jìn)行優(yōu)化,Kim D等[64]在六種優(yōu)化方案下的仿真結(jié)果表明,優(yōu)化后的系統(tǒng)同時(shí)減少了31%的溫室氣體排放,降低了11%的運(yùn)行成本,改善了2%的出水質(zhì)量.Borzooei S等[2]同時(shí)考慮了污水處理廠的污水處理和污泥處置模型,通過(guò)改變不同的操作參數(shù)在提升出水水質(zhì)的同時(shí)節(jié)省了高達(dá)5000MWh的能耗,并且發(fā)現(xiàn)引入高級(jí)濃縮和污泥預(yù)處理過(guò)程對(duì)電廠的能量平衡和溫室氣體平衡有積極的影響.

以上研究表明應(yīng)用多目標(biāo)優(yōu)化可以在不影響污水處理系統(tǒng)整體性能的基礎(chǔ)上同步實(shí)現(xiàn)環(huán)境與經(jīng)濟(jì)效益的提升,但目前的研究仍然局限于狹義上的污水處理系統(tǒng),對(duì)于污水處理系統(tǒng)的前端(污水管網(wǎng))和后端(受納水體)中溫室氣體排放對(duì)環(huán)境的整體影響研究還鮮有涉及.

污水處理行業(yè)面臨著減污降碳協(xié)同增效的任務(wù),針對(duì)此問(wèn)題我們應(yīng)該重點(diǎn)分析目標(biāo)污水處理系統(tǒng)中污染物去除與溫室氣體排放的關(guān)聯(lián)機(jī)制,給出協(xié)同核算的具體步驟,集成多目標(biāo)優(yōu)化進(jìn)行問(wèn)題求解.對(duì)于污染物去除協(xié)同控制溫室氣體的核算邊界、協(xié)同機(jī)制和核算方法,具體可參考付加鋒等[124]建立的方法,包括了協(xié)同控制效應(yīng)和協(xié)同程度的計(jì)算公式;另一方面,污泥處置及資源化過(guò)程也是溫室氣體核算的難點(diǎn),要針對(duì)不同的污泥處置路徑,根據(jù)物料平衡對(duì)全生命周期內(nèi)的污泥轉(zhuǎn)化為能源的碳足跡進(jìn)行分析.從環(huán)境效益來(lái)看,我國(guó)最常見(jiàn)的垃圾處理方式即衛(wèi)生填埋,在沒(méi)有對(duì)產(chǎn)生的甲烷進(jìn)行深度處理的情況下,碳排放量達(dá)到最大值[125].因此進(jìn)行多目標(biāo)優(yōu)化的過(guò)程中,我們應(yīng)該評(píng)估調(diào)整污泥處置技術(shù)路線對(duì)溫室氣體及能量平衡的影響,并且以熱源溫度和藥劑使用等方面作為決策變量研究碳排放的響應(yīng)機(jī)制.

5 基于數(shù)學(xué)模擬的資源回收研究

傳統(tǒng)意義上的污水處理廠運(yùn)行主要目的是為了保證一定的出水質(zhì)量,因此研究重點(diǎn)往往放在有機(jī)物和氮磷的去除上.然而隨著如今環(huán)境資源的日益緊缺,研究模式應(yīng)從污染物去除逐步轉(zhuǎn)變?yōu)橘Y源回收,從數(shù)學(xué)模型的角度來(lái)說(shuō),這意味著我們污水廠內(nèi)的線性模型將轉(zhuǎn)變?yōu)楦友h(huán)的模型[126].

5.1 數(shù)學(xué)模擬在能源回收中的應(yīng)用

對(duì)于能量回收來(lái)說(shuō),甲烷作為污水廠內(nèi)厭氧消化產(chǎn)生的回收的有機(jī)質(zhì)能源,可以為污水廠提供約25%到50%的能源需求[127-128],該部分能源進(jìn)行回收可抵消部分污水處理能源消耗,進(jìn)而減少碳排放[129].因此一些關(guān)于厭氧消化的模型被開(kāi)發(fā)出來(lái),Gracia等[130]開(kāi)發(fā)了一種模擬好氧和厭氧操作條件下污泥消化反應(yīng)器的數(shù)學(xué)模型,并成功在中溫和高溫厭氧消化以及自熱高溫好氧消化的場(chǎng)景下得以驗(yàn)證. Ruffino等[131]分別對(duì)44L和240L的反應(yīng)器下進(jìn)行了中溫消化和高溫消化之間的模擬比較,發(fā)現(xiàn)在有效的熱交換前提下,污水廠的高溫消化過(guò)程可以回收100～200Nm3/h的天然氣.此外,Sakiewicz等[132]使用人工神經(jīng)網(wǎng)絡(luò)對(duì)污水處理廠中的厭氧消化工藝進(jìn)行了模擬,經(jīng)參數(shù)敏感性分析表明,與水質(zhì)指標(biāo)相比,操作工藝參數(shù)對(duì)甲烷產(chǎn)量的重要性更高.以上研究都表明,利用數(shù)學(xué)模擬技術(shù)可探尋合適的工況條件使污水廠實(shí)現(xiàn)能源自給.

5.2 數(shù)學(xué)模擬在養(yǎng)分回收中的應(yīng)用

從污水中回收養(yǎng)分一直是研究的熱點(diǎn).其中磷作為引起自然水體的富營(yíng)養(yǎng)化問(wèn)題的主要因素而被廣泛關(guān)注,然而如今由于磷元素在自然界的短缺,被認(rèn)為是一種待回收的物質(zhì).Shu等[133]認(rèn)為通過(guò)沉淀的方式將磷轉(zhuǎn)化為鳥(niǎo)糞石具有可觀的經(jīng)濟(jì)價(jià)值.研究發(fā)現(xiàn)在厭氧消化過(guò)程引起的鳥(niǎo)糞石自發(fā)沉淀會(huì)導(dǎo)致污水廠管道堵塞[134],因此,鳥(niǎo)糞石的回收也極具工程意義.Kazadi等[135]基于BSM2平臺(tái)建立了污水廠全流程模型,研究發(fā)現(xiàn)礦物質(zhì)沉淀強(qiáng)烈影響厭氧消化池中的成分和整個(gè)廠內(nèi)的污染負(fù)荷.通過(guò)鳥(niǎo)糞石結(jié)晶進(jìn)行磷回收方案的模擬,降低了43%出水磷指標(biāo)濃度.為了進(jìn)行最佳磷管理策略的研究,Lizarralde等[136]基于WEST軟件模擬目標(biāo)污水廠進(jìn)行了全流程建模,分析了不同的運(yùn)行方案得出了以下結(jié)論:(1)低SRT可以提高鳥(niǎo)糞石的回收率(2)曝氣池中固體濃度對(duì)于鳥(niǎo)糞石回收的最佳值隨溫度的變化而變化,且在溫度在13～19℃范圍內(nèi),固體濃度為2000g/m3時(shí)最利于鳥(niǎo)糞石的回收,而在較高溫度下,鳥(niǎo)糞石回收率在較低固體濃度下達(dá)到最大.如今不乏一些從污水中回收養(yǎng)分的新興技術(shù),但大部分方法從本質(zhì)上來(lái)說(shuō)還是一種“以能換能”的形式;對(duì)于現(xiàn)階段的資源回收,借助數(shù)學(xué)模擬工具,厘清污水廠內(nèi)部可利用的資源分布,再通過(guò)調(diào)節(jié)工況而帶來(lái)的經(jīng)濟(jì)與環(huán)境效益往往是出乎意料的.

目前污水與資源的關(guān)系仍然是污水處理領(lǐng)域內(nèi)的新興問(wèn)題,數(shù)學(xué)模擬技術(shù)有助于預(yù)測(cè)不同場(chǎng)景下污水處理系統(tǒng)的能量及資源回收潛力,然而關(guān)于能源回用的評(píng)估計(jì)算涉及到諸多經(jīng)驗(yàn)參數(shù),要想提高模型預(yù)測(cè)的準(zhǔn)確度需要進(jìn)行參數(shù)的調(diào)研及修正.如今的能源和資源回用研究主要聚焦于甲烷及磷元素的回收,建議未來(lái)利用數(shù)學(xué)模型合理開(kāi)發(fā)污水處理系統(tǒng)的其他資源潛力(如氮元素、硫元素、熱能等),最大程度實(shí)現(xiàn)污水廠能源自給,最大限度的減少溫室氣體排放,為碳減排作出貢獻(xiàn).

6 結(jié)論與展望

6.1 結(jié)論

通過(guò)回顧數(shù)學(xué)模擬技術(shù)的發(fā)展和數(shù)學(xué)模擬在污水處理系統(tǒng)中的研究與應(yīng)用,展示并討論了數(shù)學(xué)模擬技術(shù)在解決污水處理系統(tǒng)中水質(zhì)預(yù)測(cè)、工藝參數(shù)控制優(yōu)化、溫室氣體排放相關(guān)的多目標(biāo)優(yōu)化問(wèn)題、以及資源回收等方面問(wèn)題獨(dú)特的優(yōu)越性和存在的問(wèn)題.

①水質(zhì)指標(biāo)的預(yù)測(cè)方面主要通過(guò)以活性污泥系列模型和機(jī)器學(xué)習(xí)方法為載體,對(duì)氮磷等相關(guān)出水指標(biāo)進(jìn)行預(yù)測(cè),研究者們往往致力于對(duì)模型過(guò)程進(jìn)行增減或者是對(duì)機(jī)器學(xué)習(xí)算法進(jìn)行優(yōu)化改進(jìn),來(lái)追求更高的預(yù)測(cè)準(zhǔn)確度和效率.

②數(shù)學(xué)模擬對(duì)污水廠出水質(zhì)量及能耗控制的關(guān)鍵工藝參數(shù)為內(nèi)回流比、外回流比、曝氣池溶解氧和污泥齡等,研究表明對(duì)曝氣系統(tǒng)建立模型進(jìn)行模擬研究尤其關(guān)鍵.此外數(shù)學(xué)模擬技術(shù)對(duì)于不同的處理污水的類型(城市污水、鄉(xiāng)鎮(zhèn)污水和工業(yè)廢水)都有技術(shù)和經(jīng)濟(jì)上的優(yōu)化案例.

③數(shù)學(xué)模擬結(jié)合溫室氣體模型可幫助研究者們制定控制策略來(lái)降低生產(chǎn)過(guò)程對(duì)環(huán)境的影響.隨著多目標(biāo)優(yōu)化模型算法的研究發(fā)展,Pareto最優(yōu)解集的收斂性與多樣性得到提高,在溫室氣體排放(GHG)、出水質(zhì)量(EQI)與運(yùn)行成本(OCI)之間的權(quán)衡問(wèn)題上可進(jìn)行大量研究.利用數(shù)學(xué)模擬工具可在不同場(chǎng)景對(duì)污水處理系統(tǒng)中溫室氣體減排的潛力進(jìn)行量化,從而做到減污降碳協(xié)同增效.

④資源回收會(huì)對(duì)污水處理系統(tǒng)產(chǎn)生環(huán)境和經(jīng)濟(jì)效益產(chǎn)生積極的影響.對(duì)于能量回收方面,目前數(shù)學(xué)模擬技術(shù)集中對(duì)厭氧消化模型進(jìn)行了研究,利用模型對(duì)消化過(guò)程產(chǎn)生的甲烷產(chǎn)量進(jìn)行優(yōu)化控制,以滿足污水處理廠的部分能源需要.同時(shí)全流程建?？梢悦枋鑫鬯幚硐到y(tǒng)內(nèi)中磷參與的反應(yīng)過(guò)程,填補(bǔ)了磷管理在決策工具和設(shè)計(jì)方法方面的空缺.利用數(shù)學(xué)模擬工具對(duì)SRT和溫度等關(guān)鍵參數(shù)的影響進(jìn)行優(yōu)化,是提高鳥(niǎo)糞石回收率的重要方法.

6.2 展望

①不同的地區(qū),其進(jìn)水組分具有不同的特征,針對(duì)目標(biāo)污水處理廠進(jìn)行數(shù)學(xué)模擬的時(shí)候,應(yīng)該在前期進(jìn)行廣泛的調(diào)研以及監(jiān)測(cè),提升硬件設(shè)施水平,建立適合當(dāng)?shù)剡M(jìn)水組分監(jiān)測(cè)體系,并確保分析方法的準(zhǔn)確性和簡(jiǎn)單性.

②盡管利用數(shù)學(xué)模擬對(duì)溫室氣體進(jìn)行模擬具有明顯的優(yōu)勢(shì),但目前還是面臨著兩個(gè)問(wèn)題:一是如今污水廠全流程范圍內(nèi)的溫室氣體形成與排放機(jī)制研究得不夠完善,過(guò)于參數(shù)化,不同水質(zhì)和不同工藝類型的反應(yīng)器中N2O形成機(jī)制的研究需要加強(qiáng);二是數(shù)據(jù)的缺乏,溫室氣體模型的建立與驗(yàn)證需要大量的監(jiān)測(cè)數(shù)據(jù),這意味著需要污水行業(yè)具備更加先進(jìn)的自動(dòng)化監(jiān)測(cè)和處理數(shù)據(jù)的能力.

③數(shù)據(jù)的共享與數(shù)據(jù)庫(kù)平臺(tái)的建立有利于推動(dòng)污水處理領(lǐng)域中數(shù)學(xué)模擬技術(shù)的發(fā)展與升級(jí).但是污水處理領(lǐng)域仍處在傳統(tǒng)的階段,研究者往往很難獲得足夠的數(shù)據(jù)支撐他們的模擬研究,因此未來(lái)污水處理行業(yè)需要在數(shù)據(jù)開(kāi)源及共享方面作出更大的努力.

④污水管網(wǎng)中的厭氧環(huán)境及污水廠出水的受納水體同樣會(huì)排放一定量的溫室氣體,而污水處理廠作為中間環(huán)節(jié),未來(lái)的模擬方向應(yīng)該將三個(gè)環(huán)節(jié)集成并進(jìn)行大量的場(chǎng)景模擬,來(lái)探究三者之間的相互作用及對(duì)環(huán)境的整體影響.

⑤整合文獻(xiàn)中已有的溫室氣體模型,建立一個(gè)利于高效長(zhǎng)期統(tǒng)計(jì)溫室氣體信息的集成模型,最終建立一個(gè)決策支持系統(tǒng),并結(jié)合技術(shù)與經(jīng)濟(jì)分析為污水處理廠提供創(chuàng)新性的運(yùn)營(yíng)指南.

⑥在推動(dòng)實(shí)現(xiàn)減污降碳協(xié)同增效的道路上,未來(lái)可借助機(jī)器學(xué)習(xí)深入分析污染物指標(biāo)與溫室氣體排放的關(guān)聯(lián)機(jī)制,提取特征信息,為相關(guān)碳排放核算技術(shù)指南、規(guī)范的制定提供參考,逐步建立起治污與控碳協(xié)調(diào)的排放標(biāo)準(zhǔn);此外,多目標(biāo)優(yōu)化應(yīng)研究不同污泥處置技術(shù)路線對(duì)碳排放及能量平衡的影響,應(yīng)對(duì)污水處理和污泥處置同步進(jìn)行全生命周期的碳足跡評(píng)估,開(kāi)發(fā)更多與污泥處置路徑相關(guān)的決策變量.

⑦針對(duì)不同類型的進(jìn)水條件,對(duì)污水處理系統(tǒng)能耗評(píng)估過(guò)程的經(jīng)驗(yàn)參數(shù)進(jìn)行測(cè)定和比較,同時(shí)利用數(shù)學(xué)模型合理開(kāi)發(fā)污水處理系統(tǒng)的其他資源潛力(氮元素、硫元素、熱能等).

[1] 彭 玉,王建輝,齊高相,等.活性污泥模型(ASMs)研究進(jìn)展及其發(fā)展前景[J]. 應(yīng)用化工, 2020,49(5):1288-1292.

Peng Y, Wang J H, Qi G X, et al. Research progress and development prospect of activated sludge models (ASMs) [J]. Applied Chemical Industry, 2020,49(5):1288-1292.

[2] Borzooei S, Campo G, Cerutti A, et al. Optimization of the wastewater treatment plant: From energy saving to environmental impact mitigation [J]. Science of The Total Environment, 2019,691:1182- 1189.

[3] Liu H, Zhang Y, Zhang H. Prediction of effluent quality in papermaking wastewater treatment processes using dynamic kernel- based extreme learning machine [J]. Process Biochemistry, 2020, 97:72-79.

[4] Fernández-Arévalo T, Lizarralde I, Fdz-Polanco F, et al. Quantitative assessment of energy and resource recovery in wastewater treatment plants based on plant-wide simulations [J]. Water Research, 2017, 118:272-288.

[5] 徐承志,操家順,羅景陽(yáng),等.活性污泥數(shù)學(xué)模型在污水處理中的研究進(jìn)展[J]. 應(yīng)用化工, 2021,50(5):1341-1347.

Xu C Z, Cao J S, Luo J Y, et al. Research progress of activated sludge mathematical model in sewage treatment [J]. Applied Chemical Industry, 2021,50(5):1341-1347

[6] 郭亞萍,陳士才,張宇坤,等.ASMs仿真軟件在污水處理廠中的應(yīng)用[J]. 中國(guó)給水排水, 2010,26(14):23-25.

Guo Y P, Chen S C, Zhang Y K, et al. Application of ASMS simulation software in sewage treatment plant [J]. China Water& Wastewater, 2010,26(14):23-25

[7] Huang R, Ma C, Ma J, et al. Machine learning in natural and engineered water systems [J]. Water Research, 2021,205:117666.

[8] Sweetapple C, Fu G, Butler D. Multi-objective optimisation of wastewater treatment plant control to reduce greenhouse gas emissions [J]. Water Research, 2014,55:52-62.

[9] Miao S, Zhou C, Alqahtani S A, et al. Applying machine learning in intelligent sewage treatment: A case study of chemical plant in sustainable cities [J]. Sustainable Cities and Society, 2021,72:103009.

[10] Chen W, Dai H, Han T, et al. Mathematical modeling and modification of a cycle operating activated sludge process via the multi-objective optimization method [J]. Journal of Environmental Chemical Engineering, 2020,8(6):104470.

[11] Torres C I, Marcus A K, Parameswaran P, et al. Kinetic experiments for evaluating the Nernst? Monod model for anode-respiring bacteria (ARB) in a biofilm anode [J]. Environmental Science &Technology, 2008,42(17):6593-6597.

[12] Acuna-Askar K, Englande Jr A J, Hu C, et al. Methyl tertiary-butyl ether (MTBE) biodegradation in batch and continuous upflow fixed- biofilm reactors [J]. Water Science and Technology, 2000,42(5/6): 153-161.

[13] McKinney R E. Mathematics of complete-mixing activated sludge [J]. Journal of the Sanitary Engineering Division, 1962,88(3):87-113.

[14] Lawrence A W, McCarty P L. Kinetics of methane fermentation in anaerobic treatment [J]. Water Pollution Control Federation, 1969: R1-R17.

[15] Andrews J F. Dynamic models and control strategies for wastewater treatment processes [J]. Water Research, 1974,8(5):261-289.

[16] James A. Investigations of sewage discharges to some British coastal waters. Chap. 8, Bacterial mortality, Pt. 3 [J]. Marine Pollution Bulletin, 1987,18(6):299-299.

[17] 邵 袁.基于WEST軟件模擬的城市污水廠的優(yōu)化運(yùn)行研究[D]. 南京:東南大學(xué), 2019.

Shao Y. Research on optimal operation of urban sewage plant based on West software simulation [D]. Nanjing: Southeast University, 2019.

[18] Dold P L, Marais G V R. Evaluation of the general activated sludge model proposed by the IAWPRC task group [J]. Water Science and Technology, 1986,18(6):63-89.

[19] Henze M, Gujer W, Mino T, et al. Wastewater and biomass characterization for the activated sludge model no. 2: biological phosphorus removal [J]. Water Science and Technology, 1995,31(2): 13-23.

[20] Henze M, Gujer W, Mino T, et al. Activated sludge model No.2D, ASM2D [J]. Water Science and Technology, 1999,39(1):165-182.

[21] Gujer W, Henze M, Mino T, et al. Activated sludge model No. 3 [J]. Water Science and Technology, 1999,39(1):183-193.

[22] Goel R, Mino T, Satoh H, et al. Comparison of hydrolytic enzyme systems in pure culture and activated sludge under different electron acceptor conditions [J]. Water Science and Technology, 1998,37(4/5): 335-343.

[23] Nasr M S, Moustafa M A, Seif H A, et al. Modelling and simulation of German BIOGEST/EL-AGAMY wastewater treatment plants-Egypt using GPS-X simulator [J]. Alexandria Engineering Journal, 2011,50 (4):351-357.

[24] Beaupré M, Rieger L, Vanrolleghem P A, et al. Implementation of a dynamic cost calculation module for Avedore WWTP using WEST [J]. Dept. of Industrial Electrical Engineering and Automation Lund University, Québec, 2008.

[25] Lei L, Gharagozian A, Start B, et al. Process alternative comparisons assisted with biowin modeling [J]. Proceedings of the Water Environment Federation, 2006,(9):3274-3289.

[26] Pedersen J. Controlling activated sludge process using EFOR [J]. Water Science and Technology, 1992,26(3/4):783-790.

[27] 袁良松,謝長(zhǎng)煥,鄧 科.試論ASM水質(zhì)特性參數(shù)測(cè)定方法[J]. 能源環(huán)境保護(hù), 2007,(5):1-3.

Yuan L S, Xie C H, Deng K. Discussion on the determination method of ASM water quality characteristic parameters [J]. Energy Environmental Protection, 2007,(5):1-3.

[28] Dold P L, Marais G V R, Ekama G A. Procedures for determining influent COD fractions and the maximum specific growth rate of heterotrophs in activated sludge systems [J]. Water Science & Technology, 1986,18(6):91.

[29] Tschui M, Siegrist H. Interpretation of experimental data with regard to the Activated Sludge Model No. 1and calibration of the model for municipal wastewatertreatment plants [J]. Water Science & Technology, 1992,25(6):167.

[30] Henze M. Characterization of wastewater for modelling of activated sludge processes [J]. Water Science & Technology, 1992,25(6):1.

[31] Melcer H. Methods for wastewater characterization in activated sludge modelling: WERF Report (Project 99-WWF-3) [R]. Iwa Publishing, 2004.

[32] 郝二成,周 軍,等.活性污泥2號(hào)模型中進(jìn)水COD組分確定方法研究[J]. 給水排水, 2008,34(4):32-36.

Hao E C, Zhou J, et al. Study on determination method of influent cod composition in activated sludge model 2 [J]. Water supply and drainage, 2008,34(4):32-36.

[33] 艾海男,張代均,何 強(qiáng),等.基于ASM_2的快速易生物降解COD組分表征方法構(gòu)建[J]. 環(huán)境工程學(xué)報(bào), 2011,5(9):2005-2008.

Ai H N, Zhang D J, He Q, et al. Based on ASM_ construction of rapid and easily biodegradable COD component characterization method [J]. Journal of environmental engineering, 2011,5(9):2005-2008.

[34] Orhon D, okgr E U. COD fractionation in wastewater characterization—the state of the art [J]. J. Chem. Technol. Biotechnol., 1997,68(3):283-293.

[35] 池春榕,董姍燕,李齊佳,等.活性污泥數(shù)學(xué)模型研究進(jìn)展[J]. 有色金屬科學(xué)與工程, 2017,8(4):111-117.

Chi C R, Dong S Y, Li Q J, et al. Research progress of activated sludge mathematical model [J]. Nonferrous MetalsScience and Engineering, 2017,8(4):111-117.

[36] 孫培德,王如意.全耦合活性污泥模型(FCASM3)Ⅰ:建模機(jī)理及數(shù)學(xué)表征[J]. 環(huán)境科學(xué)學(xué)報(bào), 2008,(12):2404-2419.

Sun P D, Wang R Y. Fully coupled activated sludge model (fcasm3) Ⅰ: modeling mechanism and mathematical characterization [J]. Acta Scientiae Circumstantiae, 2008,(12):2404-2419.

[37] Kim H, Hao O J. SBR system for phosphorus removal: ASM2 and simplified linear Model [J]. Journal of Environmental Engineering, 2001.

[38] Bahar S, Ciggin A S. A simple kinetic modeling approach for aerobic stabilization of real waste activated sludge [J]. Chemical Engineering Journal, 2016,303:194-201.

[39] Ni B J, Yu H Q. An approach for modeling two-step denitrification in activated sludge systems [J]. Chemical Engineering Science, 2008, 63(6):1449-1459.

[40] Man Z, Gong J, Yang C, et al. Simulation of the performance of aerobic granular sludge SBR using modified ASM3 model [J]. Bioresource Technology, 2012,127C:473-481.

[41] Hebb D O. The Organization of behavior; A Neuropsychological Theory [J]. The American Journal of Psychology, 1950,63(4):633.

[42] Mitchell T M. Machine Learning [M]. Machine Learning, 2003.

[43] 耿麗娟,李星毅.用于大數(shù)據(jù)分類的KNN算法研究[J]. 計(jì)算機(jī)應(yīng)用研究, 2014,31(5):1342-1344.

Geng L J, Li X Y. Research on KNN algorithm for big data classification [J] Application Research of Computer, 2014,31(5): 1342-1344.

[44] 丁世飛,齊丙娟,譚紅艷.支持向量機(jī)理論與算法研究綜述[J]. 電子科技大學(xué)學(xué)報(bào), 2011,40(1):2-10.

Ding S F, Qi B J, Tan H Y. Overview of support vector machine theory and algorithm [J] Journal of University of Electronic Science and Technology of China, 2011,40(1):2-10.

[45] Shi T, Horvath S. Unsupervised learning with random forest predictors [J]. Journal of Computational and Graphical Statistics, 2006,15(1): 118-138.

[46] Clark J W. Neural network modeling [J]. Physics in Medicine & Biology, 1991,36(10):1259.

[47] Adeli H, Karim A. Fuzzy-wavelet RBFNN model for freeway incident detection [J]. Journal of transportation Engineering, 2000,126(6): 464-471.

[48] Cho K, Van Merri?nboer B, Gulcehre C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation [J]. arXiv preprint arXiv:1406.1078, 2014.

[49] Graves A, Schmidhuber J. Framewise phoneme classification with bidirectional LSTM and other neural network architectures [J]. Neural networks, 2005,18(5/6):602-610.

[50] Liu W, Wang Z, Liu X, et al. A survey of deep neural network architectures and their applications [J]. Neurocomputing, 2017, 234:11-26.

[51] 盧宏濤,張秦川.深度卷積神經(jīng)網(wǎng)絡(luò)在計(jì)算機(jī)視覺(jué)中的應(yīng)用研究綜述[J]. 數(shù)據(jù)采集與處理, 2016,31(1):1-17.

Lu H T, Zhang Q C. Review on the application of deep convolution neural network in computer vision [J] Journal of Data Acquisition & Processing, 2016,31(1):1-17.

[52] Mowbray M, Savage T, Wu C, et al. Machine learning for biochemical engineering: A review [J]. Biochemical Engineering Journal, 2021,172: 108054.

[53] Zhong S, Zhang K, Bagheri M, et al. Machine Learning: New Ideas and Tools in Environmental Science and Engineering [J]. Environmental Science & Technology, 2021,55(19):12741-12754.

[54] 王麗萍,任 宇,邱啟倉(cāng),等.多目標(biāo)進(jìn)化算法性能評(píng)價(jià)指標(biāo)研究綜述[J]. 計(jì)算機(jī)學(xué)報(bào), 2021,44(8):1590-1619.

Wang L P, Ren Y, Qiu Q C, et al. Review on performance evaluation indexes of multi-objective evolutionary algorithms [J]. Journal of Computer Science, 2021,44(8):1590-1619.

[55] Koopmans T C. An Analysis of Production as an efficient combination of activities [J]. Analysis of Production and Allocation, 1951.

[56] Bhoskar M T, Kulkarni M O K, Kulkarni M N K, et al. Genetic Algorithm and its Applications to Mechanical Engineering: A Review [J]. Materials Today: Proceedings, 2015,2(4):2624-2630.

[57] 李紅梅.多目標(biāo)優(yōu)化演化算法研究綜述[J]. 現(xiàn)代計(jì)算機(jī)(專業(yè)版), 2009,(4):44-46.

Li H M. Review of multi-objective optimization evolutionary algorithm [J]. Modern computer (Professional Edition), 2009,(4):44- 46.

[58] Schaffer J D. Some experiments in machine learning using vector evaluated genetic algorithms [Z]. Vanderbilt Univ., Nashville, TN (USA), 1985.

[59] Zitzler E, Thiele L. Multiobjective evolutionary algorithms: a comparative case study and the strength Pareto approach [J]. IEEE transactions on Evolutionary Computation, 1999,3(4):257-271.

[60] Srinivas N, Deb K, Roy P K, et al. Comparative study of vector evaluated GA and NSGA applied to multiobjective optimization, 1995 [C]. 1995.

[61] 羅健旭,彭培培,徐 穎,等.污水處理過(guò)程的多目標(biāo)優(yōu)化[J]. 西安交通大學(xué)學(xué)報(bào), 2017,51(3):129-135.

Luo J X, Peng P P, Xu Y, et al. Multi objective optimization of sewage treatment process [J]. Journal of Xi'an Jiaotong University, 2017, 51(3):129-135.

[62] Zhang R, Xie W, Yu H, et al. Optimizing municipal wastewater treatment plants using an improved multi-objective optimization method [J]. Bioresource Technology, 2014,157:161-165.

[63] 李 宸.基于多目標(biāo)優(yōu)化的污水廠升級(jí)改造綜合評(píng)價(jià)和工藝比選[D]. 哈爾濱:哈爾濱工業(yè)大學(xué), 2019.

Li C. Comprehensive evaluation and process comparison of sewage plant upgrading based on multi-objective optimization [D]. Harbin: Harbin University of technology,2019.

[64] Kim D, Bowen J D, Ozelkan E C. Optimization of wastewater treatment plant operation for greenhouse gas mitigation [J]. Journal of Environmental Management, 2015,163:39-48.

[65] Nopens, Benedetti, Jeppsson, et al. Benchmark Simulation Model No 2:finalisation of plant layout and default control strategy RID E-1784-2011 [J]. Water Science & Technology, 2010,2010,62(9)(-): 1967-1974.

[66] Chen K, Wang H, Valverde-Pérez B, et al. Optimal control towards sustainable wastewater treatment plants based on multi-agent reinforcement learning [J]. Chemosphere, 2021,279:130498.

[67] Yang C, Zhang Y, Huang M, et al. Adaptive dynamic prediction of effluent quality in wastewater treatment processes using partial least squares embedded with relevance vector machine [J]. Journal of Cleaner Production, 2021,314:128076.

[68] Cao J, Yang E, Xu C, et al. Model-based strategy for nitrogen removal enhancement in full-scale wastewater treatment plants by GPS-X integrated with response surface methodology [J]. Science of The Total Environment, 2021,769:144851.

[69] Vitanza R, Colussi I, Cortesi A, et al. Implementing a respirometry- based model into BioWin software to simulate wastewater treatment plant operations [J]. Journal of Water Process Engineering, 2016,9: 267-275.

[70] Hu H, Liao K, Xie W, et al. Modeling the formation of microorganism-derived dissolved organic nitrogen (mDON) in the activated sludge system [J]. Water Research, 2020,174:115604.

[71] Eldyasti A, Andalib M, Hafez H, et al. Comparative modeling of biological nutrient removal from landfill leachate using a circulating fluidized bed bioreactor (CFBBR) [J]. Journal of Hazardous Materials, 2011,187(1):140-149.

[72] Dorofeev A, Nikolaev Y, Kozlov M, et al. Modeling of anammox process with the biowin software suite [J]. Applied Biochemistry & Microbiology, 2017,53(1):78-84.

[73] Huang Z, Wu R, Yi X, et al. A novel model with GA evolving FWNN for effluent quality and biogas production forecast in a full-scale Anaerobic wastewater treatment process [J]. Complexity, 2019,2468189.

[74] Corominas L, Garrido-Baserba M, Villez K, et al. Transforming data into knowledge for improved wastewater treatment operation: A critical review of techniques [J]. Environmental Modelling &Software, 2018,106:89-103.

[75] Hamed M M, Khalafallah M G, Hassanien E A. Prediction of wastewater treatment plant performance using artificial neural networks [J]. Environmental Modelling & Software, 2004,19(10): 919-928.

[76] Seshan H, Goyal M K, Falk M W, et al. Support vector regression model of wastewater bioreactor performance using microbial community diversity indices: Effect of stress and bioaugmentation [J]. Water Research, 2014,53:282-296.

[77] Perendeci A, Arslan S, Tanyola? A, et al. Effects of phase vector and history extension on prediction power of adaptive-network based fuzzy inference system (ANFIS) model for a real scale anaerobic wastewater treatment plant operating under unsteady state [J]. Bioresource Technology, 2009,100(20):4579-4587.

[78] Bagherzadeh F, Mehrani M, Basirifard M, et al. Comparative study on total nitrogen prediction in wastewater treatment plant and effect of various feature selection methods on machine learning algorithms performance [J]. Journal of Water Process Engineering, 2021,41: 102033.

[79] Alejo L, Atkinson J, Guzmán-Fierro V, et al. Effluent composition prediction of a two-stage anaerobic digestion process: Machine learning and stoichiometry techniques [J]. Environmental Science and Pollution Research, 2018,25(21):21149-21163.

[80] Manu D S, Thalla A K. Artificial intelligence models for predicting the performance of biological wastewater treatment plant in the removal of Kjeldahl Nitrogen from wastewater [J]. Applied Water Science, 2017,7(7):3783-3791.

[81] Picos-Benítez A R, López-Hincapié J D, Chávez-Ramírez A U, et al. Artificial intelligence based model for optimization of COD removal efficiency of an up-flow anaerobic sludge blanket reactor in the saline wastewater treatment [J]. Water Science and Technology, 2017,75(6): 1351-1361.

[82] Jasim N A. The design for wastewater treatment plant (WWTP) with GPS X modeling [J]. Cogent Engineering, 2020,7(1):1723782.

[83] 魏忠慶,上官海東,葉均磊,等.基于GPS-X模擬的污水處理廠提標(biāo)工藝優(yōu)化[J]. 中國(guó)給水排水, 2018,34(19):81-84.

Wei Z Q, Shangguan H D, Ye J Li, et al. Optimization of standard raising process of sewage treatment plant based on gps-x simulation [J]. China Water & Wastewater, 2018,34(19):81-84.

[84] 周 振,胡大龍,吳志超,等.基于數(shù)學(xué)模型的多模式AAO系統(tǒng)運(yùn)行優(yōu)化研究[J]. 中國(guó)環(huán)境科學(xué), 2014,34(7):1734-1739.

Zhou Z, Hu D L, Wu Z C, et al. Research on operation optimization of multi-mode AAO system based on mathematical model [J]. China Environmental Science, 2014,34(7):1734-1739.

[85] Elawwad A, Matta M, Abo-Zaid M, et al. Plant-wide modeling and optimization of a large-scale WWTP using BioWin’s ASDM model [J]. Journal of Water Process Engineering, 2019,31:100819.

[86] 柳蒙蒙,陳梅雪,齊 嶸,等.面向寒冷地區(qū)城鎮(zhèn)污水處理廠提標(biāo)改造的ASM模擬優(yōu)化及其應(yīng)用[J]. 環(huán)境工程學(xué)報(bào), 2020,14(4):1119- 1128.

Liu M M, Chen M X, Qi R, et al. ASM simulation optimization and its application for upgrading and reconstruction of urban sewage treatment plants in cold areas [J]. Chinese Journal of Environmental Engineering, 2020,14(4):1119-1128.

[87] Molinos-Senante M, Hanley N, Sala-Garrido R. Measuring the CO2shadow price for wastewater treatment: a directional distance function approach [J]. Applied Energy, 2015,144:241-249.

[88] Longo S, D Antoni B M, Bongards M, et al. Monitoring and diagnosis of energy consumption in wastewater treatment plants. A state of the art and proposals for improvement [J]. Applied Energy, 2016,179: 1251-1268.

[89] Wang S, Zou L, Li H, et al. Full-scale membrane bioreactor process WWTPs in East Taihu basin: Wastewater characteristics, energy consumption and sustainability [J]. Science of The Total Environment, 2020,723:137983.

[90] Bagherzadeh F, Nouri A S, Mehrani M, et al. Prediction of energy consumption and evaluation of affecting factors in a full-scale WWTP using a machine learning approach [J]. Process Safety and Environmental Protection, 2021,154:458-466.

[91] 郝二成,郭 毅,劉偉巖,等.基于數(shù)學(xué)模擬的污水廠運(yùn)行分析——建模與體檢[J]. 中國(guó)給水排水, 2020,36(15):23-28.

Hao E C, Guo Y, Liu W Y, et al. Operation analysis of sewage plant based on mathematical simulation--modeling and physical examination [J]. China Water & Wastewater, 2020,36(15):23-28.

[92] 郝二成,郭 毅,劉偉巖,等.基于數(shù)學(xué)模擬的污水廠運(yùn)行分析——控制與優(yōu)化[J]. 中國(guó)給水排水, 2020,36(17):23-29.

Hao E C, Guo Y, Liu W Y, et al. Operation analysis of sewage plant based on mathematical simulation - control and optimization [J]. China Water & Wastewater, 2020,36(17):23-29.

[93] 邵 袁,王華成,覃榴濱,等.鄉(xiāng)鎮(zhèn)污水處理廠仿真模擬及優(yōu)化運(yùn)行研究[J]. 水處理技術(shù), 2020,46(9):90-97.

Shao Y, Wang H C, Qin L B, et al. Study on Simulation and optimal operation of township sewage treatment plants [J]. Technology of Water Treatment, 2020,46(9):90-97.

[94] Wu X, Yang Y, Wu G, et al. Simulation and optimization of a coking wastewater biological treatment process by activated sludge models (ASM) [J]. Journal of Environmental Management, 2016,165:235- 242.

[95] Rosso D, Stenstrom M K. Comparative economic analysis of the impacts of mean cell retention time and denitrification on aeration systems [J]. Water Research, 2005,39(16):3773-3780.

[96] Water E F U. Energy conservation in water and wastewater facilities [M]. New York: WEF Press, 2009.

[97] Olsson G. Water and energy: threats and opportunities [M]. London: IWA publishing, 2015.

[98] Lizarralde I, Fernández-Arévalo T, Beltrán S, et al. Validation of a multi-phase plant-wide model for the description of the aeration process in a WWTP [J]. Water Research, 2018,129:305-318.

[99] Asteriadis I, Azis K, Ntougias S, et al. A control strategy for an intermittently aerated and fed bioreactor to reduce aeration costs: A simulation study [J]. Biochemical Engineering Journal, 2021,173: 108081.

[100] Sean W, Chu Y, Mallu L L, et al. Energy consumption analysis in wastewater treatment plants using simulation and SCADA system: Case study in northern Taiwan [J]. Journal of Cleaner Production, 2020,276:124248.

[101] Holenda B, Domokos E, Rédey A, et al. Aeration optimization of a wastewater treatment plant using genetic algorithm [J]. Optimal Control Applications and Methods, 2007,28(3):191-208.

[102] Abbasi N, Ahmadi M, Naseri M. Quality and cost analysis of a wastewater treatment plant using GPS-X and CapdetWorks simulation programs [J]. Journal of Environmental Management, 2021,284: 111993.

[103] Rothausen S G, Conway D. Greenhouse-gas emissions from energy use in the water sector [J]. Nature Climate Change, 2011,1(4):210- 219.

[104] Edenhofer O. Climate change 2014: mitigation of climate change [M]. Cambridge University Press, 2015.

[105] Protocol G R. Accurate, transparent, and consistent measurement of greenhouse gases across North America Version 1.1 [Z]. May, 2008.

[106] Ni B, Ye L, Law Y, et al. Mathematical modeling of nitrous oxide (N2O) emissions from full-scale wastewater treatment plants [J]. Environmental Science & Technology, 2013,47(14):7795-7803.

[107] Boiocchi R, Gernaey K V, Sin G. A novel fuzzy-logic control strategy minimizing N2O emissions [J]. Water Research, 2017,123:479-494.

[108] Huang F, Shen W, Zhang X, et al. Impacts of dissolved oxygen control on different greenhouse gas emission sources in wastewater treatment process [J]. Journal of Cleaner Production, 2020,274:123233.

[109] Wang H, Yang Y, Keller A A, et al. Comparative analysis of energy intensity and carbon emissions in wastewater treatment in USA, Germany, China and South Africa [J]. Applied Energy, 2016,184:873- 881.

[110] Bani Shahabadi M, Yerushalmi L, Haghighat F. Impact of process design on greenhouse gas (GHG) generation by wastewater treatment plants [J]. Water Research, 2009,43(10):2679-2687.

[111] Carr M. Reducing greenhouse gas emissions industrial biotechnology and biorefining, 2007 [C]. 2007.

[112] Parravicini V, Svardal K, Krampe J. Greenhouse gas emissions from wastewater treatment plants [J]. Energy Procedia, 2016,97:246-253.

[113] Fan Y, Bai Y, Jiao W. Estimation of GHG Emissions from Water Reclamation Plants in Beijing [J]. Water Environment Research, 2016,88(9):795-802.

[114] de Haas D, Foley J, Barr K. Greenhouse gas inventories from WWTPs–the trade-off with nutrient removal[J]. Proceedings of the Water Environment Federation, 2008,2008(6):264-285.

[115] Zhang Z, Wang B. Research on the life-cycle CO2emission of China's construction sector [J]. Energy and Buildings, 2016,112:244-255.

[116] Nep O, Iea I. The IPCC Guidelines for National Greenhouse Gas Inventories. IPCC [R]. 2006.

[117] Mannina G, Ekama G, Caniani D, et al. Greenhouse gases from wastewater treatment — A review of modelling tools [J]. Science of The Total Environment, 2016,551-552:254-270.

[118] Kyung D, Kim M, Chang J, et al. Estimation of greenhouse gas emissions from a hybrid wastewater treatment plant [J]. Journal of Cleaner Production, 2015,95:117-123.

[119] Koutsou O P, Gatidou G, Stasinakis A S. Domestic wastewater management in Greece: Greenhouse gas emissions estimation at country scale [J]. Journal of Cleaner Production, 2018,188:851-859.

[120] 陳文亮,姚重華,呂錫武.活性污泥工藝的多目標(biāo)優(yōu)化分析[J]. 環(huán)境科學(xué)學(xué)報(bào), 2013,33(7):1918-1925.

Chen W L, Yao C H, Lv X W. Multi objective optimization analysis of activated sludge process [J]. Acta Scientiae Circumstantiae, 2013, 33(7):1918-1925.

[121] 周紅標(biāo),喬俊飛.混合多目標(biāo)骨干粒子群優(yōu)化算法在污水處理過(guò)程優(yōu)化控制中的應(yīng)用[J]. 化工學(xué)報(bào), 2017,68(9):3511-3521.

Zhou H B, Qiao J F. Application of hybrid multi-objective backbone particle swarm optimization algorithm in optimal control of sewage treatment process [J]. CIESC Journal, 2017,68(9):3511-3521.

[122] Flores-Alsina X, Arnell M, Amerlinck Y, et al. Balancing effluent quality, economic cost and greenhouse gas emissions during the evaluation of (plant-wide) control/operational strategies in WWTPs [J]. Science of The Total Environment, 2014,466-467:616-624.

[123] Sweetapple C, Fu G, Butler D. Multi-objective optimisation of wastewater treatment plant control to reduce greenhouse gas emissions [J]. Water Research, 2014,55:52-62.

[124] 付加鋒,馮相昭,高慶先,等.城鎮(zhèn)污水處理廠污染物去除協(xié)同控制溫室氣體核算方法與案例研究[J]. 環(huán)境科學(xué)研究, 2021, 34 (9):2086-2093.

Fu J F, Feng X Z, Gao Q X, et al. Accounting method and case study of greenhouse gas in collaborative control of pollutant removal in urban sewage treatment plant [J] Researchof Environmental Science, 2021,V.34;No.284(9):2086-2093.

[125] Guo Q, Dai X. Analysis on carbon dioxide emission reduction during the anaerobic synergetic digestion technology of sludge and kitchen waste: Taking kitchen waste synergetic digestion project in Zhenjiang as an example [J]. Waste Management, 2017,69:360-364.

[126] Diaz-Elsayed N, Rezaei N, Guo T, et al. Wastewater-based resource recovery technologies across scale: A review [J]. Resources, Conservation and Recycling, 2019,145:94-112.

[127] Crawford G. Energy Efficiency in Wastewater treatment in North America: a compendium of best practices and case studies of novel approaches: WERF Report OWSO4R07e (WERF Report S) [M]. IWA Publishing, 2011.

[128] Puchongkawarin C, Gomez-Mont C, Stuckey D C, et al. Optimization-based methodology for the development of wastewater facilities for energy and nutrient recovery [J]. Chemosphere, 2015, 140:150-158.

[129] Manyuchi M M, Mbohwa C, Muzenda E. Anaerobic treatment of opaque beer wastewater with enhanced biogas recovery through Acti-zyme bio augmentation [J]. South African Journal of Chemical Engineering, 2018,26:74-79.

[130] de Gracia M, Grau P, Huete E, et al. New generic mathematical model for WWTP sludge digesters operating under aerobic and anaerobic conditions: Model building and experimental verification [J]. Water Research, 2009,43(18):4626-4642.

[131] Ruffino B, Cerutti A, Campo G, et al. Thermophilic vs. mesophilic anaerobic digestion of waste activated sludge: Modelling and energy balance for its applicability at a full scale WWTP [J]. Renewable Energy, 2020,156:235-248.

[132] Sakiewicz P, Piotrowski K, Ober J, et al. Innovative artificial neural network approach for integrated biogas - wastewater treatment system modelling: Effect of plant operating parameters on process intensification [J]. Renewable and Sustainable Energy Reviews, 2020, 124:109784.

[133] Shu L, Schneider P, Jegatheesan V, et al. An economic evaluation of phosphorus recovery as struvite from digester supernatant [J]. Bioresource Technology, 2006,97(17):2211-2216.

[134] Fattah K P. Assessing struvite formation potential at wastewater treatment plants [J]. International Journal of Environmental Science and Development, 2012,3(6):548.

[135] Kazadi Mbamba C, Flores-Alsina X, John Batstone D, et al. Validation of a plant-wide phosphorus modelling approach with minerals precipitation in a full-scale WWTP [J]. Water Research, 2016,100:169-183.

[136] Lizarralde I, Fernández-Arévalo T, Manas A, et al. Model-based opti mization of phosphorus management strategies in Sur WWTP, Madrid [J]. Water Research, 2019,153:39-52.

Development and comprehensive application of mathematical simulation in sewage treatment system under the trend of carbon neutralization.

CHEN Zhi-chi1, HE Qiang1, CAI Ran2,4, LUO Hua-rui3, LUO Nan2, SONG Chen-xin4, CHENG Hong1*

(1.Key Laboratory of the Three Gorges Reservoir Region's Eco-Environment, State Ministry of Education, Chongqing University, Chongqing 400045, China；2.Beijing Capital Eco-Environment Protection Group Co., Ltd., Beijing 10000, China；3.Shenzhen Huanshui Investent Group Co., Ltd, Shenzhen 518031, China；4.Sichuan Shuihui Ecological Environment Management Co., Ltd, Neijiang 641000, China)., 202242(6)：2587～2602

Mathematical simulation technology (MST) has been widely applied in wastewater treatment, therefore, in order to systematically summarize these related technologies, this study reviewed the development of MST in sewage treatment system, and the application of activated sludge model (ASM) and machine learning (ML) in water quality prediction and parameter optimization; In addition, this paper mainly discussed the models of greenhouse gas emission in sewage treatment system, and the trade-off of multi-objective optimization model in sewage treatment system with the objectives of greenhouse gas emission (GHG), effluent quality (EQI) and operating cost (OCI). Furthermore, this paper also summarized the development of MST to achieve the energy self-sufficiency and resource recovery of sewage plant. The results from this study showed that MST can accurately predict the effluent quality, quickly optimize the process parameters, weigh the relationship among greenhouse gas emission, effluent quality and the operation cost, and improve the resource recovery efficiency. Overall, MST can effectively guide the operation optimization and management of sewage treatment process, and ultimately provide technical supports for the synergy of pollution reduction and carbon reduction in sewage treatment industry.

carbon neutralization；activated sludge model；machine learning；greenhouse gases；multi-objective optimization；resource recovery

X703

A

1000-6923(2022)06-2587-16

陳治池(1997-),男,重慶潼南人,重慶大學(xué)碩士研究生,主要從事水污染控制,機(jī)器學(xué)習(xí),污水工藝數(shù)值模擬優(yōu)化.

2021-11-22

重慶市技術(shù)創(chuàng)新與應(yīng)用發(fā)展專項(xiàng)(cstc2019jscx-tjsbX0002);住房和城鄉(xiāng)建設(shè)部科學(xué)技術(shù)計(jì)劃項(xiàng)目(2020-R-027);企業(yè)自主研發(fā)課題“基于數(shù)字化手段的流域水環(huán)境綜合治理系統(tǒng)研究1.0”

* 責(zé)任作者, 講師, hong.cheng@cqu.edu.cn