壓縮包內(nèi)含有CAD圖紙和說(shuō)明書(shū),均可直接下載獲得文件，所見(jiàn)所得，電腦查看更方便。Q 197216396 或 11970985
本科生畢業(yè)設(shè)計(jì)（論文）答辯記錄
學(xué) 院 環(huán)境科學(xué)與工程學(xué)院 專(zhuān) 業(yè) 環(huán)境工程 答辯人
畢業(yè)設(shè)計(jì)（論文）題目250MW燃煤電廠煙氣除塵脫硫工程設(shè)計(jì) 記錄人
答辯小組成員：
說(shuō)明：1. 主要記錄答辯委員所提的問(wèn)題及答辯人對(duì)問(wèn)題的回答。
2. 本記錄與學(xué)生的畢業(yè)設(shè)計(jì)（論文）資料一同裝訂。
記 錄 內(nèi) 容
問(wèn)題1：布袋除塵器與電除塵器比較，你的設(shè)計(jì)中為什么選擇布袋除塵器除塵？
答：布袋除塵器除塵性能比電除塵的好，除塵效率也比電除塵高，且其處理氣量大，能耗低，運(yùn)行費(fèi)用少。
問(wèn)題2：廢氣除塵的方法？
答：重力沉降室、慣性除塵器和旋風(fēng)除塵器、濕式除塵器、電力除塵器、過(guò)濾式除塵器 (又命袋式除塵器）
問(wèn)題3：在廢氣處理過(guò)程中回收利潤(rùn)的估算是多少 ？
答：該系統(tǒng)每天能回收的石膏有8000kg,市場(chǎng)價(jià)格為0.23元/kg,則每天的收入是8000 kg×0.23元/kg＝1840元，一年就可賣(mài)得66萬(wàn)元。
布袋除塵每天回收5噸可作水泥生料得分塵，按水泥成本計(jì)算，每臺(tái)除塵器每年可創(chuàng)效益：5噸/天×60元/噸×360天＝10.8萬(wàn)元。
該除塵脫硫系統(tǒng)總投資328.05萬(wàn)元，而副產(chǎn)品石膏、煤分每年可創(chuàng)效76.8萬(wàn)元，大概5年可以回收成本。
任務(wù)書(shū)
題目名稱(chēng)
250MW燃煤電廠煙氣除塵脫硫工程設(shè)計(jì)
學(xué)生學(xué)院
環(huán)境科學(xué)與工程學(xué)院
專(zhuān)業(yè)班級(jí)
姓 名
學(xué) 號(hào)
一、畢業(yè)設(shè)計(jì)（論文）的內(nèi)容
燃煤電廠煙氣除塵脫硫工程設(shè)計(jì)，包括各種除塵脫硫的工藝原理、各種除塵脫硫的工藝方法比較、主體設(shè)備選型和非標(biāo)準(zhǔn)設(shè)備設(shè)計(jì)，管道輸送系統(tǒng)設(shè)計(jì)及工程投資概算等。
二、畢業(yè)設(shè)計(jì)（論文）的要求與數(shù)據(jù)
廢氣處理量：畢業(yè)實(shí)習(xí)收集，或者“按產(chǎn)排污系數(shù)手冊(cè)”；
廢氣成分：畢業(yè)實(shí)習(xí)收集，或者“按產(chǎn)排污系數(shù)手冊(cè)”；
畢業(yè)實(shí)習(xí)10天以上；實(shí)習(xí)報(bào)告（含資料調(diào)研報(bào)告）10000字以上；
??? 畢業(yè)設(shè)計(jì)說(shuō)明書(shū)30000字以上；
繪制工程設(shè)計(jì)圖紙8張（A4）以上。
三、畢業(yè)設(shè)計(jì)（論文）應(yīng)完成的工作
查閱和翻譯文獻(xiàn)資料；
參與畢業(yè)實(shí)習(xí)并編寫(xiě)實(shí)習(xí)報(bào)告；
編寫(xiě)畢業(yè)設(shè)計(jì)說(shuō)明書(shū)；
進(jìn)行工程概算和運(yùn)行可行性分析；
繪制工程設(shè)計(jì)圖紙。
序號(hào)
設(shè)計(jì)（論文）各階段內(nèi)容
起止日期
1
參與畢業(yè)實(shí)習(xí)
3月15日~4月12日
2
編寫(xiě)實(shí)習(xí)報(bào)告、查閱和翻譯文獻(xiàn)資料
4月13~4月25日
3
研究設(shè)計(jì)方案，進(jìn)行設(shè)計(jì)的有關(guān)計(jì)算
4月26日~5月10日
4
編寫(xiě)畢業(yè)設(shè)計(jì)說(shuō)明書(shū)
5月11日~5月25日
5
進(jìn)行工程概算和運(yùn)行可行性分析
5月26日~5月29日
6
繪制工程設(shè)計(jì)圖紙
5月30日~6月8日
7
答辯準(zhǔn)備及答辯
6月9日~6月12日
四、畢業(yè)設(shè)計(jì)（論文）進(jìn)程安排
五、應(yīng)收集的資料及主要參考文獻(xiàn)
1. 王志魁主編 . 化工原理 .第二版.北京：化學(xué)工業(yè)出版社，1998.10
2. 赫吉明 馬廣大主編 . 大氣污染控制工程. 第二版.北京：高等教育出版社，2002
3. 賀匡國(guó)主編.化工容器及設(shè)備簡(jiǎn)明設(shè)計(jì)手冊(cè).化學(xué)工業(yè)出版社，1989
4. 黃學(xué)敏.張承中主編. 大氣污染控制工程實(shí)踐教程.北京：化學(xué)工業(yè)出版社. 2003.9
5. 立本英機(jī).安部郁夫（日）主編.高尚愚譯編. 活性炭的應(yīng)用技術(shù)ü其維持管理及存在問(wèn)題.南京：東南大學(xué)出版社，2002.7
6. 林肇信主編.大氣污染控制工程.高等教育出版社.1991.5
7. 全燮.楊鳳林主編. 環(huán)境工程計(jì)算手冊(cè).中國(guó)石化出版社.2003.6
8. 吳忠標(biāo)主編 . 實(shí)用環(huán)境工程手冊(cè)ü大氣污染控制工程 化學(xué)工業(yè)出版社. 2001.9
9. 姜安璽主編. 空氣污染控制 .北京：化學(xué)工業(yè)出版社. 2003
10. 朗曉珍. 楊毅宏主編. 冶金環(huán)境保護(hù)及三廢治理技術(shù). 東北大學(xué)出版社. 2002
11. 童志權(quán)等主編. 工業(yè)廢氣污染控制與利用. 北京：化學(xué)工業(yè)出版社，1988
12. 王紹文.張殿印.徐世勤.董保澍主編. 環(huán)保設(shè)備材料手冊(cè).冶金工業(yè)出版社 2000.9
13. 朱世勇，《環(huán)境與工業(yè)氣體凈化技術(shù)》，化學(xué)工業(yè)出版社，2001
14. 李光超，《大氣污染控制技術(shù)》，化學(xué)工業(yè)出版社，2002
15. L.Ekman.LIFAC-經(jīng)濟(jì)有效的脫硫方法.芬蘭：Fortum Engineering Ltd.
16. 唐敬麟，張祿虎編. 除塵裝置系統(tǒng)及設(shè)備設(shè)計(jì)選用手冊(cè)化學(xué)工業(yè)出版社.2004
17. 《給水排水設(shè)計(jì)手冊(cè) （第11卷）》，中國(guó)建筑工業(yè)出版社，1986.
18. 趙毅，李守信，《有害氣體控制工程》，化學(xué)工業(yè)出版社，2001.
19. 陳常貴、曾敏靜、劉國(guó)雄等編，《化工原理》，天津科學(xué)技術(shù)出版社，2002
20. Licht,W《Air Pollution Control Engineering》.Publisher,New York,NY(US);Marcel Dekker,Inc.System Entry Date:2001 May 13.
21. Dry Removal of Gaseous Pollutants from Flue Gases with the GFB(FGD by CFB).Lurgi Report,Germany,1990..
22. 劉天齊主編，三廢處理工程技術(shù)手冊(cè)：廢氣卷，北京：化學(xué)工業(yè)出版社 1999.5
發(fā)出任務(wù)書(shū)日期：20xx年3月10日 指導(dǎo)教師簽名：
預(yù)計(jì)完成日期：20xx年6月12日 專(zhuān)業(yè)負(fù)責(zé)人簽章：
主管院長(zhǎng)簽章：
2
目 錄
1概述 1
2大氣污染預(yù)測(cè)方法 1
2.1空氣質(zhì)量模型 2
2.1.1箱模型 2
2.1.2高斯模型 2
2.1.3 K模型 2
2.2灰色理論模型 3
2.2.1 GM(1,1)模型 3
2.2.2 GM(1,N)模型 4
2.3投影尋蹤回歸模型 5
2.4模糊理論模型 6
2.5環(huán)境質(zhì)量計(jì)算模型 6
2.6統(tǒng)計(jì)理論方法 7
3大氣污染預(yù)測(cè)中應(yīng)注意的幾個(gè)問(wèn)題 7
4結(jié)論 8
參 考 文 獻(xiàn) 8
1 Overview 10
2 Air Pollution Forecast Method 11
2.1 Air Quality Model 11
2.1.1 A Box Model 11
2.1.2 Gaussian Model 12
2.1.3 K Model 12
2.2 Gray Theory Model 13
2.2.1 GM(1,1) Model 13
2.2.2 GM(1,N)Model 14
2.3 Projection Pursuit Regression Model 15
2.4 Fuzzy Theory Model 16
2.5 Environmental Quality Measurement Model 17
2.6 Statistical Theory Method 17
3 Some Problems Must be paid Attention To In Air Pollution Forecast 18
4 Conclusion 19
References 20
大氣污染預(yù)測(cè)方法探討
鄭博福 游海 弓曉峰 齊美富
(南昌大學(xué)環(huán)境與化學(xué)工程學(xué)院,江西南昌 330029)
摘要 綜述當(dāng)前常用的大氣污染預(yù)測(cè)方法,對(duì)空氣質(zhì)量模型、灰色理論模型、投影尋蹤回歸模型等3種主要的大氣污染預(yù)測(cè)方法進(jìn)行了詳細(xì)介紹,重點(diǎn)分析討論了這些方法的實(shí)用性和局限性,闡述了進(jìn)行大氣污染預(yù)測(cè)時(shí)應(yīng)注意的幾個(gè)問(wèn)題。
關(guān)鍵詞 大氣污染, 預(yù)測(cè), 模型
1 概 述
預(yù)測(cè)就是根據(jù)主觀的經(jīng)驗(yàn)和教訓(xùn)、客觀的資料與條件和演變的邏輯與推斷,尋求事物的發(fā)展規(guī)律,對(duì)事物未來(lái)發(fā)展趨勢(shì)和可能達(dá)到的水平做出估計(jì)和推斷。環(huán)境預(yù)測(cè)是以人口為中心,以社會(huì)經(jīng)濟(jì)預(yù)測(cè)和科學(xué)技術(shù)預(yù)測(cè)為基礎(chǔ),預(yù)計(jì)和推測(cè)隨著人口、經(jīng)濟(jì)、政策等社會(huì)因素的發(fā)展環(huán)境的變化趨勢(shì)，提出防止環(huán)境質(zhì)量惡化和改善環(huán)境質(zhì)量的對(duì)策,達(dá)到社會(huì)的發(fā)展與環(huán)境保護(hù)之間的協(xié)調(diào)與統(tǒng)一。
污染預(yù)測(cè)是環(huán)境預(yù)測(cè)的基礎(chǔ)和重要組成部分,污染預(yù)測(cè)的誤差大小直接影響著環(huán)境預(yù)測(cè)質(zhì)量。污染預(yù)測(cè)根據(jù)預(yù)測(cè)對(duì)象可以分大氣污染預(yù)測(cè)、水污染預(yù)測(cè)、土壤污染預(yù)測(cè)、噪聲污染預(yù)測(cè)等。大氣污染預(yù)測(cè)就是預(yù)測(cè)某一特定區(qū)域的大氣污染的未來(lái)變化趨勢(shì),并提出改善大氣環(huán)境質(zhì)量的對(duì)策,為決策部門(mén)在制定該區(qū)域大氣污染防治規(guī)劃與經(jīng)濟(jì)發(fā)展規(guī)劃時(shí)提供參考和依據(jù)。
2 大氣污染預(yù)測(cè)方法
目前,國(guó)內(nèi)外學(xué)者用于大氣污染預(yù)測(cè)的方法模型主要有:空氣質(zhì)量模型、灰色理論模型、投影尋蹤回歸模型、模糊理論模型、線型系統(tǒng)分析模型、環(huán)境質(zhì)量計(jì)量模型、統(tǒng)計(jì)理論方法等預(yù)測(cè)模型和方法。
2.1空氣質(zhì)量模型
常用于大氣污染預(yù)測(cè)的空氣質(zhì)量模型有:箱模型、高斯模型及K理論模型。
2.1.1箱模型
箱模型是一種最簡(jiǎn)單的城市空氣質(zhì)量模型。它把整個(gè)城市空間看作為一個(gè)或多個(gè)矩形的箱形式,其主要假設(shè)條件為:1)在一個(gè)箱體內(nèi),污染源(看作面源)的源強(qiáng)是一個(gè)常數(shù);2)污染物進(jìn)入箱體(大氣)后,立即在鉛直方向均勻分布。由于城市污染源分布比較均勻,鉛直擴(kuò)散速率較快,上述假設(shè)有一定的合理性。但是箱模型的假定與實(shí)際情況有很大差異,對(duì)近地面的濃度估算偏低。
2.1.2高斯模型
高斯模型是城市空氣質(zhì)量模型中最主要的應(yīng)用模型,因?yàn)?1)大多數(shù)平原城市及郊區(qū)的范圍在20～30km以?xún)?nèi),流場(chǎng)并不十分復(fù)雜;2)城市空氣質(zhì)量模型的誤差主要來(lái)源于模型輸入?yún)?shù),尤其是污染源資料并不可能十分準(zhǔn)確、精細(xì),使對(duì)模型本身的改進(jìn)歸于徒勞,從應(yīng)用的效果看,復(fù)雜數(shù)值模型并不優(yōu)于高斯模型;3)高斯模型對(duì)氣象資料的需求比其它空氣質(zhì)量模型對(duì)氣象資料的需求更低,而運(yùn)算效果卻明顯提高。
高斯模型具有簡(jiǎn)單實(shí)用、空間分辨率高的優(yōu)點(diǎn),但它有以下不足之處:1) 當(dāng)模擬的尺度達(dá)到幾十公里,或者因下墊面不均勻,使流場(chǎng)比較復(fù)雜時(shí),高斯煙流模型的精度就難以滿(mǎn)足要求;2)高斯模型的沉積和化學(xué)轉(zhuǎn)化過(guò)程只能十分粗略的處理,當(dāng)這些過(guò)程已相當(dāng)重要或者作為研究對(duì)象時(shí),高斯模型不適用。
2.1.3 K模型
該模型是由平流擴(kuò)散方程式經(jīng)各種簡(jiǎn)化假設(shè)而推導(dǎo)得出的,它具有如下效能:1)能夠模擬三維非定常流場(chǎng)中的輸送和擴(kuò)散,因此可以模擬復(fù)雜下墊面和較大的尺度范圍內(nèi)的空氣污染;2)污染源場(chǎng)可以任意給定,即Q=Q(x,y,z,t),3)邊界可以反射、吸收和穿透污染物質(zhì),其濃度在邊界上可變;4)可以模擬包括非線型化學(xué)反應(yīng)引起的濃度變化;5)可以模擬干、濕沉積引起的濃度的變化。
因?yàn)镵模型來(lái)源于模仿分子擴(kuò)散的梯度輸送假設(shè),它具有一定的局限性;1)梯度輸送假設(shè)要求滿(mǎn)足一定的尺度條件,使擴(kuò)散方程僅僅在煙流尺度大于占優(yōu)勢(shì)的湍渦尺度時(shí)才是正確的;2)對(duì)流條件下梯度-輸送關(guān)系不成立,可能出現(xiàn)反梯度輸送的現(xiàn)象,不能應(yīng)用K模型;3)K模型對(duì)基礎(chǔ)資料及輸入?yún)?shù)的要求很高。
空氣質(zhì)量模型的預(yù)測(cè)精度在很大程度上依賴(lài)于對(duì)污染源和氣象條件的預(yù)測(cè)精度,因此,比較適用于短時(shí)污染預(yù)測(cè),一般不用于長(zhǎng)期污染預(yù)測(cè)。
2.2灰色理論模型
由于灰色理論模型在建模、預(yù)測(cè)、控制等方面的獨(dú)到性,十多年來(lái)已在氣象、生態(tài)、環(huán)境、社會(huì)經(jīng)濟(jì)等領(lǐng)域得了廣泛應(yīng)用,常用于環(huán)境污染預(yù)測(cè)得灰色理論模型主要有:GM(1,1)模型和GM(1,N)模型。
2.2.1 GM(1,1)模型
該模型是對(duì)原始數(shù)據(jù)列|}作累加生成,弱化隨機(jī)性后建立的時(shí)間連續(xù)性微分方程,一般形式為
+ χ =μ (1)
式中: ,μ為參數(shù),可由最小二乘法解得。
離散響應(yīng)值為
(2)
此時(shí)是累加值,可以還原成預(yù)測(cè)值模型:
(3)
式中
(4)
模型建立后,k為定值
該模型相當(dāng)于指數(shù)模型y=aebx,是特定的指數(shù)曲線,形狀簡(jiǎn)單,具有單調(diào)性,其特點(diǎn)是能反映事物發(fā)展的趨勢(shì),但不能很好地?cái)M合擺動(dòng)過(guò)程,對(duì)變化幅度大、起伏大地?cái)?shù)據(jù)顯得無(wú)能為力。因而,該模型對(duì)數(shù)據(jù)的分布有一定的要求。
2.2.2 GM(1,N)模型
該模型是由N個(gè)變量組成的一階線型動(dòng)態(tài)時(shí)間連續(xù)微分方程,一般形式為
其中,是對(duì)原始數(shù)列作最后一次累加生成,即
記系數(shù)向量為,可由最小二乘求得,由此可得離散響應(yīng)值為
最后,對(duì)作一次還原生成,即
則就是未來(lái)預(yù)測(cè)值。
在用該模型進(jìn)行環(huán)境污染的預(yù)測(cè)時(shí),首先需要確定出與要預(yù)測(cè)的因子相關(guān)性較大的影響因子(如社會(huì)、經(jīng)濟(jì)因子等),在選取主要影響因子時(shí),一般要采用灰色關(guān)聯(lián)分析方法。由于該模型采用了影響因子的篩選,使環(huán)境預(yù)測(cè)與社會(huì)、經(jīng)濟(jì)等因素預(yù)測(cè)相結(jié)合,從而提高了預(yù)測(cè)的精確度和可信度。
2.3投影尋蹤回歸模型
投影尋蹤回歸模型是用來(lái)處理和分析高維非正態(tài)觀測(cè)數(shù)據(jù)的一種探索數(shù)據(jù)分析新方法。其基本思想是:把高維數(shù)據(jù)投影到低維空間上,通過(guò)極小化某個(gè)投影指標(biāo),尋找出能反映原高維數(shù)據(jù)結(jié)構(gòu)或特征的投影,以達(dá)到研究、分析高維數(shù)據(jù)的目的。
K階投影尋蹤自回歸(PPAR(k))模型的一般形式為
式中: xi為時(shí)間序列{x|在t時(shí)刻估計(jì)值,xi-1,xi-2…,xi-k為k個(gè)時(shí)序預(yù)測(cè)因子,其選取最終由數(shù)據(jù)結(jié)構(gòu)決定,為第m 個(gè)投影方向含量,滿(mǎn)足||||=1,是的最優(yōu)分段線性函數(shù),稱(chēng)為嶺函數(shù),它是一個(gè)數(shù)值函數(shù), 表示第m個(gè)嶺函數(shù)對(duì)xi的貢獻(xiàn)大小的權(quán)重系數(shù)。
尋找xi的表示方法是逐次找出最優(yōu)的:
模型的最小準(zhǔn)則是選取適當(dāng)?shù)膮?shù)求滿(mǎn)足下式的解：
該式中“E”為運(yùn)算符,表示“誤差的平方和的均值”。
K 階自回歸實(shí)際上是一個(gè)多變量輸入、單變量輸出的問(wèn)題。為了適合x(chóng)i與 xi-1,xi-2…,xi-k不呈線性關(guān)系的特點(diǎn)．PPAR(k)模型選取一系列嶺函數(shù)的“和”來(lái)逼近目標(biāo)函數(shù)的方法，即用增大 M 的方法來(lái)減少模型誤差，并且采用遙段線性函數(shù)在相應(yīng)各投影方向上不斷對(duì)觀測(cè)數(shù)據(jù)平滑逼近得到的數(shù)值函數(shù)。能使模型更客觀地反映數(shù)據(jù)本身的內(nèi)在結(jié)構(gòu)。從而增強(qiáng)預(yù)測(cè)結(jié)果的穩(wěn)定性。
2.4模糊理論模型
模糊理論知識(shí)特別是模糊聚類(lèi)分析、系統(tǒng)模型參數(shù)辨識(shí)以及模糊模型識(shí)別等知識(shí)廣泛用于環(huán)境預(yù)測(cè)中。
在大氣污染預(yù)測(cè)中,由于影響大氣污染物濃度的因素是多方面的,大氣污染物濃度的高低沒(méi)有明確的界限,具有模糊性,一般的單變量統(tǒng)計(jì)方法難以客觀反映各指標(biāo)與污染物濃度間的相關(guān)關(guān)系。而運(yùn)用模糊識(shí)別理論法,可以利用已有的實(shí)測(cè)資料,能計(jì)算得到相應(yīng)的最優(yōu)模糊分類(lèi)中心矩陣、最優(yōu)模糊劃分矩陣和指標(biāo)綜合權(quán)重矩陣等有用參數(shù),然后應(yīng)用最優(yōu)模糊識(shí)別理論進(jìn)行預(yù)測(cè),且可以隨著實(shí)測(cè)資料的增多和變化不斷進(jìn)行參數(shù)調(diào)整,使其能夠貼切地反映實(shí)際情況。
在進(jìn)行環(huán)境污染預(yù)測(cè)時(shí),模糊理論知識(shí)一般情況下都和其它預(yù)測(cè)模型結(jié)合使用,這樣能取得更好的預(yù)測(cè)效果。因?yàn)橛媚：垲?lèi)分析可以較好地把影響環(huán)境質(zhì)量的各種因素按主次區(qū)分開(kāi)來(lái),預(yù)測(cè)環(huán)境污染只需考慮必要的影響因素(主因素)而將次要因素略去,如果考慮的影響因素過(guò)多,不僅使計(jì)算量增大,還會(huì)得不到好的預(yù)測(cè)效果;如果考慮的影響因素過(guò)少,沒(méi)有把主要因素考慮進(jìn)去,則預(yù)測(cè)的結(jié)果是不可信的。
2.5環(huán)境質(zhì)量計(jì)算模型
環(huán)境質(zhì)量計(jì)量模型是通過(guò)對(duì)環(huán)境系統(tǒng)中大量隨機(jī)現(xiàn)象的研究，確定環(huán)境質(zhì)量與環(huán)境系統(tǒng)中各因素之間的相互作用形式，以數(shù)學(xué)模型來(lái)預(yù)測(cè)環(huán)境質(zhì)量變化趨勢(shì)，主要有回歸分析模型和 投入產(chǎn)出模型等幾種模型?；貧w分析所研究的數(shù)學(xué)模型主要是線性回歸和多項(xiàng)式回歸模型。對(duì)于環(huán)境質(zhì)量和影響環(huán)境質(zhì)量的幾個(gè)參數(shù)它們的k次觀察(試驗(yàn))數(shù)據(jù)是( )，x=1，2，…,k,那么，它們之間的關(guān)系可用下式表示：
式中是n+1個(gè)待估計(jì)的未知效，即回歸系數(shù)，可以用最小二乘法求得。在求線性回歸方程前，線性回歸模型只是一種假設(shè)，盡管這種假設(shè)常常不是沒(méi)有根據(jù)的，但在求出線性回歸方程后，還是需要對(duì)其進(jìn)行統(tǒng)計(jì)檢驗(yàn)，以給出肯定或者否定的結(jié)論。在模型的準(zhǔn)確性經(jīng)核實(shí)后,即可進(jìn)行預(yù)測(cè)和控制。
環(huán)境質(zhì)量計(jì)量模型中的投入產(chǎn)出法預(yù)測(cè)環(huán)境污染狀況是根據(jù)國(guó)家和各部門(mén)經(jīng)濟(jì)發(fā)展的規(guī)劃,結(jié)合環(huán)境可以忍受的污染水平來(lái)預(yù)測(cè)未來(lái)的環(huán)境狀況的。
2.6統(tǒng)計(jì)理論方法
用與環(huán)境預(yù)測(cè)的統(tǒng)計(jì)理論方法有時(shí)間序列法、自回歸分析法、多元線型回歸法、多元統(tǒng)計(jì)分析法、概率統(tǒng)計(jì)法、神經(jīng)網(wǎng)絡(luò)法、遞推預(yù)測(cè)法等,時(shí)間序列法和自回歸分析法是最常用的兩種統(tǒng)計(jì)理論預(yù)測(cè)方法。
時(shí)間序列法就是把說(shuō)明某一環(huán)境現(xiàn)象各個(gè)時(shí)期的統(tǒng)計(jì)資料,按時(shí)間先后進(jìn)行排列,從而得到這一環(huán)境現(xiàn)象依時(shí)間變化的動(dòng)態(tài)數(shù)列,通過(guò)對(duì)數(shù)列的變化分析,來(lái)預(yù)計(jì)未來(lái),這種方法比較簡(jiǎn)單易行,但由于它只依據(jù)過(guò)去資料,對(duì)環(huán)境變化有關(guān)的因素及未來(lái)可能的影響考慮較少,因此預(yù)測(cè)的精度會(huì)隨著預(yù)測(cè)時(shí)間的增長(zhǎng)而下降,在環(huán)境污染的中長(zhǎng)期很少使用這一預(yù)測(cè)方法。
自回歸分析法就是把給出的測(cè)量數(shù)據(jù)x1,x2,…,xn看作隨機(jī)過(guò)程x(t)的一個(gè)樣本函數(shù),通過(guò)對(duì)現(xiàn)實(shí)xi的分析,估計(jì)過(guò)程x(t)的總體特征,預(yù)測(cè)x(t)未來(lái)取值的概率分布,從而給出t>tn時(shí)x(t)的預(yù)報(bào)值。由于環(huán)境問(wèn)題中時(shí)間序列的復(fù)雜性,一般來(lái)說(shuō),質(zhì)量因素或污染因素不可能用一個(gè)完全的數(shù)學(xué)函數(shù)給出來(lái),但可以用一個(gè)概率分布函數(shù)組給出x(t)未來(lái)取值狀況的統(tǒng)計(jì)描述,因此適于這種描述的自回歸分析法也可以用于環(huán)境預(yù)測(cè)。
3大氣污染預(yù)測(cè)中應(yīng)注意的幾個(gè)問(wèn)題
1) 在進(jìn)行大氣污染預(yù)測(cè)時(shí),由于污染源數(shù)量多、排放方式不同,且污染物擴(kuò)散、遷移的影響因素較多,大氣污染狀況的各種影響因素之間又相互聯(lián)系和制約,作用機(jī)理不十分清楚;由于客觀環(huán)境是一個(gè)復(fù)雜的系統(tǒng),在目前,任何計(jì)算模型和計(jì)算方法都不能把所有的影響因素反映出來(lái),因此,如何確定不同類(lèi)型污染源及大氣污染狀況的主要影響因素,是大氣污染源預(yù)測(cè)中必需要注意的問(wèn)題。
2) 要進(jìn)行區(qū)域大氣污染的長(zhǎng)期預(yù)測(cè),人口、資源、經(jīng)濟(jì)、技術(shù)等因素對(duì)環(huán)境質(zhì)量的影響是不可忽略的[9]。環(huán)境預(yù)測(cè)應(yīng)以人口預(yù)測(cè)為中心,以社會(huì)經(jīng)濟(jì)預(yù)測(cè)和科學(xué)技術(shù)預(yù)測(cè)為基礎(chǔ),如果單純以污染物濃度的原始資料來(lái)預(yù)測(cè)環(huán)境的未來(lái)發(fā)展趨勢(shì)，不考慮社會(huì)、經(jīng)濟(jì)、政策等影響因素，是不現(xiàn)實(shí)、不可信的。所以，必須重視分析社會(huì)、經(jīng)濟(jì)、政策等因素的發(fā)展趨勢(shì)以及它們對(duì) 環(huán)境質(zhì)量的影響。
3) 當(dāng)前，對(duì)大氣污染進(jìn)行預(yù)測(cè)的一些方法還未擺脫 “從某些假定出發(fā)，按照一定的準(zhǔn)則,進(jìn)行模擬，作出預(yù)測(cè)”這種格式，它們?cè)诃h(huán)境預(yù)測(cè)中的應(yīng)用受到一定的限制。對(duì)于不同的實(shí)際情況，一定要選取適合該情況的預(yù)測(cè)模型，以充分利用這些預(yù)測(cè)模型的優(yōu)點(diǎn)，克服它們的缺點(diǎn)，提高預(yù)測(cè)的精度和可信度。所以，如何針對(duì)不同情況，科學(xué)地選取預(yù)測(cè)模型和建立新的預(yù)測(cè)模型是值得重視和迫切需要解決的問(wèn)題。
4結(jié)論
近些年來(lái)，環(huán)境預(yù)測(cè)正由單目標(biāo)向多目標(biāo)、單環(huán)境要素向多環(huán)境要素、靜態(tài)影響向動(dòng)態(tài)影響的方向發(fā)展，進(jìn)行預(yù)測(cè)時(shí)不再是僅僅考慮污染源、氣象條件等歷史資料，而是要同時(shí)考慮預(yù)測(cè)區(qū)域內(nèi)人口、經(jīng)濟(jì)、技術(shù)、政策等在內(nèi)的宏觀影響因素的變化對(duì)環(huán)境質(zhì)量的影響。環(huán)境預(yù)測(cè)正向科學(xué)化、系統(tǒng)化、計(jì)算機(jī)化、精確度高、可信度高的方向發(fā)展。
在區(qū)域大氣污染長(zhǎng)期預(yù)測(cè)方面，目前還缺乏系統(tǒng)、可靠、成熟的頂級(jí)測(cè)試方法。由于環(huán)境質(zhì)量狀況與人口、經(jīng)濟(jì)、技術(shù)、政策等因素密切相關(guān)，它們之間相互作用的機(jī)理非常復(fù)雜，人們?cè)谶M(jìn)行區(qū)域環(huán)境長(zhǎng)期預(yù)測(cè)時(shí)．往往只對(duì)這些影響因素和環(huán)境質(zhì)量的關(guān)系進(jìn)行定性的分析和描述，因而,預(yù)測(cè)效果一般都不太令人滿(mǎn)意。所以，在進(jìn)行區(qū)域大氣污染長(zhǎng)期預(yù)測(cè)時(shí)．在污染源、 氣象條件、人口、經(jīng)濟(jì)等歷史資料調(diào)查和現(xiàn)狀調(diào)查的基礎(chǔ)上，認(rèn)真考慮人口、經(jīng)濟(jì)、技術(shù)、政策等社會(huì)因素對(duì)大氣污染狀況的影響并定量地分析這種關(guān)系是非常重要的。
參 考 文 獻(xiàn)
[1] 中國(guó)環(huán)境科學(xué)研究院．城市大氣污染總量控制方法手冊(cè)[s] ．北京：中國(guó)環(huán)境科學(xué)出版社,1991．
[2] 劉思峰,郭天撈．灰色系統(tǒng)理論及其應(yīng)用[M]．鄭州：河南大學(xué)出版社,1991．
[3] 童恒慶．經(jīng)濟(jì)回歸模型及計(jì)算[M] ．武漢：潮北科學(xué)技術(shù)出版社,1997。
[4] 李柞泳,鄧新民．PPAR(k)技術(shù)在NOx 濃度預(yù)測(cè)中的應(yīng)用[J]．環(huán)境科學(xué)研究，1997，10(5)；32～34．
[5] 張寶泉等．計(jì)算機(jī)與環(huán)境多因紊分析[M] ．北京；中國(guó)環(huán)境科學(xué)出版社，1993．
[6]熊德琪,陳守煜．城市大氣污染物濃度預(yù)測(cè)模糊識(shí)別理論與模型 [J] ．環(huán)境科學(xué)學(xué)報(bào),1993,13(4)：482～490．
[7]黃國(guó)和．應(yīng)用多元統(tǒng)計(jì)分析理論預(yù)測(cè)城市大氣污染[j] ．環(huán)境科學(xué),1991，12(2);29-34．
[8]安鴻志,陳兆國(guó)等。時(shí)間序列的分析與應(yīng)用[M]。北京:科學(xué)出版社,1983。
[9]程水源,蒲思奇，劉德政等．區(qū)域大氣環(huán)境預(yù)測(cè)確定場(chǎng)區(qū)長(zhǎng)度的新方法[J] ．環(huán)境科學(xué)進(jìn)展．1995，3(3)；66～71．
[10]Finzi G．A．Mathematical Mode for Air Pollution Forecast and Alerm the Urban Area Arm[J]。Environ,1982．16(9)：2055--2059．
[11] 鮑強(qiáng)．環(huán)境科學(xué)進(jìn)展[J] ．1996，4(1)：1～18．
[12]David．P．Chock，Sandra L Winkler Air Quality Prdeictions Using a Fixed Layer—Depth Vetical Structure in the Urban Airshed Mode [J] ．Environ Sc iTechno1,1997，31;359～370．
An Approach on the Models for Air Pollution Forecast
Bofu Zheng Hai You Xiaofeng Gong Meifu Qi
(Environmental and Chemical Engineering School，Nanchang University, Nanchang
330029, China)
ABSTRACT This paper summarized some methods which were often applied in air pollution forceast，introduced three mainly used models of air pollution forecast in detail, such as air quality model, gray theory model and projection pursuit regression model，analysed the practice and limitness of those models，formulated some problems which would be paid attention to on forecasting air pollution.
KEY WORDS air pollution, forecast，model
1 Overview
Forecast is based on subjective experience, lessons and objective information on conditions with the evolution of the logic and inference. Seek the law of the things. Make estimates and assumptions to the things for the development of the future development trend and a level may reach．Environment forecasts is based on population projections for the center, and socio-economic forecasts and projections for the science and technology infrastructure. Forecasting and concluding the trend of environment as population, economic, social factors developing, bringing forward the policies to prevent deterioration of environmental quality and improve environmental quality measures, making social development and environmental protection harmonization and unification.
Pollution forecast is the basis and an important component of the Environmental forecast, pollution prediction errors have a direct impact on the environmental forecast quality.According to forecasts pollution target, we can divided pollution forecasts into air pollution forecasts, water pollution, soil pollution forecasts, noise pollution forecast．Atmospheric pollution is forecasting air pollution trends for the future in a particular area, and bringing forward some strategies to improve air quality, also providing a reference basis. to decision makers to work out the regional air pollution control planning and economic development planning.
2 Air Pollution Forecast Method
At present, what domestic and foreign scholars use for air pollution prediction models approach are : air quality model, a gray model, projection pursuit regression model, fuzzy model, linear model analysis, environmental quality measurement model methods of statistical theory and methods of forecasting model.
2.1 Air Quality Model
What are commonly used in atmospheric pollution forecast models are : a box model, Gaussian model and K model.
2.1.1 A Box Model
A box model is the simplest model of urban air quality. It look the entire space as the city of one or more rectangular box form, the main assumptions are: 1) in a box, sources of pollution (as sources) is a constant; 2) If went into the box , pollutants (air) immediately distribute in the vertical direction. Because urban sources of pollution distributed more equitably and vertically proliferate at a faster rate, the above assumptions are certainly reasonable. However, there is a big difference between a box model assumptions and the actual situation, as in the near ground the concentration estimates is low.
2.1.2 Gaussian Model
Gaussian model of urban air quality is the most important model for the application model because : 1) Areas of most of the plains cities and suburbs is within the 20 ~30km,and the flow field is not very complicated; 2) Errors of urban air quality model mainly come from the model input parameters, especially sources of information may not be very accurate, precise, so that the improvement of model itself is attributable to in vain, from the effects of application , complex numerical model is not better than the Gaussian model; 3) Gaussian model needs little weather information than other air quality model, but the result was markedly improved.
Gaussian model is simple and practical, high spatial resolution of the merits, but it has the following shortcomings : 1) When the simulation scale reaches tens of kilometers, or has an uneven surface, more complex flow field, precision of Gaussian plume model will be difficult to meet the requirements; 2) Gaussian model for the deposition and chemical transformation process is very rough, When these processes are important or as a research object, the Gaussian model is not applicable.
2.1.3 K Model
The model consists of advection - diffusion equation is simplified by various assumptions derivation, it has the following performance : 1)It can simulate transmission and proliferation in three-dimensional unsteady flow field, so it can simulate complex surface and larger scale of air pollution;2) Pollution sources fields can be given arbitrarily, that is Q=Q (x, y, z, t); 3) Borders have reflection, absorption and penetration of pollutants, its concentration on the border can change; 4) It can simulate non-linear chemical reaction caused by the concentration; 5) It can simulate dry and wet deposition of concentration changes.
Because K model for molecular diffusion copied from the gradient transfer assumption, it has certain limitations; 1) gradient carrier must meet the requirements of the assumption that the standard conditions diffusion equation so only smoke scale is dominant at the eddy scale that it is correct; 2)When convection conditions gradient-carrier relationship is not established, it may appear counter - gradient transport phenomena .It will be not applied to K model; 3) K model has a high demand of basic information and input parameters .
Prediction accuracy of air quality model is largely dependent on the sources of pollution and forecast accuracy of weather conditions, therefore, it is more applicable to short-term pollution forecasts generally than forecast for the long-term contamination.
2.2 Gray Theory Model
Gray theory model is unique as theoretical model in modeling, forecasting, and control ect. For this 10 years ,it is widely used in meteorology, ecology, the environment, social and economic fields , gray model which are commonly used in the pollution forecast are : GM (1,1) and GM (1, N) model.
2.2.1 GM(1,1) Model
The model is result of the cumulative production for the original data presentation |}, weakening the randomness of time after the establishment of continuity equation, and the general form is:
+ χ =μ (1)
in the formula: ,μ is parameters,which can be soluted by Least Square.
Discrete Response Value is
(2)
Here is accumulated value,which can be revert predictive value Model.
(3)
in the formula
(4)
After the model, the fixed value is k .
The model is equivalent to exponential model y = aebx ,which has a specific exponential curve, and the shape is simple and monotonous. Its characteristics is able to reflect the trend of development of things, but not a good fit for a Swing process of the magnitude of changes, it is powerless undulating land data. Thus, the model of data distribution have certain requirements.
2.2.2 GM(1,N)Model
The model consists of N variables in a dynamic time-continuous linear differential equations, the general form is
Thereinto, is the original series for a final cumulative production, that is
Modulus vetor is ,which can be soluted by Least Square.Thus,
Discrete Response Value is
At last, makea revert building, that is
So is the future forecast value.
Using the model of pollution prediction, firstly,we need to identify and predict the factors associated with greater impact factor (such as social and economic factors, etc.) When choosing the main factor affecting the selection, generally using gray relational analysis. As the model uses an impact factor screening, environmental prediction and social and economic factors combine forecast, enhances the result accuracy and credibility.
2.3 Projection Pursuit Regression Model
Projection pursuit regression model is used for the processing and analysis of high dimensional non-normal data to explore a new method for data analysis. The basic idea is : Make high-dimensional data projection to the low-dimensional space, through a projection of a very small target .Find out the projection which would reflect the original high-dimensional data structures or features of the projector to achieve the purposes for studying and analysising high-dimensional data.
The general form of k rank projection pursuit regression model (PPAR(k))model is
in the formula:xi is Time series {x| estimated value when the time is t , xi-1,xi-2…,xi-k is the k predictors factor of timing, its final selection by the data structure, is the m content in the direction of the projector to meet||||=1. is the optimal piecewise linear function of , known as Ling function, and it is a numerical function, and shows that is the m Ling-function,which has an contribution to xi of the size of the weight coefficients.
Find the method of showing xi is to find out the optimal :
The least model selection criteria is to select appropriate parameters to
work out the following formula:
"E" in the formula is the operator, said the "error of the mean square."
K- factorial autoregressive is actually problem with a multi - variable input, single-output variables . To be fit for not linear characteristics of xi and xi-1,xi-2…,xi-k., PPAR (k) model selectes a series of ridge function " summation " to make a approximation to objective function, that is, increase M to reduce the amount of the model errors. Also use linear function of the distance to make a constant observation data on the projection right direction to get the numerical approximation functions smoothly. It can reflect the internal data structure of itself more objectively. Thereby enhance the stability of the results of forecasts.
2.4 Fuzzy Theory Model
Fuzzy theory, especially fuzzy clustering analysis, Parameter identification system model and fuzzy model identification extensive knowledge is used widely in the environment forecast.
In atmospheric pollution forecast, because the concentration of air pollutants are affected by many factors, there is no clear boundaries in atmospheric concentrations of the pollutants level, and it is blurred. General univariate statistical methods is hard to reflect the objective correlated relation between indicators and the pollutant concentration. When recognizing and using fuzzy theory, we can utilize the existing datas to calculate the optimal fuzzy classification center matrix, Optimal fuzzy partition matrix indicators and comprehensive weight matrix and other useful parameters, then use the optimal fuzzy recognition theory to forecast ,also with the measured data increased and the ongoing changes in parameter adjustment, it can precisely reflect the actual situation.
When making an environmental pollution forecast, fuzzy theoretical knowledge is used under normal circumstances by combining other forecasting models, which will achieve a better prediction. Because fuzzy clustering analysis can be used to reduce the impact of environmental quality to prioritize factors.When making an environmental pollution forecast we only consider the necessary environmental impact factor (the main factors) not a secondary factor. If considering the impact of too many factors, not only increase the amount of calculation, but also be not a good forecast results. If the consideration of factors is too few, not only the main factors is taken into account, but also the projected results are unbelievable.
2.5 Environmental Quality Measurement Model
Environmental quality measurement model studys large of random phenomenon in environment system to identify environmental quality and environmental system interaction factors forms by a mathematical model to predict the trend of the environment quality changes, and there are mainly regression analysis model and the input-output model and other models. The mathematical model which Regression analysis study is mainly the linear and polynomial regression model. Some parameters of environmental quality and environmental quality isand their k observations (test) data is (), x = 1, 2, ..., k ,so that the relationship between them can be given by :
In the formule, is n+1 yet to be the unknown effect, that is the regression coefficient which can be worked out by using the method of least squares. Before working out linear regression equation, the linear regression model is only a hypothesis, although this assumption is often not unfounded. When the linear regression equation is worked out, we need to make statistical tests to give a positive or negative conclusion. If the accuracy of the model is verified, we can predict and control it.
The method input and output forecasting the contaminated status about Environmental quality measurement model is based on the environmental situation of countries and sectors of economic development planning, integrating environment can tolerate pollution levels to predict the future state of the environment.