【摘要】：隨著計算機(jī)高速度計算的發(fā)展,神經(jīng)網(wǎng)絡(luò)模型在動態(tài)環(huán)境評價管理中尤其是在污水處理廠運營管理和預(yù)測評價中的應(yīng)用逐漸成熟。由于其權(quán)重科學(xué)、客觀;能同時對多變量進(jìn)行有效地處理;具有很強(qiáng)的魯棒性、記憶力及非線性擬合能力,因此使用神經(jīng)網(wǎng)絡(luò)模型對污水處理廠運營管理進(jìn)行預(yù)測和評價,是對動態(tài)的多變量系統(tǒng)進(jìn)行評價和管理過程中常用的方法。本文綜合考慮水質(zhì)、環(huán)境、經(jīng)濟(jì)效益及監(jiān)督管理指標(biāo),基于深圳市污水處理廠現(xiàn)有評價體系進(jìn)行資料調(diào)研,獲取相關(guān)數(shù)據(jù)并通過回歸分析、殘差分析對數(shù)據(jù)進(jìn)行處理,逐步建立和優(yōu)化評價指標(biāo)體系;基于沙井污水處理廠的歷史數(shù)據(jù)和已建立的評價指標(biāo)體系初步建立BP神經(jīng)網(wǎng)絡(luò)模型;采用歷史數(shù)據(jù)對模型進(jìn)行訓(xùn)練、校驗及預(yù)測;最后通過逐層調(diào)整隱層數(shù)和節(jié)點數(shù)獲取優(yōu)化模型。對已建立的優(yōu)化模型進(jìn)行穩(wěn)定性校驗,最終獲取高穩(wěn)定性優(yōu)化模型:[6,6,7];然后獲取其權(quán)重矩陣并進(jìn)行顯著性分析,得出:單位污水絮凝劑消耗量、運營負(fù)荷率的權(quán)重值最大,分別為0.132和0.128,而在實際運營管理過程中對污水處理廠影響較大的出水COD、進(jìn)水流量的權(quán)重值卻很小;為實現(xiàn)實際運營管理中出水COD、進(jìn)水流量的權(quán)重值最大,需要逐一增減高穩(wěn)定性優(yōu)化模型的指標(biāo)變量、調(diào)節(jié)隱層數(shù)和節(jié)點數(shù),然后對其進(jìn)行穩(wěn)定性校驗,最終獲取高穩(wěn)定性深度優(yōu)化模型:[10,5,8,7]。通過比較模型的精準(zhǔn)度及預(yù)測能力,確定高穩(wěn)定性深度優(yōu)化模型為綜合評價模型;通過對比深圳市污水處理廠經(jīng)營評估模型和綜合評價模型影響因素的指標(biāo)權(quán)重,分析各個影響指標(biāo)對沙井污水處理廠運營管理的具體影響,最終從運營管理、節(jié)能減排等角度給政府和污水處理廠相關(guān)管理人員提供合理的政策建議。
[Abstract]:With the development of computer high speed calculation, the application of neural network model in dynamic environmental assessment management, especially in the operation management and prediction evaluation of sewage treatment plant has gradually matured. Because its weight is scientific, objective, it can deal with multivariable effectively at the same time, it has strong robustness, memory and nonlinear fitting ability, so the neural network model is used to predict and evaluate the operation and management of sewage treatment plant. It is a common method to evaluate and manage dynamic multivariable system. In this paper, water quality, environment, economic benefit and supervision and management index are considered synthetically. Based on the existing evaluation system of wastewater treatment plant in Shenzhen, the relevant data are obtained, and the data are processed by regression analysis and residual analysis. The evaluation index system is gradually established and optimized, the BP neural network model is established based on the historical data of manhole sewage treatment plant and the established evaluation index system, and the model is trained, calibrated and predicted by historical data. Finally, the optimization model is obtained by adjusting the number of hidden layers and the number of nodes layer by layer. The stability of the established optimization model is checked, and the high stability optimization model is obtained, and then the weight matrix is obtained and the significant analysis is carried out. It is concluded that the weight value of the consumption of flocculant per unit sewage and the operating load rate is the largest. In the process of actual operation and management, the weight value of effluent COD, influent flow is very small, and in order to realize the actual operation management, the weight value of effluent COD, influent flow is the largest. It is necessary to add and decrease the index variables of the high stability optimization model one by one, adjust the number of hidden layers and the number of nodes, and then check the stability of the model, and finally obtain the high stability depth optimization model: [10]. By comparing the accuracy and prediction ability of the model, the high stability and depth optimization model is determined as the comprehensive evaluation model, and the index weights of the influencing factors of the Shenzhen sewage treatment plant management evaluation model and the comprehensive evaluation model are compared. This paper analyzes the specific impact of each impact index on the operation and management of the sewage treatment plant of manhole, and finally provides reasonable policy recommendations to the government and the relevant management personnel of the sewage treatment plant from the aspects of operation management, energy saving and emission reduction.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號】：X703

【參考文獻(xiàn)】

相關(guān)期刊論文 前7條

1 邱紅霞;;基于熵權(quán)的灰色關(guān)聯(lián)模型在水電工程評價中的應(yīng)用[J];水科學(xué)與工程技術(shù);2010年02期

2 姚清晨;;COD和BOD_5相關(guān)關(guān)系及其在環(huán)境監(jiān)測中的應(yīng)用[J];太原科技;2009年05期

3 田妮莉;喻莉;;一種基于小波變換和FIR神經(jīng)網(wǎng)絡(luò)的廣域網(wǎng)網(wǎng)絡(luò)流量預(yù)測模型[J];電子與信息學(xué)報;2008年10期

4 程志強(qiáng);劉載文;王小藝;崔莉鳳;;基于粒子群算法的神經(jīng)網(wǎng)絡(luò)在污水處理優(yōu)化中的應(yīng)用[J];北京工商大學(xué)學(xué)報(自然科學(xué)版);2008年04期

5 錢曉東;王正歐;;ART2神經(jīng)網(wǎng)絡(luò)聚類的改進(jìn)研究[J];南京理工大學(xué)學(xué)報(自然科學(xué)版);2007年01期

6 劉洪波;張宏偉;;基于小波分解的城市供水管網(wǎng)短期水量負(fù)荷預(yù)測[J];中國給水排水;2006年17期

7 王國平,王洪光;集對分析用于污水處理廠的綜合評價[J];江蘇環(huán)境科技;2002年01期

相關(guān)碩士學(xué)位論文 前2條

1 孫艷玉;基于多級BP神經(jīng)網(wǎng)絡(luò)的無線電羅盤多故障診斷研究[D];鄭州大學(xué);2014年

2 朱文龍;基于遺傳算法的BP神經(jīng)網(wǎng)絡(luò)在多目標(biāo)優(yōu)化中的應(yīng)用研究[D];哈爾濱理工大學(xué);2009年

，

本文編號：2265064