【作者】 余颖；

【导师】 乔俊飞；

【作者基本信息】 北京工业大学 ， 模式识别与智能系统， 2005， 硕士

【摘要】 活性污泥法是利用自然界微生物的生命活动来清除污水中有机物和脱氮除磷的一种有效方法。活性污泥法污水处理过程是一个动态的多变量、强耦合过程,具有时变、高度非线性、不确定性和滞后等特点,过程建模相当困难。为保证处理过程运行良好和提高出水质量,开发精确、实用的动态模型已成为国内外专家学者普遍关心的问题。此外,由于污水处理过程是一个复杂的生化反应过程,现场试验不仅时间长且成本很高,因此,研究对污水处理过程的建模和仿真技术具有十分重要的现实意义。本文在深入分析现有研究成果的基础上,对活性污泥法污水处理过程的智能建模方法及仿真技术进行了研究。主要做了以下几方面工作: (1)在充分了解活性污泥法污水处理过程的现状及工艺流程的基础上,深入分析了现有的几种建模的方法,并且给出该过程建模和仿真技术的发展方向; (2)研究了基于BP 和RBF 神经网络的小型活性污泥法污水处理系统建模方法,证明了神经网络建立污水处理过程模型是有效可行的;研究结果表明:①采用RBF 网络建立的模型,具有方法简单、学习速度快等优点;②RBF 网络能很快逼近训练数据,收敛速度明显快于BP 网络;③BP 网络的泛化能力好于RBF网络; (3)提出了基于PCA-减法聚类-ANFIS 的活性污泥法污水处理过程建模方法,并且与基于PCA-BP 神经网络污水处理模型相比,具有更好的拟合能力及收敛速度。(4)采用LabVIEW 软件平台开发了活性污泥法污水处理过程仿真软件,初步实现了曝气池和二沉池的反应机理。本文的研究对我国污水处理过程的模型发展和仿真技术的开发有一定借鉴意义,但由于条件和时间的限制,尚有许多方面有待改进和深入探讨。更多还原

【Abstract】 Activated sludge process is a kind of effective method, which consumes the substrates and phosphate release and denitrification in wastewater by microorganism’s metabolism. Activated sludge process is a multi-variables、strong coupling system, which has characteristic of time varying、high non-linearity、indetermination and lag, and difficult to build process model. For assuring treatment process run well and advancing effluent quality, it has already turned into the universal careful problem by many researchers of the scientific world to study the precise and applied dynamic model. Besides, due to wastewater process, as a complicated biochemistry reaction process, its’locale experiments need long time and are expensive. So it has especially essential realistic value to study the modeling and simulation technology of wastewater process. In the paper, on the basis of thorough analyzing past achievements, the author studied intelligent modeling and simulation technology of activated sludge wastewater treatment process. The main content of this paper is as follows: (1)On the basis of understanding the actuality of wastewater process and technics, the paper analyzed thorough present modeling methods, and pointed out the development direction of modeling and simulation technology of the process; (2) The paper studied the model method of minitype activated sludge system by BP and RBF neural networks, and testified these wastewater treatment models by neural network are effective; The results show that: ①The RBF neural network model has characteristic of simple method、rapid study speed, etc; ②The RBF neural network model has better convergence ability and impending speed than the BP neural network model; ③The BP neural network model has better generalization than the RBF neural network model; (3) The paper put forward the modeling method of activated sludge process on the basis of PCA-subtractive clustering-ANFIS network. And comparing with PCA-BP wastewater treatment model, it has better convergence ability and更多还原