【作者】 迟明杰；

【导师】 綦星光；

【作者基本信息】 山东轻工业学院 ， 检测技术与自动化装置， 2010， 硕士

【摘要】 随着经济的发展和城市化进程的加快,水资源的污染程度也越来越严重。为加大对水资源的保护。近几年,我国建立了大量的污水处理厂。序列间歇式活性污泥(SBR)法作为一种利用自然界微生物生命活动来去除污水中的有机物以及脱氮除磷的有效方法,在新建污水处理厂中得到广泛应用。然而由于污水处理过程具有非线性、多变量、和大时滞的特点,因此运用传统的数学建模方法很难为污水处理过程建立精确的模型,从而在一定程度上制约了污水处理行业自动控制水平的提高。因此,加强污水处理过程建模及控制算法的研究,将有利于水处理行业的快速发展。本文从解析ASM1模型入手,较为全面的了解了污水处理过程的内部机理。在污水处理行业建模经验的基础上,分别用BP和RBF两种神经网络建立了两个可以预测出水水质中化学需氧量(COD)含量的预测模型。通过Matlab软件进行了仿真,并分析了选择不同隐含层神经元个数和不同隐含层激励函数对模型预测精度的影响。仿真结果显示:两种模型都能较好的完成对出水COD的预测,其中BP网络模型的相对误差在3%以内, RBF网络模型的最大相对误差达到了9%,但从训练时间上来看RBF网络要远小于BP网络。另外,从污水处理过程来看,溶解氧(DO)作为影响污水处理效果的重要参数,对它的控制效果直接决定了出水水质的好坏,因此对DO的控制将是污水处理过程中的控制重点。但是对于城市生活污水来讲,在不同的时间段,污水中各种成分的含量波动比较大,所以对DO的控制又是一个控制难点。针对传统PID控制算法在污水处理过程中控制效果不尽理想的现状,结合污水处理过程本身的特点。本文引入了带自修正因子的模糊控制算法,并建立了相应的模糊控制器,实现了对DO含量的控制。通过Matlab软件对普通模糊控制器和带自调整因子的模糊控制器进行了仿真,结果显示:带自修正因子的模糊控制器可以很好的实现控制要求。带自修正因子的模糊控制系统比普通的模糊控制系统有着更好稳定性、实时性和鲁棒性。更多还原

【Abstract】 With economic development and accelerated urbanization,the pollution of water resources are becoming increasingly serious. To increase the water resources protection, a large number of sewage treatment plants under construction. Activated sludge process is a kind of effective method, which consumes the substrates and phosphate release and denitrification in wastewater by microorganism’s metabolism. In the new sewage treatment plant, it also has been widely used. However, due to the sewage treatment process with the characteristics of nonlinear, multivariable, and large time delay, it is difficult to establish a precisely model for the sewage treatment process. To a certain extent, it restricts the control level of the sewage treatment process. Therefore, to enhance wastewater treatment process modeling and control algorithm, will be conducive to the rapid development of the water treatment industryIn this paper, from the analytical model of the ASM1 to understand the internal mechanism of the sewage treatment. On the basis of modeling experience in sewage treatment industry, has established two prediction models to predict COD. It was simulated by Matlab software, and analyzed the effects of the number of hidden layer neurons and different hidden layer activation function to the model prediction accuracy. Simulation results show that: both models can better predict the content of the effluent COD, and the relative error of BP network model is less than 3%, while the maximum relative error of the RBF network model achieves 9%, however, the RBF network training time is far less than the BP network.In addition, the dissolved oxygen (DO) is an important parameter of sewage treatment, it’s control result determines the water quality. Therefore, the control of the DO is the control focus of the sewage treatment process. But the components of urban sewage is Volatile at different time, so,it is difficult to control the DO very well. as the traditional PID control method does not work very well in sewage treatment process, the author introduces the fuzzy controller based on self- adjustment factor and established a control system for the DO concentration. The ordinary fuzzy control and with self-adjusting factor fuzzy control is simulated by Matlab software.The simulation results show that: with self-modifying factor fuzzy controller can achieve good control requirements. It has better stability, real-time and robustness than ordinary fuzzy control system.更多还原