基于BP算法污水处理报告.docVIP

模式识别期末作业 —基于BP算法分类器 ROC曲线 Abstract：With the development of the industry, the lack of water resources, environmental protection and energy saving requirements, sewage recycling has become increasingly conspicuous. The characteristics of the wastewater treatment process is a multivariable, nonlinear, time-varying and randomness, so it is difficult to establish precise mathematical model. At the same time, the control of sewage treatment system and belongs to the multi-objective control, the need to control several water indicators, and also needs to restrain the change of external environment. Artificial neural network (ANN) is the biggest advantage is can approximate any complex nonlinear relations, fully has the strong learning ability and fault tolerance, able to handle the quantitative and qualitative data at the same time, can use the connection structure combined with other control methods and artificial intelligence, has become a research focus in the sewage treatment process. The purpose of this paper is to adopt the BP model of artificial neural network of a sewage treatment plant of wastewater treatment system was simulated, in order to provide theoretical basis for water quality evaluation. 摘要:随着工业的发展、水资源的缺乏、环保及节能的需求,污水循环利用已经变得日益引人注目。污水 处理过程的特点是多变量、非线性、时变性与随机性,故建立精确的数学模型比较困难。同时,污水处理系统的控制又属于多目标控制,需要控制几种出水指标,也需要抑制外部环境的变化。人工神经网络(ANN)的最大优点是可以充分逼近任意复杂的非线性关系,有较强的学习能力和容错性,同时能够处理定量、定性数据,能够利用连接的结构与其他控制方法及人工智能相结合,已成为污水处理过程的研究热点。 本文旨在采用人工神经网络的BP模型对某污水处理厂的污水处理系统进行模拟,以便对出水水质评价提供理论依据。 简介 人工神经网络方法是建立在现代神经科学研究，成果基础上的一种抽象的数学模型,它反映了大脑功能的若干基本特征,如抽象和模拟。使机器具有人脑那样的感知、学习和推理能力,能够更好地适应环境和系统特性的变化,适合于复杂系统的建模和控制。它具有大规模并行、分布式存储与处理、自组织、自适应与自学习的能力,特别适用于处理需要同时考虑许多因素和条件的、不精确的和模糊的信息处理问题。这使得它在模式识别、控制优化、智能信息处理以及故障诊断等方面都有广泛的应用。近年来,人工神经网络在水资源系统、饮用水及污水处理系统、环境治理系统等水处理系统中广泛应用,主要应用于臭氧生物活性炭系统的建模;废水处理中出水水质评价。BP人工神经网络通常是指基于误差反向传播算法的多层前向神经网络。BP网络在某些方面显示出独特的优越性和广泛的推广性,已成为日前研究应用最为广泛的神经网络学习算法,据统计近90%的神经网络应用是基于BP网络算法的。与感知器和线性神经网络等其他网络不同的是,BP网络的神经元采用的传递函数通常是Sigmoid型可微函数,所以可