污水处理出水COD软测量预测建模方法的研究

发布时间：2018-02-19 17:50

本文关键词： 污水处理神经网络支持向量机出水COD 融合　出处：《安徽工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,在国家经济发展的同时水资源污染问题变得也愈发严重,污水治理已经是社会急需解决的问题。因此,污水处理系统的研究与应用有着重要的意义。本文以安徽省某环保科技公司的污水处理为研究对象,研究污水出水COD参数的预测。污水处理的过程具有复杂性以及水体的变化呈现非线性和未知性等特性。传统的污水处理使用的ASM相关模型由于其考虑参数和过程较多,增加了测量参数的干扰因素,最终给所建预测模型带来不理想的效果。人工神经网络有着较好的容错性、自适应的能力以及可以进行大规模的数据并行计算,适用于多因素、多条件的场合,以及对数据的模糊处理。因此,将神经网络应用于对污水水质预测具有一定的优势,本文的研究内容主要包括:首先,针对软测量应用的优势,建立了神经网络污水处理软测量模型,并进行了仿真验证,针对模型预测效果的不足增加了基于BP算法及Elman算法的改进方法,使模型的准确度有一定的提升。其次,研究了支持向量机基本原理、核函数与参数的选定方法,并将支持向量机应用于污水的预测建模研究,选取了基于网格搜索优化以及粒子群优化两种方法分别对该模型进行参数优化,取得了良好的预测效果。最后,针对神经网络在建模时存在的局部极小值及辅助变量导致的矛盾数据问题,建立了基于支持向量机和改进神经网络相联合的预测模型。同时为了解决组合模型中单一神经网络之间存在的相关性问题,加入了主元回归(PCR)对子模型的输出进行数据加权融合。最终,构建的软测量模型在预测准确性上有较大的提升以及拥有较好的鲁棒性和泛化能力。本论文的贡献在于:建立了基于支持向量机与改进神经网络相组合的软测量出水COD模型,并仿真验证对比单神经网络模型以及支持向量机模型,验证了此方法的优良预测效果,实现了对出水COD数据更加精准的预测。
[Abstract]:In recent years, with the development of national economy, the problem of water resources pollution has become more and more serious. Sewage treatment has become a problem urgently needed to be solved by society. The research and application of sewage treatment system is of great significance. The prediction of COD parameters of sewage effluent is studied. The process of sewage treatment is complicated and the variation of water body is nonlinear and unknown. The traditional ASM model used in wastewater treatment is more concerned with parameters and processes. The disturbance factors of the measurement parameters are added, and the prediction model is finally brought unsatisfactory effect. The artificial neural network has better fault tolerance, adaptive ability and large scale data parallel computation, so it is suitable for many factors. Therefore, the application of neural network to the prediction of sewage water quality has some advantages. The main research contents of this paper include: firstly, aiming at the advantages of soft sensing application, The soft sensor model of sewage treatment based on neural network is established, and the simulation results are verified. The improved method based on BP algorithm and Elman algorithm is added to improve the accuracy of the model. The basic principle of support vector machine (SVM), the method of selecting kernel function and parameters are studied, and the SVM is applied to the prediction modeling of sewage. Two methods based on grid search optimization and particle swarm optimization are selected to optimize the parameters of the model, and good prediction results are obtained. In order to solve the problem of local minima and auxiliary variables in the modeling of neural network, A prediction model based on support vector machine and improved neural network is established. The principal component regression (PCR) is added for data weighted fusion of the output of the sub-model. Finally, The proposed soft-sensing model has better prediction accuracy, robustness and generalization ability. The contribution of this paper lies in the establishment of a soft-sensing water COD model based on support vector machine and improved neural network. Compared with the single neural network model and the support vector machine model, the simulation verifies the good prediction effect of this method, and realizes the more accurate prediction of the effluent COD data.
【学位授予单位】：安徽工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X832;TP18

【参考文献】