基于人工神经网络建立污水处理的软测量模型.doc

摘　要为解决目前日趋严重的水体环境污染问题，实现污水处理过程中水质参数的有效检测是十分重要的。其中，对于出水总磷(TP)测量方法的研究，是目前国际上的热点问题。由于污水处理过程具有机理复杂、高度耦合、与高度非线性等特点，传统的检测仪表往往无法提供及时、准确的测量结果。此外，由于TP的检测方法的复杂性，其检测仪表不仅价格昂贵，提供的检测结果通常具有滞后性(超过15分钟)。针对上述缺陷，本文提出通过搭建TP软测量模型的方法，实现TP的实时准确测量。在众多软测量建模方法的研究中，基于人工神经网络搭建软测量模型的方法得到了广泛的关注。但是，在污水处理工业中，基于神经网络的软测量模型依然存在测量精度不高以及无法进行在线修正等缺陷。针对此问题，本文提出了一种基于径向基神经网络(RBFNN)的TP软测量模型，该模型可以实现TP的在线测量，并获得较高的测量精度。本文主要工作内容如下：1.搭建污水处理中TP的软测量模型。以温度、厌氧末端ORP、好氧前端DO、好氧前端TSS以及出水PH作为输入变量，出水TP作为输出变量，建立一种多输入单输出(MISO)的TP软测量模型；2.提出了一种自组织RBF神经网络以及新的学习算法，并使用真实数据对网络进行训练。经实验结果可证明，本算法可以有效的缩短软测量模型的预测时间，并能够提高模型的预测精度；3.将该软测量模型嵌入至LabVIEW中，开发了前台显示界面，使软测量模型的预测效果更具有可视性。关键词：污水处理；出水总磷；软测量；自组织RBF神经网络AbstractTo solve the increasing water pollution problems,?it is important to achieve effective detection of water quality parameters in the sewage treatment process. Among all the water quality parameter measuring methods, the measuring methods of effluent total phosphorus (TP), is currently the hot issues in the world. Due to the mechanism of wastewater treatment process is complex, highly coupled, and highly nonlinear, the traditional detection instruments are often unable to provide timely and accurate measurement results. In addition, because of the complexity of the detection method of TP, the test instrument is not only expensive, but also provide the test results with a lag (more than 15 minutes). According to the above defects, this paper proposes a method of constructing TP soft measurement model to realize the real-time and accurate measurements of TP.In many of the soft measurement modeling method study, the method of?building soft measurement model, which based on artificial neural network?has been widely concerned. But in the sewage treatment industry, the soft measurement model based on neural network still remains the problems which measurement accuracy is not high and could not be online correction of defects.To solve these problems, this paper proposes a TP soft measurement model which based on RBF neural network (RBFN