多参数影响下污水总氮浓度预测最优方法研究

发布时间：2019-03-27 08:30

【摘要】：污水总氮(TN)深度脱除是当前我国污水处理领域的重大科技需求.TN的去除受到多种环境及操作条件的影响,开发多参数条件下稳健的TN浓度预测方法是降低污水厂能耗、实现智能化控制的重要前提.针对以上问题,以某实际污水处理厂反硝化深床滤池为例,采用BP神经网络(BP)、量子遗传算法优化的BP神经网络(QGA_BP)、改进的QGA_BP和支持向量回归机(SVR),在进水流量和碳源投加量等13种变量条件下,对滤池出水TN进行了模拟预测.共选取147组数据,其中130组用于出水水质和工艺参数的拟合模拟,17组用于结果验证.将总氮实测值依次与BP,QGA_BP和改进的QGA_BP神经网络以及SVR预测结果进行对比,相关系数R2依次增大,分别为0.221,0.275,0.826和0.951,即预测值与实测值之间的拟合度逐渐升高.SVR克服了神经网络预测误差较大的问题,对多参数影响下TN浓度的预测具有较高的准确性和稳定性,用其替代常用的神经网络算法具有明显的优势.
[Abstract]:The deep removal of total nitrogen (TN) in wastewater is a major scientific and technological demand in the field of sewage treatment in China at present. TN removal is affected by a variety of environment and operating conditions. To develop a robust TN concentration prediction method under multi-parameter conditions is to reduce the energy consumption of wastewater treatment plants. The important premise of realizing intelligent control. In view of the above problems, taking the denitrification deep bed filter of a practical sewage treatment plant as an example, the BP neural network (BP), quantum genetic algorithm optimized BP neural network (QGA_BP), improved QGA_BP and support vector regression machine (SVR),) were used to optimize the BP neural network (QGA_BP). Under the conditions of 13 variables, such as influent flow and carbon source dosage, the effluent TN of the filter was simulated and predicted. A total of 147 sets of data were selected, of which 130 groups were used for fitting simulation of effluent quality and process parameters, and 17 groups were used to verify the results. Comparing the measured values of total nitrogen with BP,QGA_BP, improved QGA_BP neural network and SVR prediction, the correlation coefficients R2 are 0.221, 0.275, 0.826 and 0.951, respectively, and the correlation coefficients are 0.221, 0.275, 0.826 and 0.951, respectively. The prediction error of neural network is overcome, and the accuracy and stability of the prediction of TN concentration under the influence of multi-parameters is high. Using it to replace the common neural network algorithm has obvious advantages.
【作者单位】：南京大学环境学院污染控制与资源化研究国家重点实验室;江苏中宜金大环保产业技术研究院有限公司;南京大学宜兴环保研究院;
【基金】：国家水专项课题(2017ZX07204001) 江苏省重点研发计划项目(BE2017632) 江苏省科技成果转化专项资金项目(BA2016012) 中央高校基本科研业务费项目(021114380046)
【分类号】：TP18;X703

【相似文献】