摘要：

针对序批式生物膜系统难以构建水质模型的问题,采用神经网络技术进行建模方法研究。根据拉伊达准则剔除异常数据,并用训练样本调整网络连接权值,用检验样本实时动态监控训练过程,用LM算法构建了一个7-12-3结构的BP神经网络模型。将模型输出结果与实测数据进行比较,其相关系数分别为R_COD=0.857,R_NH4⁺-N=0.918,R_PO4^3--P=0.942,能够满足污水处理过程建模的要求。

关键词: 序批式生物膜反应器, 神经网络, 建模

Abstract:

It is difficult to build the model of sequencing batch biofilm reactor.This problem has been studied and solved by using the neural network technique.The 7-12-3 back-propagation neural network technique is developed for the system with excluding abnormal data according to pauta criterion,adjusting the network connection weights by training samples,monitoring the training process timely with test samples and the LM algorithm.The model output result being compared with actually measured data,the coefficient of COD is 0.857,ammonia is 0.918,and phosphate is 0.942,meeting the modeling requirement of sewage treatment process.

Key words: sequencing batch biofilm reactor, neural network, modeling

中图分类号: 

• X703