引用本文:蒙西,张寅,乔俊飞.基于动态模糊神经网络的出水含氮参数软测量方法[J].控制理论与应用,2024,41(12):2383~2392.[点击复制]
MENG Xi,ZHANG Yin,QIAO Junfei.Soft-sensing method for effluent nitrogen parameters based on a dynamic fuzzy neural network[J].Control Theory and Technology,2024,41(12):2383~2392.[点击复制]
基于动态模糊神经网络的出水含氮参数软测量方法
Soft-sensing method for effluent nitrogen parameters based on a dynamic fuzzy neural network
摘要点击 3350  全文点击 46  投稿时间：2022-07-26  修订日期：2024-08-28
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2023.20667
  2024,41(12):2383-2392
中文关键词  城市污水处理过程  模糊神经网络  分级更新  出水含氮量  软测量
英文关键词  municipal wastewater treatment process  fuzzy neural network  hierarchical updating  effluent nitrogen concentration  soft-sensing
基金项目  国家自然科学基金项目(61903012, 622731013, 61890930–5, 62021003), 科技创新2030—“新一代人工智能”重大项目(2021ZD0112301, 2021ZD 0112302), 国家重点研发计划项目(2019YFC1906004–2)资助.
作者单位E-mail
蒙西* 北京工业大学信息学部 mengxi@bjut.edu.cn 
张寅 北京工业大学信息学部  
乔俊飞 北京工业大学信息学部  
中文摘要
      针对城市污水处理过程出水氨氮(NH4+-N)和出水总氮(TN)难以实时准确检测的问题, 文中提出了一种基于动态模糊神经网络(DFNN)的出水含氮参数软测量方法. 首先, 采用自组织增删机制和快速二阶学习算法构建模糊神经网络(FNN), 以快速获得结构精简的软测量模型; 其次, 引入自适应激活强度阈值设计FNN分级更新策略, 确保软测量模型在非平稳环境下的预测精度; 最后, 通过基准仿真1号模型(BSM1)平台的数据验证了DFNN软测量方法的有效性, 实验结果表明, 所提出的方法能够实现出水NH4+-N和出水TN的在线精准检测.
英文摘要
      Aiming at the real-time and accurate measurements of the effluent ammonium nitrogen (NH4+-N) and the effluent total nitrogen (TN) in municipal wastewater treatment process, a soft-sensing method for effluent nitrogen parameters based on a dynamic fuzzy neural network (DFNN) is proposed in this paper. First, by utilizing a self-organizing growing-and-pruning mechanism and an improved second-order learning algorithm, a fuzzy neural network (FNN) is constructed in order to obtain a soft-sensing model with a simplified structure. Then, by introducing an adaptive firing strength threshold, a hierarchical updating strategy of FNN is designed, which can effectively ensure the prediction accuracy of the soft-sensing model under non-stationary environments. Finally, the effectiveness of the proposed DFNN soft-sensing method is verified based on the simulation data which were provided by the benchmark simulation model No. 1 (BSM1) platform. The simulation results show that the proposed soft-sensing method can achieve online and accurate measurements of the effluent NH4+-N and the effluent TN.