【摘要】 针对污水处理运行过程的重要指标出水总磷(Total Phosphorous,TP)和出水氨氮(Ammonia Nitrogen,NH4-N)难以实时测量的问题,文中提出了一种基于模糊神经网络的多变量软测量方法。首先,利用主元分析法对污水处理过程运行数据进行分析,获得TP和NH4-N的相关主元变量;其次,设计了一种基于模糊神经网络的多输入多输出软测量方法,利用自适应二阶算法对模型参数进行调整,提高了软测量方法的精度;最后,将设计的软测量方法进行封装,并将其应用于污水处理过程中试平台。实验结果表明:基于模糊神经网络的软测量方法能够同时实现TP和NH4-N的实时测量,并且具有较好的测量精度。更多还原

【Abstract】 Effluent total phosphorus(TP) and effluent ammonia nitrogen(NH4-N) both are important indexes of wastewater treatment process. Due to the difficulties to measure these two variables online simultaneously, a multiple-variable soft-sensor method based on the fuzzy neural network(FNN), was proposed in this paper. Firstly, the principal component analysis method was used to analyze the operation data of wastewater treatment process to obtain the principal component variables of TP and NH4-N. Secondly, a multiple-input multiple-output(MIMO) soft-senor method, based on the FNN, was designed. Then, the parameters of the proposed soft-sensor method were adjusted by the adaptive second-order algorithm, improved the accuracy. Finally, the proposed soft-sensor method was encapsulated and applied to a real wastewater treatment plant. The results indicated that the FNN-based soft-sensor can predict TP and NH4-N simultaneously with suitable prediction accuracy.更多还原